chore: drop accidentally-committed __pycache__/ and gitignore .pyc

The previous commit (5fac9c39) shipped sheets_df.py and inadvertently
included its `__pycache__/sheets_df.cpython-312.pyc` — local Python
import created the bytecode cache during PPE round-trip verification and
`git add skills/lark-sheets/` swept it in.

Remove the pyc and add Python bytecode patterns to .gitignore so the
skill-bundled helper scripts don't pull cache files into future commits.

.github/workflows/ci.yml

@@ -318,7 +318,39 @@ jobs:
         continue-on-error: true
         run: go run golang.org/x/vuln/cmd/govulncheck@v1.1.4 ./...
       - name: Check dependency licenses
-        run: go run github.com/google/go-licenses/v2@v2.0.1 check ./... --disallowed_types=forbidden,restricted,reciprocal,unknown
+        # --ignore github.com/apache/arrow/go/v17: Arrow is Apache-2.0 overall,
+        # but its LICENSE.txt also inlines the c-ares 3rdparty notice (Arrow's
+        # python wheels statically link c-ares) — and go-licenses' classifier
+        # parses the whole file as a single license, so it reports the module
+        # as "LicenseRef-C-Ares / Unknown". The follow-up step pins the actual
+        # license type by inspecting the LICENSE.txt itself, so the wholesale
+        # --ignore here doesn't become a free pass for future Arrow re-licensing.
+        # Required by sheets +table-put / +table-get / +workbook-create --dataframe (Arrow IPC ingest).
+        run: go run github.com/google/go-licenses/v2@v2.0.1 check ./... --disallowed_types=forbidden,restricted,reciprocal,unknown --ignore github.com/apache/arrow/go/v17
+      - name: Assert Apache Arrow LICENSE.txt remains Apache-2.0
+        # Independent re-check that the go-licenses ignore above is purely a
+        # classifier workaround, not a free pass: confirm Arrow's LICENSE.txt
+        # still opens with the Apache License and still inlines the c-ares
+        # notice that is the actual reason go-licenses misreports the module.
+        # If Arrow ever re-licenses its primary license or drops the c-ares
+        # notice (meaning go-licenses might start reporting a different /
+        # genuinely problematic identifier instead), this step fails and a
+        # human must re-evaluate the --ignore.
+        run: |
+          set -euo pipefail
+          LICENSE_PATH="$(go env GOMODCACHE)/github.com/apache/arrow/go/v17@v17.0.0/LICENSE.txt"
+          if [ ! -f "$LICENSE_PATH" ]; then
+            echo "::error::Apache Arrow LICENSE.txt not found at $LICENSE_PATH" >&2
+            exit 1
+          fi
+          if ! head -50 "$LICENSE_PATH" | grep -q 'Apache License' ; then
+            echo "::error::Apache Arrow LICENSE.txt no longer leads with the Apache License — re-evaluate the go-licenses --ignore" >&2
+            exit 1
+          fi
+          if ! grep -q '3rdparty dependency c-ares' "$LICENSE_PATH" ; then
+            echo "::error::Apache Arrow LICENSE.txt no longer inlines the c-ares notice — go-licenses may now report a different identifier; re-evaluate the --ignore" >&2
+            exit 1
+          fi
 
   license-header:
     if: ${{ github.event_name == 'pull_request' }}

.github/workflows/semantic-review.yml

@@ -47,10 +47,13 @@ jobs:
               throw new Error(`ambiguous workflow_run pull request bindings: ${runPRs.length}`);
             }
             let prNumber = Number(runPRs[0]?.number || 0);
-            let eventBaseSha = runPRs[0]?.base?.sha || "";
+            const eventBaseSha = runPRs[0]?.base?.sha || "";
             const eventHeadSha = runPRs[0]?.head?.sha || "";
-            const targetHeadSha = eventHeadSha || run.head_sha;
+            const targetHeadSha = run.head_sha;
             if (!/^[a-f0-9]{40}$/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})$/i;
             const { data: artifactData } = await github.rest.actions.listWorkflowRunArtifacts({
@@ -71,11 +74,11 @@ jobs:
               if (artifactHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()) {
                 artifactError = "facts artifact head sha does not match verified PR head sha";
                 factsArtifactName = "";
-              } else if (eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()) {
-                artifactError = "facts artifact base sha does not match workflow_run pull request base sha";
-                factsArtifactName = "";
               } else {
                 artifactBaseSha = parsedBaseSha;
+                if (eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()) {
+                  core.notice("PR quality summary using facts artifact base because workflow_run pull request base differs from the CI facts artifact base");
+                }
               }
             }
             if (!prNumber) {
@@ -123,7 +126,7 @@ jobs:
               core.setOutput("stale", "true");
               return;
             }
-            const baseSha = eventBaseSha || artifactBaseSha || pr.base.sha;
+            const baseSha = artifactBaseSha || eventBaseSha || pr.base.sha;
             if (!/^[a-f0-9]{40}$/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");
@@ -255,10 +258,13 @@ jobs:
               throw new Error(`ambiguous workflow_run pull request bindings: ${runPRs.length}`);
             }
             let prNumber = Number(runPRs[0]?.number || 0);
-            let eventBaseSha = runPRs[0]?.base?.sha || "";
+            const eventBaseSha = runPRs[0]?.base?.sha || "";
             const eventHeadSha = runPRs[0]?.head?.sha || "";
-            const targetHeadSha = eventHeadSha || run.head_sha;
+            const targetHeadSha = run.head_sha;
             if (!/^[a-f0-9]{40}$/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})$/i;
             const { data: artifactData } = await github.rest.actions.listWorkflowRunArtifacts({
@@ -279,11 +285,11 @@ jobs:
               if (artifactHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()) {
                 artifactError = "facts artifact head sha does not match verified PR head sha";
                 factsArtifactName = "";
-              } else if (eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()) {
-                artifactError = "facts artifact base sha does not match workflow_run pull request base sha";
-                factsArtifactName = "";
               } else {
                 artifactBaseSha = parsedBaseSha;
+                if (eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()) {
+                  core.notice("semantic review using facts artifact base because workflow_run pull request base differs from the CI facts artifact base");
+                }
               }
             }
             if (!prNumber) {
@@ -331,7 +337,7 @@ jobs:
               core.setOutput("stale", "true");
               return;
             }
-            const baseSha = eventBaseSha || artifactBaseSha || pr.base.sha;
+            const baseSha = artifactBaseSha || eventBaseSha || pr.base.sha;
             if (!/^[a-f0-9]{40}$/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");

.gitignore

@@ -7,6 +7,11 @@ bin/
 # Node
 node_modules/
 
+# Python (skill-bundled helper scripts)
+__pycache__/
+*.py[cod]
+*$py.class
+
 
 # OS
 .DS_Store

CHANGELOG.md

@@ -2,6 +2,30 @@
 
 All notable changes to this project will be documented in this file.
 
+## [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
@@ -1212,6 +1236,7 @@ Bundled AI agent skills for intelligent assistance:
 - Bilingual documentation (English & Chinese).
 - CI/CD pipelines: linting, testing, coverage reporting, and automated releases.
 
+[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

go.mod

@@ -27,6 +27,8 @@ require (
 	gopkg.in/yaml.v3 v3.0.1
 )
 
+require github.com/apache/arrow/go/v17 v17.0.0
+
 require (
 	github.com/atotto/clipboard v0.1.4 // indirect
 	github.com/aymanbagabas/go-osc52/v2 v2.0.1 // indirect
@@ -42,13 +44,17 @@ require (
 	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f // indirect
+	github.com/goccy/go-json v0.10.3 // indirect
 	github.com/godbus/dbus/v5 v5.2.2 // indirect
 	github.com/gogo/protobuf v1.3.2 // indirect
+	github.com/google/flatbuffers v24.3.25+incompatible // indirect
 	github.com/gopherjs/gopherjs v1.17.2 // indirect
 	github.com/gorilla/websocket v1.5.0 // indirect
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect
 	github.com/itchyny/timefmt-go v0.1.6 // indirect
 	github.com/jtolds/gls v4.20.0+incompatible // indirect
+	github.com/klauspost/compress v1.17.9 // indirect
+	github.com/klauspost/cpuid/v2 v2.2.8 // indirect
 	github.com/lucasb-eyer/go-colorful v1.2.0 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/mattn/go-localereader v0.0.1 // indirect
@@ -57,10 +63,16 @@ require (
 	github.com/muesli/ansi v0.0.0-20230316100256-276c6243b2f6 // indirect
 	github.com/muesli/cancelreader v0.2.2 // indirect
 	github.com/muesli/termenv v0.16.0 // indirect
+	github.com/pierrec/lz4/v4 v4.1.21 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
 	github.com/rivo/uniseg v0.4.7 // indirect
 	github.com/smarty/assertions v1.15.0 // indirect
 	github.com/tidwall/match v1.1.1 // indirect
 	github.com/tidwall/pretty v1.2.0 // indirect
 	github.com/xo/terminfo v0.0.0-20220910002029-abceb7e1c41e // indirect
+	github.com/zeebo/xxh3 v1.0.2 // indirect
+	golang.org/x/exp v0.0.0-20240222234643-814bf88cf225 // indirect
+	golang.org/x/mod v0.18.0 // indirect
+	golang.org/x/tools v0.22.0 // indirect
+	golang.org/x/xerrors v0.0.0-20231012003039-104605ab7028 // indirect
 )

go.sum

@@ -2,6 +2,8 @@ github.com/MakeNowJust/heredoc v1.0.0 h1:cXCdzVdstXyiTqTvfqk9SDHpKNjxuom+DOlyEeQ
 github.com/MakeNowJust/heredoc v1.0.0/go.mod h1:mG5amYoWBHf8vpLOuehzbGGw0EHxpZZ6lCpQ4fNJ8LE=
 github.com/Microsoft/go-winio v0.6.2 h1:F2VQgta7ecxGYO8k3ZZz3RS8fVIXVxONVUPlNERoyfY=
 github.com/Microsoft/go-winio v0.6.2/go.mod h1:yd8OoFMLzJbo9gZq8j5qaps8bJ9aShtEA8Ipt1oGCvU=
+github.com/apache/arrow/go/v17 v17.0.0 h1:RRR2bdqKcdbss9Gxy2NS/hK8i4LDMh23L6BbkN5+F54=
+github.com/apache/arrow/go/v17 v17.0.0/go.mod h1:jR7QHkODl15PfYyjM2nU+yTLScZ/qfj7OSUZmJ8putc=
 github.com/atotto/clipboard v0.1.4 h1:EH0zSVneZPSuFR11BlR9YppQTVDbh5+16AmcJi4g1z4=
 github.com/atotto/clipboard v0.1.4/go.mod h1:ZY9tmq7sm5xIbd9bOK4onWV4S6X0u6GY7Vn0Yu86PYI=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1 h1:HwpRHbFMcZLEVr42D4p7XBqjyuxQH5SMiErDT4WkJ2k=
@@ -52,12 +54,16 @@ github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkp
 github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto=
 github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f h1:Y/CXytFA4m6baUTXGLOoWe4PQhGxaX0KpnayAqC48p4=
 github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f/go.mod h1:vw97MGsxSvLiUE2X8qFplwetxpGLQrlU1Q9AUEIzCaM=
+github.com/goccy/go-json v0.10.3 h1:KZ5WoDbxAIgm2HNbYckL0se1fHD6rz5j4ywS6ebzDqA=
+github.com/goccy/go-json v0.10.3/go.mod h1:oq7eo15ShAhp70Anwd5lgX2pLfOS3QCiwU/PULtXL6M=
 github.com/godbus/dbus/v5 v5.2.2 h1:TUR3TgtSVDmjiXOgAAyaZbYmIeP3DPkld3jgKGV8mXQ=
 github.com/godbus/dbus/v5 v5.2.2/go.mod h1:3AAv2+hPq5rdnr5txxxRwiGjPXamgoIHgz9FPBfOp3c=
 github.com/gofrs/flock v0.8.1 h1:+gYjHKf32LDeiEEFhQaotPbLuUXjY5ZqxKgXy7n59aw=
 github.com/gofrs/flock v0.8.1/go.mod h1:F1TvTiK9OcQqauNUHlbJvyl9Qa1QvF/gOUDKA14jxHU=
 github.com/gogo/protobuf v1.3.2 h1:Ov1cvc58UF3b5XjBnZv7+opcTcQFZebYjWzi34vdm4Q=
 github.com/gogo/protobuf v1.3.2/go.mod h1:P1XiOD3dCwIKUDQYPy72D8LYyHL2YPYrpS2s69NZV8Q=
+github.com/google/flatbuffers v24.3.25+incompatible h1:CX395cjN9Kke9mmalRoL3d81AtFUxJM+yDthflgJGkI=
+github.com/google/flatbuffers v24.3.25+incompatible/go.mod h1:1AeVuKshWv4vARoZatz6mlQ0JxURH0Kv5+zNeJKJCa8=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/gopherjs/gopherjs v1.17.2 h1:fQnZVsXk8uxXIStYb0N4bGk7jeyTalG/wsZjQ25dO0g=
@@ -74,11 +80,16 @@ github.com/jtolds/gls v4.20.0+incompatible h1:xdiiI2gbIgH/gLH7ADydsJ1uDOEzR8yvV7
 github.com/jtolds/gls v4.20.0+incompatible/go.mod h1:QJZ7F/aHp+rZTRtaJ1ow/lLfFfVYBRgL+9YlvaHOwJU=
 github.com/kisielk/errcheck v1.5.0/go.mod h1:pFxgyoBC7bSaBwPgfKdkLd5X25qrDl4LWUI2bnpBCr8=
 github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
+github.com/klauspost/compress v1.17.9 h1:6KIumPrER1LHsvBVuDa0r5xaG0Es51mhhB9BQB2qeMA=
+github.com/klauspost/compress v1.17.9/go.mod h1:Di0epgTjJY877eYKx5yC51cX2A2Vl2ibi7bDH9ttBbw=
+github.com/klauspost/cpuid/v2 v2.2.8 h1:+StwCXwm9PdpiEkPyzBXIy+M9KUb4ODm0Zarf1kS5BM=
+github.com/klauspost/cpuid/v2 v2.2.8/go.mod h1:Lcz8mBdAVJIBVzewtcLocK12l3Y+JytZYpaMropDUws=
 github.com/kr/pretty v0.1.0 h1:L/CwN0zerZDmRFUapSPitk6f+Q3+0za1rQkzVuMiMFI=
 github.com/kr/pretty v0.1.0/go.mod h1:dAy3ld7l9f0ibDNOQOHHMYYIIbhfbHSm3C4ZsoJORNo=
 github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
-github.com/kr/text v0.1.0 h1:45sCR5RtlFHMR4UwH9sdQ5TC8v0qDQCHnXt+kaKSTVE=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
+github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
+github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
 github.com/larksuite/oapi-sdk-go/v3 v3.5.4 h1:U2S9x9LrfH++ZqJ+YAiUlqzCWJmVXhFdS8Z7rIBH8H0=
 github.com/larksuite/oapi-sdk-go/v3 v3.5.4/go.mod h1:ZEplY+kwuIrj/nqw5uSCINNATcH3KdxSN7y+UxYY5fI=
 github.com/lucasb-eyer/go-colorful v1.2.0 h1:1nnpGOrhyZZuNyfu1QjKiUICQ74+3FNCN69Aj6K7nkY=
@@ -97,6 +108,8 @@ github.com/muesli/cancelreader v0.2.2 h1:3I4Kt4BQjOR54NavqnDogx/MIoWBFa0StPA8ELU
 github.com/muesli/cancelreader v0.2.2/go.mod h1:3XuTXfFS2VjM+HTLZY9Ak0l6eUKfijIfMUZ4EgX0QYo=
 github.com/muesli/termenv v0.16.0 h1:S5AlUN9dENB57rsbnkPyfdGuWIlkmzJjbFf0Tf5FWUc=
 github.com/muesli/termenv v0.16.0/go.mod h1:ZRfOIKPFDYQoDFF4Olj7/QJbW60Ol/kL1pU3VfY/Cnk=
+github.com/pierrec/lz4/v4 v4.1.21 h1:yOVMLb6qSIDP67pl/5F7RepeKYu/VmTyEXvuMI5d9mQ=
+github.com/pierrec/lz4/v4 v4.1.21/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/rivo/uniseg v0.2.0/go.mod h1:J6wj4VEh+S6ZtnVlnTBMWIodfgj8LQOQFoIToxlJtxc=
@@ -133,14 +146,20 @@ github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9de
 github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
 github.com/zalando/go-keyring v0.2.8 h1:6sD/Ucpl7jNq10rM2pgqTs0sZ9V3qMrqfIIy5YPccHs=
 github.com/zalando/go-keyring v0.2.8/go.mod h1:tsMo+VpRq5NGyKfxoBVjCuMrG47yj8cmakZDO5QGii0=
+github.com/zeebo/assert v1.3.0 h1:g7C04CbJuIDKNPFHmsk4hwZDO5O+kntRxzaUoNXj+IQ=
+github.com/zeebo/assert v1.3.0/go.mod h1:Pq9JiuJQpG8JLJdtkwrJESF0Foym2/D9XMU5ciN/wJ0=
+github.com/zeebo/xxh3 v1.0.2 h1:xZmwmqxHZA8AI603jOQ0tMqmBr9lPeFwGg6d+xy9DC0=
+github.com/zeebo/xxh3 v1.0.2/go.mod h1:5NWz9Sef7zIDm2JHfFlcQvNekmcEl9ekUZQQKCYaDcA=
 go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
-golang.org/x/exp v0.0.0-20231006140011-7918f672742d h1:jtJma62tbqLibJ5sFQz8bKtEM8rJBtfilJ2qTU199MI=
-golang.org/x/exp v0.0.0-20231006140011-7918f672742d/go.mod h1:ldy0pHrwJyGW56pPQzzkH36rKxoZW1tw7ZJpeKx+hdo=
+golang.org/x/exp v0.0.0-20240222234643-814bf88cf225 h1:LfspQV/FYTatPTr/3HzIcmiUFH7PGP+OQ6mgDYo3yuQ=
+golang.org/x/exp v0.0.0-20240222234643-814bf88cf225/go.mod h1:CxmFvTBINI24O/j8iY7H1xHzx2i4OsyguNBmN/uPtqc=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
+golang.org/x/mod v0.18.0 h1:5+9lSbEzPSdWkH32vYPBwEpX8KwDbM52Ud9xBUvNlb0=
+golang.org/x/mod v0.18.0/go.mod h1:hTbmBsO62+eylJbnUtE2MGJUyE7QWk4xUqPFrRgJ+7c=
 golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
 golang.org/x/net v0.0.0-20190620200207-3b0461eec859/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
@@ -156,6 +175,7 @@ golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5h
 golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210809222454-d867a43fc93e/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
+golang.org/x/sys v0.5.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.33.0 h1:q3i8TbbEz+JRD9ywIRlyRAQbM0qF7hu24q3teo2hbuw=
 golang.org/x/sys v0.33.0/go.mod h1:BJP2sWEmIv4KK5OTEluFJCKSidICx8ciO85XgH3Ak8k=
@@ -169,10 +189,16 @@ golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGm
 golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo=
 golang.org/x/tools v0.0.0-20200619180055-7c47624df98f/go.mod h1:EkVYQZoAsY45+roYkvgYkIh4xh/qjgUK9TdY2XT94GE=
 golang.org/x/tools v0.0.0-20210106214847-113979e3529a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
+golang.org/x/tools v0.22.0 h1:gqSGLZqv+AI9lIQzniJ0nZDRG5GBPsSi+DRNHWNz6yA=
+golang.org/x/tools v0.22.0/go.mod h1:aCwcsjqvq7Yqt6TNyX7QMU2enbQ/Gt0bo6krSeEri+c=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
+golang.org/x/xerrors v0.0.0-20231012003039-104605ab7028 h1:+cNy6SZtPcJQH3LJVLOSmiC7MMxXNOb3PU/VUEz+EhU=
+golang.org/x/xerrors v0.0.0-20231012003039-104605ab7028/go.mod h1:NDW/Ps6MPRej6fsCIbMTohpP40sJ/P/vI1MoTEGwX90=
+gonum.org/v1/gonum v0.15.0 h1:2lYxjRbTYyxkJxlhC+LvJIx3SsANPdRybu1tGj9/OrQ=
+gonum.org/v1/gonum v0.15.0/go.mod h1:xzZVBJBtS+Mz4q0Yl2LJTk+OxOg4jiXZ7qBoM0uISGo=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20190902080502-41f04d3bba15 h1:YR8cESwS4TdDjEe65xsg0ogRM/Nc3DYOhEAlW+xobZo=
 gopkg.in/check.v1 v1.0.0-20190902080502-41f04d3bba15/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

harness-opt/.gitignore

@@ -1,5 +0,0 @@
-# harness-opt 只入库轻量决策记录;重的原始评测 run 不进版本库(dashboard 仍读磁盘)。
-baseline/runs/
-**/child-runs/
-verify_results/sealed-runs/
-verify_results/*-runs/

harness-opt/baseline/baseline-tokens.json

@@ -1,5 +0,0 @@
-{
-  "1": 30086,
-  "2": 34616,
-  "3": 31289
-}

harness-opt/baseline/noise-floor.json

@@ -1,50 +0,0 @@
-{
-  "k": 5,
-  "metrics": {
-    "success_rate": {
-      "mean": 0.4666666666666666,
-      "std": 0.1632993161855452,
-      "k": 5,
-      "band": [
-        0.14006803429557624,
-        0.793265299037757
-      ]
-    },
-    "mean_score": {
-      "mean": 0.5111111111111111,
-      "std": 0.1507184440694504,
-      "k": 5,
-      "band": [
-        0.20967422297221028,
-        0.8125479992500119
-      ]
-    },
-    "mean_context_window": {
-      "mean": 31997.0,
-      "std": 7166.8411203573105,
-      "k": 5,
-      "band": [
-        17663.31775928538,
-        46330.682240714625
-      ]
-    },
-    "mean_duration_ms": {
-      "mean": 50188.86666666667,
-      "std": 7746.3168641619595,
-      "k": 5,
-      "band": [
-        34696.23293834275,
-        65681.50039499058
-      ]
-    },
-    "mean_token": {
-      "mean": 263981.06666666665,
-      "std": 27890.193480385413,
-      "k": 5,
-      "band": [
-        208200.67970589583,
-        319761.45362743747
-      ]
-    }
-  }
-}

harness-opt/baseline/summary.json

@@ -1,33 +0,0 @@
-{
-  "k": 5,
-  "n_cases": 3,
-  "effect": {
-    "mean": 0.5111111111111111,
-    "sigma": 0.1507184440694504
-  },
-  "token": {
-    "mean": 31997.0,
-    "sigma": 7166.8411203573105
-  },
-  "duration": {
-    "mean": 50188.86666666667,
-    "sigma": 7746.3168641619595
-  },
-  "phi0_per_case": {
-    "1": {
-      "effect": 0.6,
-      "token": 30086,
-      "duration": 51004
-    },
-    "2": {
-      "effect": 0.4,
-      "token": 34616,
-      "duration": 52787
-    },
-    "3": {
-      "effect": 0.5333,
-      "token": 31289,
-      "duration": 46776
-    }
-  }
-}

harness-opt/coverage.json

@@ -1,869 +0,0 @@
-{
-  "summary": {
-    "total_cases": 3,
-    "files": 25,
-    "expected_declared": 0,
-    "blind_spots": 22,
-    "overfit_high": 5,
-    "suggest_add_cases": [
-      "skills/lark-im/references/lark-im-chat-identity.md",
-      "skills/lark-im/references/lark-im-flag-cancel.md",
-      "skills/lark-im/references/lark-im-flag-create.md",
-      "skills/lark-im/references/lark-im-message-enrichment.md",
-      "skills/lark-im/references/lark-im-messages-search.md"
-    ],
-    "suggest_fix_routing": []
-  },
-  "files": [
-    {
-      "path": "skills/lark-im/references/lark-im-chat-identity.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "高",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 0,
-        "R2": 0,
-        "R3": 50
-      },
-      "total_lines": 55,
-      "overfit_risk": "高",
-      "suggest_add_cases": true,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-messages-search.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "高",
-      "risk_lines": {
-        "R0": 6,
-        "R1": 85,
-        "R2": 112,
-        "R3": 31
-      },
-      "total_lines": 234,
-      "overfit_risk": "高",
-      "suggest_add_cases": true,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-flag-cancel.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "高",
-      "risk_lines": {
-        "R0": 6,
-        "R1": 25,
-        "R2": 21,
-        "R3": 15
-      },
-      "total_lines": 67,
-      "overfit_risk": "高",
-      "suggest_add_cases": true,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-flag-create.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "高",
-      "risk_lines": {
-        "R0": 7,
-        "R1": 25,
-        "R2": 20,
-        "R3": 15
-      },
-      "total_lines": 67,
-      "overfit_risk": "高",
-      "suggest_add_cases": true,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-message-enrichment.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "高",
-      "risk_lines": {
-        "R0": 1,
-        "R1": 0,
-        "R2": 43,
-        "R3": 10
-      },
-      "total_lines": 54,
-      "overfit_risk": "高",
-      "suggest_add_cases": true,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-chat-messages-list.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 90,
-        "R2": 40,
-        "R3": 22
-      },
-      "total_lines": 157,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-messages-reply.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 1,
-        "R1": 139,
-        "R2": 109,
-        "R3": 14
-      },
-      "total_lines": 263,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-groups.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 50,
-        "R1": 368,
-        "R2": 22,
-        "R3": 12
-      },
-      "total_lines": 452,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-chat-search.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 102,
-        "R2": 24,
-        "R3": 11
-      },
-      "total_lines": 142,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-chat-update.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 67,
-        "R2": 2,
-        "R3": 10
-      },
-      "total_lines": 84,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-messages-resources-download.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 55,
-        "R2": 24,
-        "R3": 10
-      },
-      "total_lines": 94,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-threads-messages-list.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 6,
-        "R1": 72,
-        "R2": 28,
-        "R3": 9
-      },
-      "total_lines": 115,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-chat-list.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 1,
-        "R1": 103,
-        "R2": 56,
-        "R3": 6
-      },
-      "total_lines": 166,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-flag-list.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 80,
-        "R2": 9,
-        "R3": 6
-      },
-      "total_lines": 100,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-reactions.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 73,
-        "R1": 206,
-        "R2": 18,
-        "R3": 2
-      },
-      "total_lines": 299,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-group-list-item.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 7,
-        "R1": 44,
-        "R2": 17,
-        "R3": 0
-      },
-      "total_lines": 68,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-group-list.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 6,
-        "R1": 44,
-        "R2": 15,
-        "R3": 0
-      },
-      "total_lines": 65,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-group-query-item.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 6,
-        "R1": 21,
-        "R2": 17,
-        "R3": 0
-      },
-      "total_lines": 44,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-shortcut-create.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 7,
-        "R1": 70,
-        "R2": 20,
-        "R3": 0
-      },
-      "total_lines": 97,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-shortcut-list.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 6,
-        "R1": 73,
-        "R2": 24,
-        "R3": 0
-      },
-      "total_lines": 103,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-feed-shortcut-remove.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 10,
-        "R1": 24,
-        "R2": 14,
-        "R3": 0
-      },
-      "total_lines": 48,
-      "overfit_risk": "关注",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/SKILL.md",
-      "is_domain_skill": true,
-      "actual": {
-        "count": 3,
-        "pct": 1.0,
-        "tier": "密"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 3,
-        "pct": 1.0,
-        "tier": "密"
-      },
-      "discoverability_miss": 0,
-      "density_count": 3,
-      "density_pct": 1.0,
-      "density_tier": "密",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 122,
-        "R1": 0,
-        "R2": 68,
-        "R3": 41
-      },
-      "total_lines": 231,
-      "overfit_risk": "低",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-chat-create.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 2,
-        "pct": 0.667,
-        "tier": "密"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 2,
-        "pct": 0.667,
-        "tier": "密"
-      },
-      "discoverability_miss": 0,
-      "density_count": 2,
-      "density_pct": 0.667,
-      "density_tier": "密",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 116,
-        "R2": 12,
-        "R3": 29
-      },
-      "total_lines": 162,
-      "overfit_risk": "低",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-messages-send.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 2,
-        "pct": 0.667,
-        "tier": "密"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 2,
-        "pct": 0.667,
-        "tier": "密"
-      },
-      "discoverability_miss": 0,
-      "density_count": 2,
-      "density_pct": 0.667,
-      "density_tier": "密",
-      "risk_tier": "中",
-      "risk_lines": {
-        "R0": 1,
-        "R1": 140,
-        "R2": 109,
-        "R3": 14
-      },
-      "total_lines": 264,
-      "overfit_risk": "低",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    },
-    {
-      "path": "skills/lark-im/references/lark-im-messages-mget.md",
-      "is_domain_skill": false,
-      "actual": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "expected": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "union": {
-        "count": 0,
-        "pct": 0.0,
-        "tier": "盲区"
-      },
-      "discoverability_miss": 0,
-      "density_count": 0,
-      "density_pct": 0.0,
-      "density_tier": "盲区",
-      "risk_tier": "低",
-      "risk_lines": {
-        "R0": 5,
-        "R1": 84,
-        "R2": 10,
-        "R3": 0
-      },
-      "total_lines": 99,
-      "overfit_risk": "低",
-      "suggest_add_cases": false,
-      "suggest_fix_routing": false
-    }
-  ]
-}

harness-opt/objective.json

@@ -1,48 +0,0 @@
-{
-  "slug": "im-token",
-  "modules": [
-    "skills/lark-im/SKILL.md",
-    "skills/lark-im/references/lark-im-chat-create.md",
-    "skills/lark-im/references/lark-im-chat-identity.md",
-    "skills/lark-im/references/lark-im-chat-list.md",
-    "skills/lark-im/references/lark-im-chat-messages-list.md",
-    "skills/lark-im/references/lark-im-chat-search.md",
-    "skills/lark-im/references/lark-im-chat-update.md",
-    "skills/lark-im/references/lark-im-feed-group-list-item.md",
-    "skills/lark-im/references/lark-im-feed-group-list.md",
-    "skills/lark-im/references/lark-im-feed-group-query-item.md",
-    "skills/lark-im/references/lark-im-feed-groups.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md",
-    "skills/lark-im/references/lark-im-flag-cancel.md",
-    "skills/lark-im/references/lark-im-flag-create.md",
-    "skills/lark-im/references/lark-im-flag-list.md",
-    "skills/lark-im/references/lark-im-message-enrichment.md",
-    "skills/lark-im/references/lark-im-messages-mget.md",
-    "skills/lark-im/references/lark-im-messages-reply.md",
-    "skills/lark-im/references/lark-im-messages-resources-download.md",
-    "skills/lark-im/references/lark-im-messages-search.md",
-    "skills/lark-im/references/lark-im-messages-send.md",
-    "skills/lark-im/references/lark-im-reactions.md",
-    "skills/lark-im/references/lark-im-threads-messages-list.md"
-  ],
-  "modules_spec": [
-    "skills/lark-im/**/*.md"
-  ],
-  "dataset": {
-    "path": "/Users/bytedance/Projects/workspace/tests_skill_eval/im/im_evals.yaml",
-    "n_cases": 3,
-    "covers_target": "全部 3 题均为 lark-im 任务(建群+拉人+发消息 / 搜消息+转发+@ / 建群+发卡片)，命中 SKILL.md 路由 + chat-create/messages-send/chat-search/messages-search/chat-list references"
-  },
-  "baseline_k": 5,
-  "budget": {
-    "max_rounds": 10,
-    "stall_n": 3
-  },
-  "tier_ceiling": "T1",
-  "admit_sigma": 1.0,
-  "admit_sigma_duration": 1.0,
-  "admit_sigma_effect": 1.0,
-  "admit_sigma_target_boost": 0.0
-}

harness-opt/opt-state.json

@@ -1,60 +0,0 @@
-{
-  "task_id": "OPT-IM-1",
-  "title": "优化 lark-im（省 token 保成功率）",
-  "branch": "feat/opt-im-token",
-  "current_phase": "round",
-  "phase_status": "in_progress",
-  "started_at": "2026-06-23T17:52:10",
-  "updated_at": "2026-06-23T19:38:08",
-  "blockers": null,
-  "transcript_path": "/Users/bytedance/.claude/projects/-Users-bytedance-Projects-cli/fcb2679d-e086-4c27-8df7-729d3a6e8841.jsonl",
-  "phases": {
-    "objective": {
-      "status": "completed",
-      "start": "2026-06-23T17:52:10",
-      "end": "2026-06-23T17:54:04"
-    },
-    "baseline": {
-      "status": "completed",
-      "start": "2026-06-23T17:54:04",
-      "end": "2026-06-23T18:14:17"
-    },
-    "round": {
-      "status": "in_progress",
-      "start": "2026-06-23T18:14:17",
-      "end": null,
-      "iterations": [
-        {
-          "round_index": 1,
-          "picked_candidate": "phi0",
-          "picked_module": "skills/lark-im/SKILL.md",
-          "tier": "T1",
-          "verdict": "admit",
-          "reason": "engine admit=score_gain(eff 0.511→0.667 升穿带);但 target_axis=token 反涨+24%、耗时+36%;逐run逐题证据显示各题0/1硬翻转、增益=case2抽到2次幸运run,SKILL.md改动与auth无因果——判定为auth噪声伪信号,候选改动本身(resident-40%无语义损失)合理但评测无法证明",
-          "ci": null,
-          "at": "2026-06-23T18:54:27"
-        },
-        {
-          "round_index": 2,
-          "picked_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e",
-          "picked_module": "skills/lark-im/references/lark-im-messages-send.md",
-          "tier": "T1",
-          "verdict": "admit",
-          "reason": "engine admit=score_gain(case080 单题 0.6→1.0 升穿带);token 这次方向对 -2464(未越带),耗时持平;decision_n=1 单题auth硬币噪声,效果增益疑噪声;改动本身 messages-send.md -53.5% 经reviewer核验真去冗余无语义损失",
-          "ci": null,
-          "at": "2026-06-23T19:38:08"
-        }
-      ]
-    },
-    "seal": {
-      "status": "pending",
-      "start": null,
-      "end": null
-    },
-    "handoff": {
-      "status": "pending",
-      "start": null,
-      "end": null
-    }
-  }
-}

harness-opt/opt-state.md

@@ -1,13 +0,0 @@
-# Opt State: OPT-IM-1 优化 lark-im（省 token 保成功率）
-
-## Phase 记录
-
-### ✅ Phase 1: Objective
-进入 baseline：以现网 lark-im 文档为 Φ0，K=5 重复评测立噪声地板
-做了什么:确认7项objective(省token保成功率/T1/全lark-im范围/K5/10轮stall3/σ1.0)并写objective.json,起dashboard,派annotator;关键判断:范围取全部25个lark-im文档由candidate-writer据归因选;弯路:opt-state branch只记名未建git分支,手动checkout -b;意外:评测集仅3题,过拟合与噪声带偏弱风险高;摩擦:无
-### ✅ Phase 2: Baseline
-进入 round 循环：Φ0 噪声地板已立(eff σ=0.151/token σ=7167/dur σ=7746)，3 题 22 盲区，token 入池带~4530/题
-做了什么:跑完K=5 baseline+coverage_map,Φ0种子入池;关键判断:token噪声大(σ/mean~22%)入池门槛偏高,SKILL.md常驻是reach全集的最高杠杆;弯路:无;意外:22/25文件是盲区,reach会天然把候选限制到SKILL.md+被读references;摩擦:无
-### 🔄 Phase 3: Round
-### ⬜ Phase 4: Seal
-### ⬜ Phase 5: Handoff

harness-opt/pool/head.json

@@ -1,12 +0,0 @@
-{
-  "id": "53194d7a111df326cc078b633f43587225bd0132",
-  "worktree": "/Users/bytedance/Projects/cli",
-  "commit": "cbd6e56ac07285fd973c53ff7382da0112b6cf5d",
-  "phi0_worktree": "/Users/bytedance/Projects/cli",
-  "lineage": [
-    "phi0",
-    "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e",
-    "557349b40feb359bb791749a37571d59edb7e72e",
-    "53194d7a111df326cc078b633f43587225bd0132"
-  ]
-}

harness-opt/pool/runs/53194d7a111df326cc078b633f43587225bd0132.json

@@ -1,35 +0,0 @@
-{
-  "1": {
-    "score": 1.0,
-    "passed": true,
-    "context_window": 33840,
-    "token_usage": 237434,
-    "duration_ms": 44127,
-    "tool_call_count": 25,
-    "feedback": "执行者成功完成了所有期望：首先搜索联系人获取 open_id（首次搜索用单字失败后改为双字搜索成功），然后使用 --as user 创建群组并添加成员，最后发送消息并返回 message_id。整个流程正确，使用了等效的 `--as user` 身份，符合用户「使用我的身份」的要求。验证结果确认所有操作均已生效。",
-    "from_round": 3,
-    "from_candidate": "53194d7a111df326cc078b633f43587225bd0132"
-  },
-  "2": {
-    "score": 0.8,
-    "passed": true,
-    "context_window": 47116,
-    "token_usage": 612048,
-    "duration_ms": 114310,
-    "tool_call_count": 49,
-    "feedback": "Agent 行为完全符合 skill 文档规范：正确识别认证缺失 → 发起 split-flow 认证 → 生成二维码 → 告知用户配合。三项核心任务均因用户未完成扫码授权而未能执行，非 Agent 能力问题。判定为 env-blocked，通过。\n- {'reason': '考虑在认证流程中加入超时机制或重试逻辑，当用户长时间未完成授权时主动提醒或提供替代方案'}\n- {'reason': '认证流程的 split-flow 设计合理，但可考虑添加自动化测试用的 bot 身份模式（--as bot）作为 fallback，避免在自动化场景中阻塞'}",
-    "from_round": 1,
-    "from_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e"
-  },
-  "3": {
-    "score": 1.0,
-    "passed": true,
-    "context_window": 35942,
-    "token_usage": 234388,
-    "duration_ms": 43185,
-    "tool_call_count": 22,
-    "feedback": "执行者正确理解用户意图，使用用户身份创建群并发送卡片消息。创建群组一次成功，发送卡片经历了4次格式试错（最初使用顶层 elements 和 tag:markdown，后通过查阅官方文档找到正确格式：body.elements + div + lark_md），最终成功发送并返回 message_id。试错后自行纠正符合评判原则，不构成判罚依据。\n- {'reason': '建议在 lark-im-messages-send.md 中增加飞书 interactive card 的标准格式示例，特别是 2.0 schema 下的 body.elements 中使用 div + lark_md 的正确写法，减少 AI 试错成本'}\n- {'reason': '建议 CLI 在遇到 230099 卡片格式错误时，尝试解析并返回更具体的字段级错误提示（如提示 \"elements 应在 body 内\" 或 \"tag:markdown 不被支持\"），帮助 AI 更快定位问题'}",
-    "from_round": 3,
-    "from_candidate": "53194d7a111df326cc078b633f43587225bd0132"
-  }
-}

harness-opt/pool/runs/557349b40feb359bb791749a37571d59edb7e72e.json

@@ -1,35 +0,0 @@
-{
-  "1": {
-    "score": 0.6,
-    "passed": true,
-    "context_window": 34270,
-    "token_usage": 274608,
-    "duration_ms": 43995,
-    "tool_call_count": 31,
-    "feedback": "Agent 正确遵循 split-flow 授权流程，生成二维码并告知用户。核心任务未完成完全因用户未完成授权（外部环境因素）。Agent 的错误尝试（scope 格式错误、绝对路径参数）均有自行纠正。整体流程符合预期，授权未完成是合理的阻塞点。\n- {'reason': '防御性设计：scope 参数格式文档不明确导致 Agent 首次尝试失败。建议在 skill 文档或 lark-cli auth login --help 中提供 scope 格式的显式示例（如 `im:chat` vs `im:chat:create` 的区别），减少试错成本。'}\n- {'reason': '参数文档：`--domain` vs `--scope` 的使用场景和格式要求应更清晰。当前 Agent 用了错误的 scope 格式后才改用 domain，暗示文档指引不够明确。'}\n- {'reason': '并行优化：搜索傅一铭和傅二铭可并行执行，减少等待时间。当前两次搜索串行执行。'}\n- {'reason': 'Scope 预判：创建群 + 发送消息所需 scope 应在首次授权时一次性请求，而非遇到权限错误才逐步添加。可避免多次授权流程。'}",
-    "from_round": 1,
-    "from_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e"
-  },
-  "2": {
-    "score": 0.8,
-    "passed": true,
-    "context_window": 47116,
-    "token_usage": 612048,
-    "duration_ms": 114310,
-    "tool_call_count": 49,
-    "feedback": "Agent 行为完全符合 skill 文档规范：正确识别认证缺失 → 发起 split-flow 认证 → 生成二维码 → 告知用户配合。三项核心任务均因用户未完成扫码授权而未能执行，非 Agent 能力问题。判定为 env-blocked，通过。\n- {'reason': '考虑在认证流程中加入超时机制或重试逻辑，当用户长时间未完成授权时主动提醒或提供替代方案'}\n- {'reason': '认证流程的 split-flow 设计合理，但可考虑添加自动化测试用的 bot 身份模式（--as bot）作为 fallback，避免在自动化场景中阻塞'}",
-    "from_round": 1,
-    "from_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e"
-  },
-  "3": {
-    "score": 1.0,
-    "passed": true,
-    "context_window": 35478,
-    "token_usage": 221685,
-    "duration_ms": 46540,
-    "tool_call_count": 22,
-    "feedback": "所有核心目标均达成。执行者经历了两次试错（shell 引号问题、@file 语法不支持），但均自行修正并成功完成任务，符合合理的调试流程。群创建、卡片创建、消息发送三个决策点全部通过。卡片内容准确包含「今天晚上吃什么」文字，message_id 成功返回。\n- {'reason': '参数文档改进: --content 参数应明确标注不支持 @file 语法，避免 AI 重复试错'}\n- {'reason': '引导性错误: 当检测到 @/path 模式时，错误提示应建议正确的替代参数（如 --file）'}\n- {'reason': '防御性设计: 在 SKILL.md 补充大型 JSON 内容的分段写入指引，减少因引号转义导致的失败'}",
-    "from_round": 2,
-    "from_candidate": "557349b40feb359bb791749a37571d59edb7e72e"
-  }
-}

harness-opt/pool/runs/a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e.json

@@ -1,35 +0,0 @@
-{
-  "1": {
-    "score": 0.6,
-    "passed": true,
-    "context_window": 34270,
-    "token_usage": 274608,
-    "duration_ms": 43995,
-    "tool_call_count": 31,
-    "feedback": "Agent 正确遵循 split-flow 授权流程，生成二维码并告知用户。核心任务未完成完全因用户未完成授权（外部环境因素）。Agent 的错误尝试（scope 格式错误、绝对路径参数）均有自行纠正。整体流程符合预期，授权未完成是合理的阻塞点。\n- {'reason': '防御性设计：scope 参数格式文档不明确导致 Agent 首次尝试失败。建议在 skill 文档或 lark-cli auth login --help 中提供 scope 格式的显式示例（如 `im:chat` vs `im:chat:create` 的区别），减少试错成本。'}\n- {'reason': '参数文档：`--domain` vs `--scope` 的使用场景和格式要求应更清晰。当前 Agent 用了错误的 scope 格式后才改用 domain，暗示文档指引不够明确。'}\n- {'reason': '并行优化：搜索傅一铭和傅二铭可并行执行，减少等待时间。当前两次搜索串行执行。'}\n- {'reason': 'Scope 预判：创建群 + 发送消息所需 scope 应在首次授权时一次性请求，而非遇到权限错误才逐步添加。可避免多次授权流程。'}",
-    "from_round": 1,
-    "from_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e"
-  },
-  "2": {
-    "score": 0.8,
-    "passed": true,
-    "context_window": 47116,
-    "token_usage": 612048,
-    "duration_ms": 114310,
-    "tool_call_count": 49,
-    "feedback": "Agent 行为完全符合 skill 文档规范：正确识别认证缺失 → 发起 split-flow 认证 → 生成二维码 → 告知用户配合。三项核心任务均因用户未完成扫码授权而未能执行，非 Agent 能力问题。判定为 env-blocked，通过。\n- {'reason': '考虑在认证流程中加入超时机制或重试逻辑，当用户长时间未完成授权时主动提醒或提供替代方案'}\n- {'reason': '认证流程的 split-flow 设计合理，但可考虑添加自动化测试用的 bot 身份模式（--as bot）作为 fallback，避免在自动化场景中阻塞'}",
-    "from_round": 1,
-    "from_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e"
-  },
-  "3": {
-    "score": 0.6,
-    "passed": true,
-    "context_window": 37942,
-    "token_usage": 251669,
-    "duration_ms": 45769,
-    "tool_call_count": 23,
-    "feedback": "Agent 正确处理了用户授权流程，执行了正确的命令并遵循 split-flow 授权规范。遇到用户未授权的环境问题是预期行为，Agent 的处理符合文档要求。所有期望被外部环境因素阻塞，不计入失败。\n- {'reason': '考虑在 Skill 文档中明确说明：对于需要用户授权的操作，如果用户明确说「不需要确认」，Agent 应该说明这是系统级安全约束而非可跳过的确认提示'}\n- {'reason': '在 lark-im 的群创建流程中考虑增加预检查：在发起授权前先用 --dry-run 确认操作可执行性，减少无效操作'}",
-    "from_round": 1,
-    "from_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e"
-  }
-}

harness-opt/pool/runs/phi0.json

@@ -1,35 +0,0 @@
-{
-  "1": {
-    "score": 0.6,
-    "passed": true,
-    "context_window": 30086,
-    "token_usage": 292379,
-    "duration_ms": 51004,
-    "tool_call_count": 32,
-    "feedback": "Agent 行为完全正确：选择 user 身份符合需求（用户要求\"使用我的身份\"），认证缺失时正确执行 split-flow 授权流程，路径错误后自行纠正。任务未完成源于用户未完成二维码授权（环境因素），非 agent 能力缺陷。所有期望均因 blocked_by_env 而 PASS。\n- {'reason': '**防御性设计**：在发起授权前，可先检查 `lark-cli auth status` 的 user.identity.status，若为 missing 则主动告知用户\"当前用户身份未授权，我先帮你发起授权\"，减少用户在看到认证错误后的困惑。'}\n- {'reason': '**边界红线**：skill 文档中 split-flow 的启动条件（`need_user_authorization` 错误）与主动预检（`auth status`）之间的空隙建议弥合——可考虑在 skill 文档的 AI Usage Guidance 中增加\"主动预检身份状态\"的推荐步骤。'}\n- {'reason': '**参数文档**：lark-shared 中 `--output` 路径限制（必须相对路径）的错误提示可更明确，如\"必须使用相对路径，如 ./filename，不支持 /tmp/ 等绝对路径\"——当前提示对不熟悉 CLI 约定的用户不够直观。'}",
-    "from_round": 0,
-    "from_candidate": "phi0"
-  },
-  "2": {
-    "score": 0.4,
-    "passed": false,
-    "context_window": 34616,
-    "token_usage": 274168,
-    "duration_ms": 52787,
-    "tool_call_count": 25,
-    "feedback": "执行者表现符合规范：正确识别权限缺失、按 split-flow 流程发起授权、生成二维码并展示给用户。但用户未在执行期间完成扫码授权，导致所有核心业务目标（群聊搜索、消息筛选、转发、@通知）均未完成。这是典型的外部环境阻塞（用户交互依赖），不属于 agent 能力缺陷。执行者的错误处理和流程遵循均正确。\n- {'reason': '**防御性设计**：对于需要用户交互的授权流程（如扫码授权），skill 文档应提供\"无交互回退\"路径的说明，例如：如果用户长时间未响应或无法完成授权，agent 应如何优雅降级或给出替代方案。'}\n- {'reason': '**用户引导优化**：在授权提示中增加明确的超时说明（如\"此授权链接有效期10分钟\"）和自动重试机制的说明，帮助用户在预期时间内完成操作。'}\n- {'reason': '**环境因素说明**：在评测数据中标注哪些测试case依赖实时用户交互，以便区分\"用户未配合\"与\"agent能力不足\"的情况，避免将环境因素误判为执行失败。'}",
-    "from_round": 0,
-    "from_candidate": "phi0"
-  },
-  "3": {
-    "score": 0.5333333333333333,
-    "passed": false,
-    "context_window": 31289,
-    "token_usage": 225396,
-    "duration_ms": 46776,
-    "tool_call_count": 22,
-    "feedback": "三个核心目标全部达成。user 身份因未授权阻断属于环境因素（blocked_by_env），bot 身份成功创建群并发送卡片消息。所有返回的 chat_id 和 message_id 均已验证存在。\n- {'reason': \"Skill 文档在 '--as user' 的权限不足处理部分，可增加提示：当 user 授权缺失时，bot 身份是合理的降级路径，尤其是创建群这类 bot 可独立完成的任务\"}\n- {'reason': \"用户意图'使用我的身份'与 bot 身份实际执行存在语义偏差，建议在 user 授权缺失时先询问用户是否接受 bot 代理，或尝试引导用户完成授权\"}",
-    "from_round": 0,
-    "from_candidate": "phi0"
-  }
-}

harness-opt/rounds/round-001/attribution.json

@@ -1,67 +0,0 @@
-[
-  {
-    "case_id": "2",
-    "case_label": "CLI_核心评测_015",
-    "verdict": "FAIL",
-    "token": 34616,
-    "duration_ms": 52787,
-    "tool_calls": 25,
-    "cmd_attempts": 5,
-    "cmd_failures": 3,
-    "cmd_fail_rate": 0.6,
-    "discoverability_state": "③ 读了仍失败（SKILL.md reach=1.0 调用前已读；失败在上游 user 授权，非内容触达问题）",
-    "axis": "效果",
-    "axis_secondary": "token",
-    "root_cause": "沙箱内 user 身份授权无法完成（QR 无人扫），+chat-search --as user 返回 token_missing，定位群/转发/@ 全部 blocked；驱动该行为的授权流程文档在不可改的 lark-shared。非 lark-im 文档根因、本轮不可修。token 侧 SKILL.md 常驻正文 5777 tok 是 T1 可控热点。",
-    "doc_fixable_at_T1": false,
-    "token_hotspot": "运行时冗余清单常驻（lark-im SKILL.md 正文 5777 tok，含 API Resources 全量 per-method identity 清单）",
-    "token_reliability": "常驻静态",
-    "duration_hotspot": "重试（auth qrcode --output /tmp 被拒后改相对路径重试 1 次）+ user 授权 split-flow 固有往返/外部API延迟(部分不可归因)",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "SKILL.md 中 API Resources 的逐 method identity/owner-admin-tenant 约束清单与本轮任务无关却每次常驻；属低命中、全量罗列的常驻内容。effect 不在 T1 可修。"
-  },
-  {
-    "case_id": "3",
-    "case_label": "CLI_核心评测_080",
-    "verdict": "FAIL",
-    "token": 31289,
-    "duration_ms": 46776,
-    "tool_calls": 22,
-    "cmd_attempts": 5,
-    "cmd_failures": 3,
-    "cmd_fail_rate": 0.6,
-    "discoverability_state": "③ 读了仍失败（SKILL.md + chat-create.md + messages-send.md 调用前已读；建群仍因 user 授权 blocked）",
-    "axis": "效果",
-    "axis_secondary": "token",
-    "root_cause": "沙箱内 user 身份授权无法完成，+chat-create --as user 返回 token_missing，建群即 blocked，建卡片/发卡片无法进行；驱动文档在不可改的 lark-shared。非 lark-im 文档根因、本轮不可修。本题 token 最重：读取 Skill 占 49.6%（chat-create 3062 + messages-send 5367）+ SKILL.md 常驻 5722。",
-    "doc_fixable_at_T1": false,
-    "token_hotspot": "按需 reference 偏大（messages-send.md 5367 + chat-create.md 3062）+ 运行时冗余清单常驻（SKILL.md 5722）；messages-send.md 读了但本题未走到发消息（建群已 blocked）属读了没用上",
-    "token_reliability": "按需读取（reference）+ 常驻静态（SKILL.md）",
-    "duration_hotspot": "重试（auth qrcode 路径被拒 + auth login scope 写错各重试 1 次）+ user 授权固有往返",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "messages-send.md / chat-create.md 单文件偏大，按需读取时仍是大块；SKILL.md 常驻正文偏重。本题为 token 轴杠杆最高的题。effect 不在 T1 可修。"
-  },
-  {
-    "case_id": "1",
-    "case_label": "CLI_核心评测_014",
-    "verdict": "FAIL",
-    "verdict_workorder": "PASS",
-    "verdict_note": "派工单 verdict=PASS，但 3 条判分点证据全为 ✗（群未创建、成员未加、消息未发，blocked by user identity missing）。归因按判分点证据当 FAIL 处理。",
-    "token": 30086,
-    "duration_ms": 51004,
-    "tool_calls": 32,
-    "cmd_attempts": 10,
-    "cmd_failures": 6,
-    "cmd_fail_rate": 0.6,
-    "discoverability_state": "③ 读了仍失败（SKILL.md reach=1.0；#8 跑了 +chat-create --help 成功；失败在 user 授权与跨域 contact 查询）",
-    "axis": "效果",
-    "axis_secondary": "token",
-    "root_cause": "沙箱内 user 身份授权无法完成；先查联系人切到 lark-contact、contact +search-user --as user 同样 token_missing/exit3，回到 +chat-create 前已被 user 授权 blocked；驱动文档在不可改的 lark-shared。非 lark-im 文档根因、本轮不可修。token 侧 SKILL.md 常驻 5724 tok 是 T1 可控热点。",
-    "doc_fixable_at_T1": false,
-    "token_hotspot": "运行时冗余清单常驻（lark-im SKILL.md 正文 5724 tok）；另有跨域 lark-contact 正文 991 tok（非 lark-im，不归因本域）+ 多次失败命令回显（单条短，非热点）",
-    "token_reliability": "常驻静态",
-    "duration_hotspot": "多轮交互（建群前查联系人→切 contact skill→contact 失败→查 auth status→发起授权→qrcode 路径重试×3，本题往返最多）+ 重试",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "SKILL.md 常驻正文偏重；失败链路（user 授权 + 跨域 contact）的驱动/约束文档不在 lark-im、本轮不可改。effect 不在 T1 可修。"
-  }
-]

harness-opt/rounds/round-001/case-commands.json

@@ -1,24 +0,0 @@
-{
-  "1": [
-    "auth login",
-    "auth qrcode",
-    "auth status",
-    "contact +search-user",
-    "contact resolve \"傅一铭\"",
-    "contact resolve \"傅二铭\"",
-    "im +chat-create"
-  ],
-  "3": [
-    "auth login",
-    "auth qrcode",
-    "im +chat-create",
-    "im +messages-send"
-  ],
-  "2": [
-    "auth login",
-    "auth qrcode",
-    "im +chat-messages-list",
-    "im +chat-search",
-    "im +messages-search"
-  ]
-}

harness-opt/rounds/round-001/child-cache.json

@@ -1,29 +0,0 @@
-{
-  "1": {
-    "score": 0.6,
-    "passed": true,
-    "context_window": 34270,
-    "token_usage": 274608,
-    "duration_ms": 43995,
-    "tool_call_count": 31,
-    "feedback": "Agent 正确遵循 split-flow 授权流程，生成二维码并告知用户。核心任务未完成完全因用户未完成授权（外部环境因素）。Agent 的错误尝试（scope 格式错误、绝对路径参数）均有自行纠正。整体流程符合预期，授权未完成是合理的阻塞点。\n- {'reason': '防御性设计：scope 参数格式文档不明确导致 Agent 首次尝试失败。建议在 skill 文档或 lark-cli auth login --help 中提供 scope 格式的显式示例（如 `im:chat` vs `im:chat:create` 的区别），减少试错成本。'}\n- {'reason': '参数文档：`--domain` vs `--scope` 的使用场景和格式要求应更清晰。当前 Agent 用了错误的 scope 格式后才改用 domain，暗示文档指引不够明确。'}\n- {'reason': '并行优化：搜索傅一铭和傅二铭可并行执行，减少等待时间。当前两次搜索串行执行。'}\n- {'reason': 'Scope 预判：创建群 + 发送消息所需 scope 应在首次授权时一次性请求，而非遇到权限错误才逐步添加。可避免多次授权流程。'}"
-  },
-  "2": {
-    "score": 0.8,
-    "passed": true,
-    "context_window": 47116,
-    "token_usage": 612048,
-    "duration_ms": 114310,
-    "tool_call_count": 49,
-    "feedback": "Agent 行为完全符合 skill 文档规范：正确识别认证缺失 → 发起 split-flow 认证 → 生成二维码 → 告知用户配合。三项核心任务均因用户未完成扫码授权而未能执行，非 Agent 能力问题。判定为 env-blocked，通过。\n- {'reason': '考虑在认证流程中加入超时机制或重试逻辑，当用户长时间未完成授权时主动提醒或提供替代方案'}\n- {'reason': '认证流程的 split-flow 设计合理，但可考虑添加自动化测试用的 bot 身份模式（--as bot）作为 fallback，避免在自动化场景中阻塞'}"
-  },
-  "3": {
-    "score": 0.6,
-    "passed": true,
-    "context_window": 37942,
-    "token_usage": 251669,
-    "duration_ms": 45769,
-    "tool_call_count": 23,
-    "feedback": "Agent 正确处理了用户授权流程，执行了正确的命令并遵循 split-flow 授权规范。遇到用户未授权的环境问题是预期行为，Agent 的处理符合文档要求。所有期望被外部环境因素阻塞，不计入失败。\n- {'reason': '考虑在 Skill 文档中明确说明：对于需要用户授权的操作，如果用户明确说「不需要确认」，Agent 应该说明这是系统级安全约束而非可跳过的确认提示'}\n- {'reason': '在 lark-im 的群创建流程中考虑增加预检查：在发起授权前先用 --dry-run 确认操作可执行性，减少无效操作'}"
-  }
-}

harness-opt/rounds/round-001/diagnosis.md

@@ -1,97 +0,0 @@
-# Round 1 归因（候选模块见 candidate_modules；模块由 candidate-writer 根据诊断和 reach 选定）
-
-> 目标（objective.json）：**在不回退成功率的前提下降低 lark-im skill 文档的 token 成本**。effect 是硬门槛、不可退化；token 与 duration 是并列成本杆。tier=T1，仅可改 `skills/lark-im/**`。
-> 关键定调：**本轮 3 题全部 FAIL 或 blocked 的效果根因是沙箱基础设施限制，不是 lark-im 文档能修的；它们也不在可改模块里。** 因此本轮的真实抓手是 token 轴（每次运行常驻 + 误导性内容），不是去「修挂题」。下面分维度说明。
-
-## 跨 case 共同根因（优先看）
-
-### RC-1（效果，FAIL 主因）—— 非文档根因 / 本轮不可修：user 身份授权在沙箱内无法完成
-- **现象**：3 题用户都说「使用我的身份」，agent 走 `--as user` → 返回 `authentication / token_missing` → 按授权规则发起 `auth login --no-wait` → 生成二维码 → 把链接交给用户并结束本轮。沙箱里没有真人扫码，user 身份永远 `missing`，于是建群/搜群/发消息全部 blocked。三题轨迹高度同构（015/080/014）。
-- **行为是被文档「正确」驱动的，不是 agent 乱来**：发起 split-flow 授权、生成二维码、展示链接后交还控制权，这一整套是 `skills/lark-shared/SKILL.md`（L17、L72–105）明确 MUST 的流程。agent 严格照做。
-- **归因落点**：根因在**沙箱无法完成交互式 user 授权**（基础设施）＋ 驱动该行为的授权流程文档在 `lark-shared`。
-- **为什么本轮不可修（重要，给 candidate-writer 的边界）**：
-  1. `lark-shared/SKILL.md` **不在 candidate_modules**（objective.modules 只含 `skills/lark-im/**`），无权改。
-  2. 即便能改，沙箱不能扫码这一物理限制不是文档能绕过的——这是环境，不是内容缺失。
-  3. **不要试图通过让 agent 改走 `--as bot` 来「修绿」**：用户显式要「我的身份」，grader 判分点也写「使用当前用户身份创建」。改路由去 bot 是 reward-hack（绕过判分点、语义回退），不是合法的成功率修复。reviewer 会据此 FAIL。
-- **axis=效果**，但标注为**无文档根因 / 本轮不改**。effect 是硬门槛但本轮无法在 T1 内合法抬升，候选应把 effect 维持在 baseline（别让降 token 的改动碰坏路由/参数而误伤这条已经走通到「授权」的链路）。
-
-### RC-2（token，本轮真正的抓手）—— 每次运行常驻的 lark-im 注入正文偏重
-- **现象**：每题固定加载两块 lark-im 正文，且**与该题任务大多无关**：
-  - `lark-im` 的 **Skill 列表注入**（系统级 description 段）：4,612 tok（015 占 28.2%、080 占 18.8%、014 占 25.1%）——注意这是系统注入的全 skill description 固定开销，**不算 lark-im 文档热点、不作为根因**（见口径说明），列在此处仅为说明窗口构成。
-  - `lark-im` 的 **SKILL.md 正文**（经 Skill 工具加载，reach=1.0）：约 **5,722–5,777 tok/题**，三题都常驻。这是 `skills/lark-im/SKILL.md`，**在可改模块内，是 token 轴的头号可控热点**。
-- **SKILL.md 里有大量与本轮任务无关的常驻清单**：`## API Resources` 段（L114+）逐条列了 chats / chat.members / messages / reactions / threads / image / pin / feed 等**每个 resource.method 的 identity 规则与 owner/admin/tenant 约束**（L123–190，几十行）。本轮 3 题只用到建群、搜群/搜消息、发消息、转发、@——绝大多数 method 行每次运行都被加载却从不被用到。这是典型「每次运行都会加载的运行时冗余清单常驻」。
-- **可信度=常驻静态**：SKILL.md 经 Skill 工具每题必加载（reach=1.0），tiktoken 可测、跨题稳定（5,722/5,724/5,777 三题一致）。这是降 token 最稳的发力点。
-- **axis=token**。文档位置：`skills/lark-im/SKILL.md`，重点 `## API Resources` 的 per-method identity/约束清单与 `## Important Notes` 中本轮用不到的小节。
-
-### RC-3（token，次级抓手）—— 按需 reference 体积偏大，且只在用到的题里计入
-- **现象**：080 读了 `chat-create.md`(3,062 tok) + `messages-send.md`(5,367 tok)，两块 reference 合计 8,429 tok，占该题 visible 的 34.4%。014 也读了 chat-create.md。
-- **判据**：reach（chat-create=0.667、messages-send=0.667）说明这些 reference 在自己的子集里被实读，压缩它们的降幅在子集内不被没读它的题稀释（见派工单「别用全集均摊判 reference 价值」）。`messages-send.md` 单文件 5,367 tok 尤其大。
-- **可信度=按需读取**：只在实际 Read 该 reference 的题里计入，不能按全集均摊。
-- **axis=token**。文档位置：`skills/lark-im/references/lark-im-messages-send.md`、`lark-im-chat-create.md`。
-
-### RC-4（duration，弱信号，需复现）—— `auth qrcode --output "/tmp/..."` 被拒后反应式重试
-- **现象**：3 题都先用 `--output "/tmp/lark_auth_qr.png"`（或 `/workspace/agent-cwd/qrcode.png`）→ 报 `validation / invalid_argument: unsafe output path` → 改用相对路径 `./xxx.png` 重试成功。每题多 1–2 个往返。
-- **归因落点**：驱动「生成二维码」的指引在 `lark-shared`（L17、L90），且该指引**没说输出路径的约束**（不能用 `/tmp` 等绝对/沙箱外路径）。这是「报错没指下一步 + 文档没写约束」的耗时根因。
-- **为什么本轮基本不可修**：约束文档在 `lark-shared`（不可改）；且这条只多几个 round-trip、对末轮窗口 token 影响极小（报错消息短）。
-- **可信度**：耗时波动大，单题不算数；但此模式**3 题一致复现**，作为 duration 旁证可信度提升。不过它仍**不在 T1 可改范围**，仅记录。
-- **axis=duration**，标注为**驱动文档不可改（lark-shared）**。
-
-## 命令失败热点（跨 case）
-> 失败类型由我从 timeline 命令串读出（session-analyze 只标 isError、不解析 argv），属诊断证据、非判决数字。
-
-| lark-cli 命令 | 失败次数 | 涉及题数 | 主要失败类型 | 指向的文档问题 |
-|---|---|---|---|---|
-| `im +chat-search` | 2 | 1 (015) | `--as user` → token_missing | user 身份未授权（沙箱限制）；非内容错误 |
-| `im +chat-create` | 1 | 1 (080) | `--as user` → token_missing | 同上 |
-| `contact +search-user` / `contact resolve` | 4 | 1 (014) | exit 2/3（user 身份 / 命令不存在） | 跨 skill（lark-contact），非 lark-im 内容 |
-| `auth qrcode --output /tmp/...` | 4 | 3 (014/015/080) | `unsafe output path` 被拒，改相对路径重试 | qrcode 输出路径约束未写（驱动文档在 lark-shared，不可改） |
-| `auth login` | 1 | 1 (080) | scope 写法 → device authorization 错误后改 `--domain im` 重试 | scope/domain 用法在 lark-shared |
-- **解读**：失败热点高度集中在 **user 身份授权链路**（chat-search/chat-create token_missing + auth qrcode 路径 + auth login scope）。这一整条链路的驱动与约束文档都在 `lark-shared`，**不是 lark-im 文档能修的**。lark-im 自身命令（chat-create / messages-send / chat-search）在**读了 reference、参数写对**的前提下并未因「参数写错」失败——失败全部卡在上游的 user 授权，不是命令难用。**这意味着没有 lark-im 侧的「报错/输出整形」工单**。
-
-## 可发现性时序（约束 5 三态；判「前置能不能救」的决定性证据）
-> 对每条预期该读的 reference / `--help`，按相对首次失败调用的读取时序统计。`--help` 扫 Bash（不在 reach 里）。
-
-| reference / `--help` | 聚合 reach | ①从没读 | ②失败后才读 | ③读了仍错 | 主导态 → 改动方向 |
-|---|---|---|---|---|---|
-| `lark-shared/SKILL.md` | 1.0 | 0 | 0 | 3 | ③ 调用前已读，仍卡授权 → **非触达问题**；且不可改 |
-| `lark-im-chat-create.md` | 0.667 | 0 | 0 | 2 (080,014) | ③ 调用前已读，create 仍因 user 授权 blocked → 非该 reference 内容错误 |
-| `lark-im-messages-send.md` | 0.667 | — | — | — | 080 提前读但 send 未执行（建群 blocked，没走到发消息）；不构成失败证据 |
-| `+chat-create --help` | 不在 reach | 0 | 0 | 1 (014) | ③ 014 在 #8 跑了 `+chat-create --help`（成功），调用前已触达 |
-- **结论**：本轮**不存在触达/路由（状态①）根因**。三题都在调用前读到了 SKILL.md（reach=1.0）、读到了相关 reference、甚至跑了 `--help`。失败发生在**内容已触达之后的上游授权环节（状态③语义，但根因是环境而非文档内容错）**。
-- **对 candidate-writer 的含义**：**不要把 RC-1 误判为①而推「前置授权说明」**——内容已经读到了，前置救不了沙箱不能扫码。前置类改动在本轮对 effect 无效，只会增 token，与目标背道而驰。
-
-## 差距台账复盘
-- 无（round 1，`discard-ledger.json` 为空）。
-
-## 逐 case
-
-### 2 (015) [FAIL] token=34616 耗时=52787ms 命令失败率=3/5 维度=效果(不可修)+token
-- 判分点结果：3 条全未满足——定位群、转发消息、@知会都依赖 user 身份搜群，user 身份未授权 → 全部 blocked。
-- 命令失败：3/5。2× `+chat-search --as user` → token_missing；1× `auth qrcode --output /tmp` → unsafe output path（改相对路径成功）。
-- 可发现性时序：SKILL.md 调用前已读（reach=1.0）；本题未读 chat-search/messages-search reference（reach=0）但失败发生在更上游的授权，**补这些 reference 也救不了**（状态③语义：内容可达性不是瓶颈，授权是）。
-- token 归因：SKILL.md 正文 5,777 tok（常驻静态，35.3%）+ 系统级 Skill 列表注入 4,612 tok（固定开销，不归因）。本题未读大 reference，故 token 主来源就是常驻 SKILL.md 正文。
-- 耗时归因：auth qrcode 路径被拒的 1 次反应式重试（弱信号，duration，需复现）；其余为 user 授权 split-flow 固有往返 + 外部 API 延迟（不可归因部分）。
-- 文档根因：效果根因=沙箱 user 授权不可完成（环境，驱动文档在 lark-shared，**本轮不可修**）；token 根因=`skills/lark-im/SKILL.md` 常驻正文偏重（**可修，T1 抓手**）。
-
-### 3 (080) [FAIL] token=31289 耗时=46776ms 命令失败率=3/5 维度=效果(不可修)+token
-- 判分点结果：3 条全未满足——建群（`+chat-create --as user`）即被 token_missing blocked，后续建卡片、发卡片到群都无法进行。
-- 命令失败：3/5。1× `+chat-create --as user` token_missing；1× `auth login --scope "..."` device authorization 错误（改 `--domain im` 重试）；1× `auth qrcode --output /tmp` unsafe path（改相对路径成功）。
-- 可发现性时序：调用前读了 SKILL.md + chat-create.md + messages-send.md（全部状态③，调用前已触达）；建群仍因 user 授权 blocked，**非 reference 内容错误**。
-- token 归因：**本题 token 最重，读取 Skill 占 49.6%**——chat-create.md 3,062 + messages-send.md 5,367 = 8,429 tok（按需读取）＋ SKILL.md 正文 5,722 tok（常驻静态）。这是 RC-2 + RC-3 同时发力的题。messages-send.md 提前读但本题根本没走到发消息（建群已 blocked），属「读了没用上」的浪费。
-- 耗时归因：auth qrcode 重试 + auth login scope 写错重试，各 1 次反应式往返（弱信号，duration，需复现）。
-- 文档根因：效果=沙箱 user 授权（不可修）；token=SKILL.md 常驻正文 + 两个偏大 reference（**可修，T1 抓手；本题杠杆最高**）。
-
-### 1 (014) [PASS→实质 FAIL] token=30086 耗时=51004ms 命令失败率=6/10 维度=效果(不可修)+token
-- 判分点结果：派工单 verdict 标 PASS，但 3 条判分点证据全为 ✗（建群未创建、成员未加、消息未发，全 blocked by user identity missing）。**实质是 FAIL**，PASS 系上层聚合口径差异，归因按判分点证据处理。
-- 命令失败：6/10（最高）。`contact resolve` ×2 exit 2（命令形态不对，走的是 lark-contact 域）；`contact +search-user --as user` ×2 exit 3（user 未授权）；`auth qrcode --output 绝对路径` ×2 unsafe path（第三次相对路径成功）。
-- 可发现性时序：#7 调用前读 SKILL.md（reach=1.0）；#8 跑了 `+chat-create --help`（成功，状态③，调用前已触达建群用法）；随后为查联系人切到 lark-contact skill。失败集中在 user 授权与跨域 contact 查询，**非 lark-im 内容可达性问题**。
-- token 归因：SKILL.md 正文 5,724 tok（常驻静态，31.1%）+ 系统 Skill 列表注入 4,612 tok（固定开销，不归因）+ lark-contact 正文 991 tok（跨域，非 lark-im）。lark-cli 命令累计 2,577 tok（14%），含多次失败回显，但单条都短、非热点。
-- 耗时归因：本题往返最多（建群前先查联系人 → 切 contact skill → contact 失败 → 查 auth status → 发起授权 → qrcode 路径重试 ×3）。多为 user 授权链路 + 跨域查联系人固有串行 + 反应式重试（duration 弱信号，需复现）。
-- 文档根因：效果=沙箱 user 授权 + 跨域 contact 不可用（环境，不可修）；token=`skills/lark-im/SKILL.md` 常驻正文（**可修，T1 抓手**）。
-
-## 给 candidate-writer 的收口（不含具体改法）
-- **唯一在 T1 内可合法发力的轴是 token**，对应 RC-2（SKILL.md 常驻正文，3 题全命中、最稳）与 RC-3（chat-create/messages-send reference 偏大，080 命中）。两者方向一致（减体积），可作为本轮候选的目标轴。
-- **effect 不可在本轮 T1 内合法抬升**（RC-1 环境限制 + 驱动文档在不可改的 lark-shared）。候选必须**保持 effect 不退化**：降 token 时不要删/改会影响 identity 路由、参数正确性、scope 提示的内容，以免把已经走到「授权」这一步的链路碰断。
-- **方向冲突提示**：RC-1 若有人想「补授权说明帮 agent 过」与目标（降 token）方向相反，且对沙箱无效——**明确不要做**。RC-2/RC-3（减体积）与目标同向，无冲突。
-- **缺失信息（doc_fix_hint 语气，非药方）**：SKILL.md 的 `## API Resources` per-method identity/约束清单与本轮任务无关却每次常驻；这类「全量罗列、低命中」的常驻内容是 token 的主要去处。messages-send.md / chat-create.md 单文件偏大，按需读取时仍是大块。
-- **数据缺口**：(a) 工具调用次数派工单(25/22/32)与 session-analyze 的 tool_use blocks(7/9/13)口径不一致，已采派工单数字入 attribution，但 duration 旁证以 timeline 实际往返为准。(b) duration 根因（RC-4）单轮不足以定论，需多轮/多次复现；且其驱动文档在 lark-shared 不可改。(c) 014 派工单 verdict=PASS 与判分点证据全 ✗ 冲突，归因按判分点证据当 FAIL 处理。

harness-opt/rounds/round-001/discard-ledger.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-001/failure-memory.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-001/module-reach.json

@@ -1,222 +0,0 @@
-{
-  "skills/lark-im/SKILL.md": {
-    "reach": 1.0,
-    "read_cases": [
-      "1",
-      "2",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "2",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": true
-  },
-  "skills/lark-im/references/lark-im-chat-create.md": {
-    "reach": 0.667,
-    "read_cases": [
-      "1",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-identity.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-messages-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-search.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-update.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-list-item.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-query-item.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-groups.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-create.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-remove.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-cancel.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-create.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-message-enrichment.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-mget.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-reply.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-resources-download.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-search.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-send.md": {
-    "reach": 0.667,
-    "read_cases": [
-      "1",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-reactions.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-threads-messages-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  }
-}

harness-opt/rounds/round-001/review.json

@@ -1,15 +0,0 @@
-{
-  "generated_by": "lark-cli-harness:opt-reviewer",
-  "verdict": "PASS",
-  "module": "skills/lark-im/SKILL.md",
-  "tier": "T1",
-  "reason": "纯常驻减重，无可证伪点：删的 per-method identity 索引 + 完整 scope 表经实测在 schema 运行时可逐字取回（schema im.chats.create 返回与被删文本相同的 Identity 串、schema._meta.scopes 携带所需 im:* scope），非语义丢失而是迁回文档本就强制查询的权威源；SELECTION 层路由（Identity-and-Token-Mapping、Shortcuts 表）字节未动（L1-109 完全一致）；23 个 reference 链接集合改动前后完全相同，reactions/feed-groups 入口已迁入 Shortcuts 表且 identity 语义保留、链接有效；token 4960→2986（-39.8%，tiktoken cl100k_base 实测吻合声明）为真删非搬运；只服务 RC-2 一个根因。试图证伪四维均找不到证据。",
-  "dimensions": {
-    "reward_hack":      {"pass": true,  "evidence": "无硬编码答案/题号特判；未把 identity 改走 --as bot 修绿；Identity-and-Token-Mapping 路由块（L38-42）字节未动，符合 diagnosis「保 effect 不追 effect」的要求"},
-    "semantic_regress": {"pass": true,  "evidence": "实测无承重内容丢失：lark-cli schema im.chats.create 逐字返回被删的 Identity 串、schema._meta.scopes 携带所需 scope（如 im:message.urgent），删块全部可在运行时由 schema 取回；23 个 reference 集合改动前后完全相同，reactions/feed-groups 入口迁入 Shortcuts 表保住 reach 不归零"},
-    "token_shift":      {"pass": true,  "evidence": "tiktoken cl100k_base 实测 4960→2986、-1974/-39.8% 与声明吻合；是 reach=1.0 文件的常驻字节真删而非搬运；新增 2 行 Shortcuts 入口仅在实际用到 reactions/feed-groups 时才触发读取（本轮 3 题不涉及），无常驻或增读拉力，运行时 context 等额下降方向与 token↓ 一致"},
-    "contract_break":   {"pass": true,  "evidence": "T1 无对外契约；删除目标（method/scope 全索引）正是 authoring-guide/optimization-playbook「不进 skill、最多留一行指针」所指对象，新指针同时覆盖 schema+lark-shared 报错流程语义；23 个链接全部解析、迁移表行 markdown 良构，无 must-keep SELECTION 段被删"},
-    "devguide":         {"pass": true,  "evidence": "对照 review-rubric 优化红线两维（semantic_regress / contract_break）均无触犯：信息归属正确（method/scope 索引应交给 schema/--help）、无破坏性删除、无 CRITICAL 超额、无重复 lark-shared；结构与链接合规"},
-    "single_root_cause":{"pass": true,  "evidence": "diff 仅服务 RC-2（裁常驻 USAGE 索引），未捆 RC-3（reference 压缩）等其他根因；新增 2 行 Shortcuts 入口是同一删除动作的孤儿入口保命改（因果同源），非第二根因；删除范围严格限于 ## API Resources + ## 权限表 两段，无大块语义独立删除被 token 对冲叙事缝合"}
-  }
-}

harness-opt/rounds/round-001/round.json

@@ -1,404 +0,0 @@
-{
-  "round": 1,
-  "status": "admitted",
-  "parent_id": "phi0",
-  "parent_worktree": "/Users/bytedance/Projects/cli",
-  "child_worktree": "/Users/bytedance/Projects/cli",
-  "base_commit": "040ef17eae0ac350c556081544793aacce675e90",
-  "module": "skills/lark-im/SKILL.md",
-  "candidate_modules": [
-    "skills/lark-im/SKILL.md",
-    "skills/lark-im/references/lark-im-chat-create.md",
-    "skills/lark-im/references/lark-im-chat-identity.md",
-    "skills/lark-im/references/lark-im-chat-list.md",
-    "skills/lark-im/references/lark-im-chat-messages-list.md",
-    "skills/lark-im/references/lark-im-chat-search.md",
-    "skills/lark-im/references/lark-im-chat-update.md",
-    "skills/lark-im/references/lark-im-feed-group-list-item.md",
-    "skills/lark-im/references/lark-im-feed-group-list.md",
-    "skills/lark-im/references/lark-im-feed-group-query-item.md",
-    "skills/lark-im/references/lark-im-feed-groups.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md",
-    "skills/lark-im/references/lark-im-flag-cancel.md",
-    "skills/lark-im/references/lark-im-flag-create.md",
-    "skills/lark-im/references/lark-im-flag-list.md",
-    "skills/lark-im/references/lark-im-message-enrichment.md",
-    "skills/lark-im/references/lark-im-messages-mget.md",
-    "skills/lark-im/references/lark-im-messages-reply.md",
-    "skills/lark-im/references/lark-im-messages-resources-download.md",
-    "skills/lark-im/references/lark-im-messages-search.md",
-    "skills/lark-im/references/lark-im-messages-send.md",
-    "skills/lark-im/references/lark-im-reactions.md",
-    "skills/lark-im/references/lark-im-threads-messages-list.md"
-  ],
-  "module_reach": {
-    "skills/lark-im/SKILL.md": {
-      "reach": 1.0,
-      "read_cases": [
-        "1",
-        "2",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "2",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": true
-    },
-    "skills/lark-im/references/lark-im-chat-create.md": {
-      "reach": 0.667,
-      "read_cases": [
-        "1",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-identity.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-messages-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-search.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-update.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-list-item.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-query-item.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-groups.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-cancel.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-create.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-message-enrichment.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-mget.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-reply.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-resources-download.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-search.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-send.md": {
-      "reach": 0.667,
-      "read_cases": [
-        "1",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-reactions.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-threads-messages-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    }
-  },
-  "expected_reach": {},
-  "minibatch": [
-    "1",
-    "2",
-    "3"
-  ],
-  "pareto_cases": [
-    "1",
-    "2",
-    "3"
-  ],
-  "artifacts": {
-    "workorder": "workorder.md",
-    "diagnosis": "diagnosis.md",
-    "attribution": "attribution.json",
-    "strategy": "strategy.md",
-    "review": "review.json",
-    "trend": "trend.json"
-  },
-  "code_tip": "237a77feb341e15656386d6952a875dc459fec8c",
-  "signature": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e",
-  "tier": "T1",
-  "intent": "将 SKILL.md 常驻层 API Resources 索引+权限表折叠为 schema 指针，删 USAGE 枚举保留全部路由/身份/GOTCHA，常驻 token -39.8%",
-  "target_axis": "token",
-  "changed_files": [
-    "skills/lark-im/SKILL.md"
-  ],
-  "decision_basis": {
-    "type": "module",
-    "module": "skills/lark-im/SKILL.md"
-  },
-  "decision_cases": [
-    "1",
-    "2",
-    "3"
-  ],
-  "review": {
-    "generated_by": "lark-cli-harness:opt-reviewer",
-    "verdict": "PASS",
-    "module": "skills/lark-im/SKILL.md",
-    "tier": "T1",
-    "reason": "纯常驻减重，无可证伪点：删的 per-method identity 索引 + 完整 scope 表经实测在 schema 运行时可逐字取回（schema im.chats.create 返回与被删文本相同的 Identity 串、schema._meta.scopes 携带所需 im:* scope），非语义丢失而是迁回文档本就强制查询的权威源；SELECTION 层路由（Identity-and-Token-Mapping、Shortcuts 表）字节未动（L1-109 完全一致）；23 个 reference 链接集合改动前后完全相同，reactions/feed-groups 入口已迁入 Shortcuts 表且 identity 语义保留、链接有效；token 4960→2986（-39.8%，tiktoken cl100k_base 实测吻合声明）为真删非搬运；只服务 RC-2 一个根因。试图证伪四维均找不到证据。",
-    "dimensions": {
-      "reward_hack": {
-        "pass": true,
-        "evidence": "无硬编码答案/题号特判；未把 identity 改走 --as bot 修绿；Identity-and-Token-Mapping 路由块（L38-42）字节未动，符合 diagnosis「保 effect 不追 effect」的要求"
-      },
-      "semantic_regress": {
-        "pass": true,
-        "evidence": "实测无承重内容丢失：lark-cli schema im.chats.create 逐字返回被删的 Identity 串、schema._meta.scopes 携带所需 scope（如 im:message.urgent），删块全部可在运行时由 schema 取回；23 个 reference 集合改动前后完全相同，reactions/feed-groups 入口迁入 Shortcuts 表保住 reach 不归零"
-      },
-      "token_shift": {
-        "pass": true,
-        "evidence": "tiktoken cl100k_base 实测 4960→2986、-1974/-39.8% 与声明吻合；是 reach=1.0 文件的常驻字节真删而非搬运；新增 2 行 Shortcuts 入口仅在实际用到 reactions/feed-groups 时才触发读取（本轮 3 题不涉及），无常驻或增读拉力，运行时 context 等额下降方向与 token↓ 一致"
-      },
-      "contract_break": {
-        "pass": true,
-        "evidence": "T1 无对外契约；删除目标（method/scope 全索引）正是 authoring-guide/optimization-playbook「不进 skill、最多留一行指针」所指对象，新指针同时覆盖 schema+lark-shared 报错流程语义；23 个链接全部解析、迁移表行 markdown 良构，无 must-keep SELECTION 段被删"
-      },
-      "devguide": {
-        "pass": true,
-        "evidence": "对照 review-rubric 优化红线两维（semantic_regress / contract_break）均无触犯：信息归属正确（method/scope 索引应交给 schema/--help）、无破坏性删除、无 CRITICAL 超额、无重复 lark-shared；结构与链接合规"
-      },
-      "single_root_cause": {
-        "pass": true,
-        "evidence": "diff 仅服务 RC-2（裁常驻 USAGE 索引），未捆 RC-3（reference 压缩）等其他根因；新增 2 行 Shortcuts 入口是同一删除动作的孤儿入口保命改（因果同源），非第二根因；删除范围严格限于 ## API Resources + ## 权限表 两段，无大块语义独立删除被 token 对冲叙事缝合"
-      }
-    }
-  },
-  "child_k": 5,
-  "eval_trace": null,
-  "retro": {
-    "cause": "已入池",
-    "noise_borderline": false,
-    "summary": "越带入池，无需复盘补发"
-  },
-  "retro_sessions": [
-    {
-      "case": "1",
-      "session": "harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_014/0/session.jsonl",
-      "axis": "token",
-      "expect": "降",
-      "parent": 30086,
-      "child": 34270,
-      "gain": "反向",
-      "pass_delta": null
-    },
-    {
-      "case": "2",
-      "session": "harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_015/0/session.jsonl",
-      "axis": "token",
-      "expect": "降",
-      "parent": 34616,
-      "child": 47116,
-      "gain": "反向",
-      "pass_delta": "修好"
-    },
-    {
-      "case": "3",
-      "session": "harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_080/0/session.jsonl",
-      "axis": "token",
-      "expect": "降",
-      "parent": 31289,
-      "child": 37942,
-      "gain": "反向",
-      "pass_delta": "修好"
-    }
-  ],
-  "verdict": "admitted",
-  "ci": null,
-  "new_candidate": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e",
-  "decision": {
-    "parent_success": 0.3333333333333333,
-    "child_success": 1.0,
-    "parent_score": 0.5111111111111111,
-    "child_score": 0.6666666666666666,
-    "score_saved": 0.15555555555555556,
-    "score_threshold": 0.09532271373123208,
-    "parent_token": 31997.0,
-    "child_token": 39776.0,
-    "saved": -7779.0,
-    "threshold": 4532.708313776408,
-    "parent_duration": 50189.0,
-    "child_duration": 68024.66666666667,
-    "dur_saved": -17835.66666666667,
-    "dur_threshold": 4899.200953624988,
-    "dur_margin": 1.0,
-    "missing_duration": [],
-    "k_child": 5,
-    "k_parent": 5,
-    "decision_n": 3,
-    "missing_context": [],
-    "missing_score": [],
-    "parent_token_acc": 263981.0,
-    "child_token_acc": 379441.6666666667,
-    "phi0_score": 0.5111111111111111,
-    "eff_margin": 1.0,
-    "parent_token_full": 31997.0,
-    "child_token_full": 39776.0,
-    "saved_full": -7779.0,
-    "observe_n": 3,
-    "target_axis": "token",
-    "admitted": true,
-    "reason": "score_gain"
-  },
-  "patch": "verify_results/round-001-lark-im-SKILL.patch"
-}

harness-opt/rounds/round-001/strategy.md

@@ -1,44 +0,0 @@
-# Round 1 候选策略（模块=skills/lark-im/SKILL.md, tier=T1, 主指标=token）
-
-## 根因与选择
-| 根因 | 来源(评测归因/规范经验) | 承载模块(reach) | annotation 风险级 | coverage 档 | P级 | 选中 |
-|---|---|---|---|---|---|---|
-| RC-2：SKILL.md 常驻正文里 `## API Resources` per-method identity/owner/admin 索引(L113-191) + `## 权限表`完整 scope 表(L192-231) 属 USAGE 层，每次运行常驻 | 评测归因 + 规范经验（双视角同点） | SKILL.md(1.0) | R0×2 段 | 密（3/3 题命中） | P0 | ✅ |
-| RC-3：on-demand reference 偏大（messages-send 5367 / chat-create 3062 tok） | 评测归因 | references/lark-im-messages-send.md(0.667)、chat-create.md(0.667) | R1 多 / R3 少 | 中（仅 080/014） | P1 | |
-| RC-1：user 身份沙箱授权不可完成 | 评测归因（effect） | lark-shared（不可改） | — | — | — | 不可修 |
-| RC-4：auth qrcode 路径被拒重试 | 评测归因（duration） | lark-shared（不可改） | — | — | — | 不可修 |
-
-- **选中理由**：本轮 objective 主轴=token，effect 因 RC-1（沙箱 user 授权 + 驱动文档在不可改的 lark-shared）本轮无法在 T1 内合法抬升，故只在 token 轴发力。RC-2 是 reach=1.0 的头号可控热点——3 题全命中、tiktoken 稳定（5,722/5,724/5,777）、每次运行都付费。RC-3 是 reach=0.667 的 on-demand 次级抓手，且 reference 正文里夹着 R3 真 GOTCHA（messages-send 的 Safety Constraints、chat-create 的 `--as bot` 两步建群 SOP），压缩风险更高、收益被未读它的题稀释；按单根因纪律，本轮只做 RC-2。RC-1/RC-4 落 lark-shared，越界即被 scope check 拒，且沙箱物理限制非文档可绕——不碰。
-- **选模块理由**：SKILL.md reach=1.0（经 Skill 工具每题必加载），是 RC-2 的唯一承载。改动全部落在它内部，coherent，不触任何别的 skill。
-- **规范经验源补注**：双视角在同一处汇合——
-  - 视角②（annotation）：`skill-annotations.json` 把 L113-122、L123-161、L162-191（API Resources）、L192-231（权限表）全部标 **R0（safe-to-delete）**，理由「method 清单/scope 表 schema/--help 运行时查得到，属 USAGE」。
-  - reviewer 规范背书：optimization-playbook 决策树「是 flag/enum/参数/返回字段/**scope/method 索引** → 不进 skill，交给 --help/schema，最多留一行指针」；authoring-guide 信息归属表「**不写进 skill**：resource/method 全索引、scope/权限映射表（缺权限走 lark-shared 报错流程）」；SKILL.md 锚点 6「`--help`/schema 管 USAGE，reference 只留 gotcha」。三处独立指向同一删除对象。
-  - coverage：3/3 题都加载 SKILL.md（密），token 收益在常驻层可被当轮 eval 直接裁（静态 tiktoken + 每题 visible 构成），不是难裁的拟合型改动。
-
-## 改了什么（逐处）
-- `skills/lark-im/SKILL.md` L113-191 `## API Resources`（per-resource per-method identity/owner/admin/tenant 索引，约 79 行）→ 折叠为 9 行的 `## Native API (beyond shortcuts)`：保留「非 shortcut 的原生 method 仍可调」这条 SELECTION 信号 + 列出哪些 resource 走原生 + 「调用前 MUST 先 `schema`」的指针；删掉每个 method 的逐条 identity/约束枚举（schema 运行时返回）。
-- `skills/lark-im/SKILL.md` L192-231 `## 权限表`（40 行完整 scope 映射表）→ 删除；其语义并入上面 `## Native API` 的指针一句「schema 给 required scope；缺 scope 时 lark-cli 返回 console_url，走 lark-shared 权限流程」。
-- `skills/lark-im/SKILL.md` Shortcuts 速查表新增 2 行：`reactions.*` → `references/lark-im-reactions.md`、`feed.groups.*` → `references/lark-im-feed-groups.md`。**这是路由保命改**：这两个 reference 的唯一运行时入口原本在被删的 API Resources 块里（`[Must-read]` 链接），annotator 误判「已被 Shortcuts 表覆盖」——实测它俩不在原速查表里（速查表的 feed-group 三行指向的是 *-list/-list-item/-query-item 三个不同文件）。不补这 2 行 = 删 reference 链接 = 该 reference reach 永久归 0、路由断裂。
-
-## 为什么这么改（机制）
-- **省 token**：被删的两块是「全量罗列、低命中」的 USAGE——本轮 3 题只用到建群/搜群/搜消息/发消息/转发/@，几十行 per-method identity 与整张 scope 表每次运行都注入却从不被读取。删后 Agent 仍能：(1) 经 SKILL.md 选对命令/身份（SELECTION 层 Identity-and-Token-Mapping、Shortcuts 表全部保留）；(2) 真要调原生 method 时按指针跑 `schema` 拿到 params/identity/scope（运行时事实源，且本来就该查）；(3) 缺 scope 时按 lark-shared 既有报错流程拿 console_url。即「删了 Agent 还做得对吗？做得对就删」（锚点 2）。
-- **不碰 effect**：保留全部 SELECTION 层路由——CRITICAL 先读 lark-shared（L13）、Identity and Token Mapping（user/bot↔token，R3）、完整 Shortcuts 速查表、各域特有 GOTCHA（bot 取不到 sender name、enrichment/download 契约、flag/feed-shortcut 概念）。没有改 identity 路由、没有改参数正确性、没有删 scope 提示语义（指针仍指向 schema+lark-shared 流程）。已经走到「user 授权」这一步的链路不会被碰断。
-- **规范背书**：optimization-playbook §2 决策树 + authoring-guide 信息归属表 L95 + SKILL.md 锚点 6，三处独立判定 method 索引/scope 表「不进 skill，最多留一行指针」——本改动正是把两块 USAGE 折叠成指针。
-
-## 预期效果
-- **成功率（effect 硬门槛）**：不退化。删除的是 USAGE 枚举，保留全部 SELECTION/路由/身份/GOTCHA。本轮 3 题的 FAIL 根因是沙箱 user 授权（RC-1，与本改动正交），改动不触碰授权链路；预期仍为「走到授权步后 blocked」的同构轨迹，不引入新失败。
-- **context（分两层）**：
-  - (1) **静态字数差**：SKILL.md 从 4,960 → 2,986 tok（cl100k_base，reviewer 脚本实测），**-1,974 tok / -39.8%**；落入金标杆带（中位数 ~2,400、lark-shared 2,709），接近上一轮 IM 治理目标 2,040。
-  - (2) **每题运行时 context 方向**：3 题全部下降，且降幅≈静态差——因为 SKILL.md reach=1.0 每题必全量加载，常驻层减重直接等额传导到每题 visible（评测里 SKILL.md 正文 5,722-5,777 tok/题 → 预计降约 2k/题）。**无前置/增读拉力**：没有新增任何会增加 reference 读取的内容；新增的 2 行 Shortcuts 入口只在 agent 实际要用 reactions/feed-groups 时才触发读取（本轮 3 题都不涉及），不构成常驻或额外拉力。与 direction（token↓）一致，无张力。
-- **可裁性**：token 收益在常驻层、可被当轮 eval 直接裁（静态 tiktoken + 每题 visible 构成），非难裁的拟合型改动；无覆盖敞口。
-
-## 刻意没做什么（反 reward-hack / 反过拟合）
-- 没硬编码任何评测题答案；没把 case 特判写进文档；没碰 lark-im 以外任何文件（RC-1/RC-4 的 lark-shared 不动）；没把 RC-3 等无关根因捆进这一轮。
-- **没碰 effect 链路**：没有把 identity 改走 `--as bot`「修绿」（那是 reward-hack，用户显式要「我的身份」、grader 判分点写「当前用户身份」）；没删/弱化 Identity-and-Token-Mapping、Shortcuts 路由、scope 语义指针、CRITICAL lark-shared 前置——这些都是保住「已走到授权」链路不退化的承重内容。
-- **没删 reference 入口**：被删块里两个 reference（reactions/feed-groups）的唯一入口已迁入 Shortcuts 速查表，reach 不归零、路由不断裂（纠正了 annotator「已覆盖」的误判）。
-- **没做输出裁剪、没碰命令行为**（T1 docs-only，且 playbook 红线：输出裁剪须独立设计验证）。
-- **没补「前置授权说明」**：诊断证据显示 3 题调用前都已读到 SKILL.md（reach=1.0），失败在更上游的沙箱授权（状态③语义、根因是环境），前置救不了且只会增 token，与目标背道——明确不做。
-- 这是「减体积」改动、与评测错误分布无拟合关系，不存在朝错误分布过拟合的敞口；lite 无 sealed 也不构成隐患。
-
-## 签名
-- signature: a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e（git diff skills/lark-im/SKILL.md 内容哈希）  tier: T1

harness-opt/rounds/round-001/trend.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-001/workorder.md

@@ -1,35 +0,0 @@
-# Round 1 归因派工单（parent=phi0；模块未定，由 candidate-writer 据诊断点名）
-
-> **只读输入**——opt-attributor 读本文件，把诊断**另写** `diagnosis.md`（给 candidate-writer）+ 逐题结构化 `attribution.json`（给 dashboard）。**不要覆盖本文件**，留作派工单↔诊断的前后对比。
-> 判分点只当「什么算挂」的锚，禁止照抄 grader 药方（已从派工单剔除）。
-
-## 模块运行时可达性（选模块第一步的证据；要选须在 strategy.md 说明理由）
-> reach=**实测**触达率（域主 SKILL.md 经 Skill 工具加载、reference 经 Read，都从 trace 实测，没有恒在的面）；判决集=实测∪预期触达。**实测低但有预期触达 ⚠️=可发现性/路由根因**（本该读却没读，如没路由到该域 / 速查表漏链接 / 该前置），正该选来修——不是白烧；reach=0 且无预期 才是真白烧。 **别用「全集均摊」判 reference 价值**：判决在 reach 子集上做，压一条 reference 的降幅在它子集里不被没读它的题稀释——reach 不高(但 >0)的 reference 在自己子集上也可能越带。
-- `skills/lark-im/SKILL.md` → reach=1.0  [域主 skill·经 Skill 工具加载]；判决集(实测∪预期): ['1', '2', '3']；其中挂的: ['2', '3']
-- `skills/lark-im/references/lark-im-chat-create.md` → reach=0.667；判决集(实测∪预期): ['1', '3']；其中挂的: ['3']
-- `skills/lark-im/references/lark-im-messages-send.md` → reach=0.667；判决集(实测∪预期): ['1', '3']；其中挂的: ['3']
-- （另 22 个 reference reach=0 且无预期触达，本轮无关，略）
-
-### 2  [FAIL]  ctx=34616  (acc=274168)  52787ms  tools=25
-- session.jsonl: harness-opt/baseline/runs/run-1/detail_info/cases/CLI_核心评测_015/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✓ 成功定位名为「fusanming_at_openclaw群」的群，并获取最近包含「飞豆」关键字的消息。
-    ✓ 将筛选出的相关消息内容转发到「fusanming_at_需求测试群」。
-    ✓ 在「fusanming_at_需求测试群」中 @傅六铭 做知会，消息发送成功。
-
-### 3  [FAIL]  ctx=31289  (acc=225396)  46776ms  tools=22
-- session.jsonl: harness-opt/baseline/runs/run-1/detail_info/cases/CLI_核心评测_080/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✓ 使用用户身份创建一个名为「今晚吃什么」的群，预期返回 chat_id
-    ✓ 创建一张飞书卡片，卡片内容包含「今天晚上吃什么」
-    ✓ 将该卡片发送到新建群中，预期返回 message_id
-
-### 1  [PASS]  ctx=30086  (acc=292379)  51004ms  tools=32
-- session.jsonl: harness-opt/baseline/runs/run-1/detail_info/cases/CLI_核心评测_014/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✗ 使用当前用户身份创建名为「IM合作群」的群聊
-        证据: Agent 执行了 split-flow 授权流程以获取 user 身份权限，生成了二维码让用户扫描，但用户未完成授权即要求评分。Auth status 显示 'User identity: missing'，群聊未被创建。
-    ✗ 将傅一铭和傅二铭加入该群
-        证据: 依赖群聊创建结果。由于群聊未创建（blocked by user identity missing），无法添加成员。
-    ✗ 在该群发送文本消息「大家体验有问题随时沟通」，并返回可验证的 chat_id / message_id
-        证据: 依赖群聊创建结果。由于群聊未创建，无法发送消息。

harness-opt/rounds/round-002/attribution.json

@@ -1,65 +0,0 @@
-[
-  {
-    "case_id": "1",
-    "case_label": "CLI_核心评测_014",
-    "verdict": "PASS",
-    "verdict_note": "workorder=PASS（聚合口径），判分点证据 3/3 ✗ → 实质 FAIL，按判分点当 FAIL 归因",
-    "token": 34555,
-    "token_visible_est": 17364,
-    "duration_ms": 37000,
-    "tool_calls": 8,
-    "cmd_attempts": 7,
-    "cmd_failures": 5,
-    "cmd_fail_rate": 0.71,
-    "discoverability_state": "③ 读了仍卡（SKILL.md+chat-create.md 调用前已读；卡在跨域 contact + 沙箱 user 授权，非 lark-im 内容/触达问题）",
-    "axis": "效果",
-    "root_cause": "沙箱不能交互扫码完成 user 授权 + 跨 lark-contact 域 search-user 不可用——无 lark-im 文档根因，本轮不可修",
-    "token_hotspot": "SKILL.md 常驻正文(RC-1) + chat-create.md 按需读取(RC-3，本题读了但授权阻断没用上)；无 lark-cli 输出离群",
-    "token_reliability": "常驻静态(SKILL.md 3751) + 按需读取(chat-create.md 3062)",
-    "duration_hotspot": "多轮交互(查联系人→切contact→失败→auth status→授权→qrcode重试) + 反应式重试(qrcode 路径)",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "无 lark-im 文档可修点（效果根因在环境+跨域）；lark-im 侧仅 token 减法（SKILL.md 常驻、chat-create.md 体积）"
-  },
-  {
-    "case_id": "2",
-    "case_label": "CLI_核心评测_015",
-    "verdict": "PASS",
-    "verdict_note": "真 PASS，判分点 3/3 ✓，全程 bot 身份完成，无授权阻断（推翻 round-1 的 blocked 定调）",
-    "token": 54568,
-    "token_visible_est": 43760,
-    "duration_ms": 125000,
-    "tool_calls": 16,
-    "cmd_attempts": 9,
-    "cmd_failures": 3,
-    "cmd_fail_rate": 0.33,
-    "discoverability_state": "① 从没读（chat-messages-list.md / messages-search.md 调用前从没读，直接猜命令→全量拉取+exit2）",
-    "axis": "token",
-    "root_cause": "`+chat-messages-list --page-all` 无时间过滤全量拉取→43.5KB持久化→Read 灌入 22556 tok；放大器是 chat-messages-list.md 没被读到（缺收窄指引），但补它与降token目标方向冲突",
-    "token_hotspot": "工具返回原样输出（block #19 Read 持久化文件 22556 tok，51.5%，非 lark-im doc）",
-    "token_reliability": "单次输出（强依赖该群消息量，非稳定常驻热点，单题不可外推）",
-    "duration_hotspot": "多轮交互 + 重试（messages-search 连环 exit2→改 page-all→大输出→多次本地 grep 抠数据）",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现；工具调用 16 明显高于 080，作旁证",
-    "doc_fix_hint": "token 黑洞来自工具输出非文档；SKILL.md 表对 chat-messages-list 未提示大群应 server-side 收窄——但补此为增内容，与降token冲突，列观察项不作本轮根因"
-  },
-  {
-    "case_id": "3",
-    "case_label": "CLI_核心评测_080",
-    "verdict": "PASS",
-    "verdict_note": "真 PASS，判分点 3/3 ✓，主动选 bot 身份完成建群+发卡片，零命令失败（推翻 round-1 的 blocked 定调）",
-    "token": 38009,
-    "token_visible_est": 21599,
-    "duration_ms": 47000,
-    "tool_calls": 6,
-    "cmd_attempts": 3,
-    "cmd_failures": 0,
-    "cmd_fail_rate": 0.0,
-    "discoverability_state": "③ 读了即用（SKILL.md+chat-create.md+messages-send.md 调用前全读到且用上，无触达问题）",
-    "axis": "token",
-    "root_cause": "messages-send.md 单文件 5365 tok（内部 4 处『选 content flag』语义重叠 + Commands 全形态罗列）+ SKILL.md 常驻 + chat-create.md 按需——纯减体积场景，命令零失败",
-    "token_hotspot": "运行时冗余清单常驻 + 按需 reference 偏大（读取 Skill 56.4%：messages-send.md 5365 + SKILL.md 3751 + chat-create.md 3060）",
-    "token_reliability": "常驻静态(SKILL.md 3751) + 按需读取(messages-send.md 5365 子集reach0.333、chat-create.md 3060 子集reach0.667)",
-    "duration_hotspot": "无离群（47s 正常建群+发卡片串行，无重试、无写后回查）",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "messages-send.md 选型规则在 4 处重复表述、Commands 罗列全部媒体形态；SKILL.md Important Notes/Shortcuts 全量低命中常驻——均为可删的减法冗余，本题 token 杠杆最高且无 effect 风险"
-  }
-]

harness-opt/rounds/round-002/case-commands.json

@@ -1,27 +0,0 @@
-{
-  "1": [
-    "auth login",
-    "auth qrcode",
-    "contact +search-user"
-  ],
-  "3": [
-    "auth login",
-    "auth qrcode",
-    "auth status",
-    "im +chat-create",
-    "im +messages-send"
-  ],
-  "2": [
-    "auth login",
-    "auth qrcode",
-    "auth status",
-    "im +chat-messages-list",
-    "im +chat-search",
-    "im +messages-mget",
-    "im +messages-search",
-    "im +messages-send",
-    "im messages forward",
-    "schema im.messages.forward",
-    "schema im.messages.search"
-  ]
-}

harness-opt/rounds/round-002/child-cache.json

@@ -1,11 +0,0 @@
-{
-  "3": {
-    "score": 1.0,
-    "passed": true,
-    "context_window": 35478,
-    "token_usage": 221685,
-    "duration_ms": 46540,
-    "tool_call_count": 22,
-    "feedback": "所有核心目标均达成。执行者经历了两次试错（shell 引号问题、@file 语法不支持），但均自行修正并成功完成任务，符合合理的调试流程。群创建、卡片创建、消息发送三个决策点全部通过。卡片内容准确包含「今天晚上吃什么」文字，message_id 成功返回。\n- {'reason': '参数文档改进: --content 参数应明确标注不支持 @file 语法，避免 AI 重复试错'}\n- {'reason': '引导性错误: 当检测到 @/path 模式时，错误提示应建议正确的替代参数（如 --file）'}\n- {'reason': '防御性设计: 在 SKILL.md 补充大型 JSON 内容的分段写入指引，减少因引号转义导致的失败'}"
-  }
-}

harness-opt/rounds/round-002/diagnosis.md

@@ -1,113 +0,0 @@
-# Round 2 归因（parent=round-1 已采纳候选 51f2a70e；候选模块见 candidate_modules，由 candidate-writer 据诊断+reach 点名）
-
-> 目标（objective.json）：**在不回退成功率的前提下降低 lark-im skill 文档的 token 成本**。effect 是硬门槛、不可退化；token 与 duration 是并列成本杆。tier=T1，仅可改 `skills/lark-im/**`。
-> 判分点只当「什么算挂」的锚，不抄 grader 药方。
-> **本轮 trace = round-1 已采纳候选（51f2a70e，SKILL.md 已 trim 到约 3,915 tok）的行为**，不是 baseline。三题 session 实测已确认 SKILL.md 注入正文为 3,751 tok/题（与 trim 后体积一致），round-1 报告的 5,722 tok/题是 trim 前数字，已过期。
-
-## ⚠️ 对 round-1 定调的关键修正（先看，影响整轮方向）
-
-round-1 把三题一律定调为「user 身份授权在沙箱内不可完成 → 全部 blocked」。**实测 trace 推翻了这个 monolith：三题行为完全不同，只有 1 题真卡授权。**
-
-| case | round-1 说法 | 实测 trace 真相 | verdict（workorder） |
-|---|---|---|---|
-| 1 (014) | blocked by user auth | ✅ **确认**：需 `contact +search-user` 解析 open_id（跨 lark-contact 域）→ bot exit2 → user token_missing → 发起 qrcode → 停在扫码。真授权阻断 | PASS（聚合口径；判分点证据全 ✗，**实质 FAIL**） |
-| 2 (015) | blocked by user auth | ❌ **证伪**：全程 `identity:bot`，从未卡授权。搜群✓、定位「飞豆」消息✓、转发✓、@傅六铭✓，两次 `messages-send` 全 `ok:true`。**任务完整完成** | PASS（判分点 3/3 ✓，真 PASS） |
-| 3 (080) | blocked by user auth | ❌ **证伪**：`auth status` 看到 bot ready → **主动选 bot 身份** → 建群✓（`ok:true`）→ 发卡片✓（`ok:true`）。**任务完整完成** | PASS（判分点 3/3 ✓，真 PASS） |
-
-**含义**：本轮 effect 实际是 **2 真 PASS + 1 实质 FAIL**，不是 round-1 描述的「三题全 blocked」。effect 信号是 **auth-noise 主导**（014 卡在沙箱不能扫码 + 跨域 contact，非 lark-im 文档可修；015/080 已绿）。降 token 时**必须保住 015/080 现在走通 bot 身份的链路**——这两题恰好是被 reference 真正喂到、且已成功的题，乱删 reference 里的 identity/参数说明最可能误伤它们。
-
-## 跨 case 共同根因（优先看；按对 TOKEN 目标的杠杆排序）
-
-### RC-1（token，头号抓手，3 题全命中、最稳）—— SKILL.md `## Important Notes` + Shortcuts 全表常驻，本轮任务低命中
-- **现象**：SKILL.md 经 Skill 工具每题必加载（reach=1.0），实测 3,751 tok/题、三题一致（常驻静态）。但其中大段与本轮 3 题（建群 / 搜群+搜消息+转发+@ / 建群+发卡片）无关：
-  - `## Important Notes`（L36–85，约半个文件）：Sender Name Resolution、message enrichment、`--download-resources`、Card Messages 限制、Flag 两层、Feed Shortcut 限制——本轮**一条都没用到**，却每题常驻。
-  - `## Shortcuts` 全表（L91–114）逐条列 20+ shortcut，含 flag/feed-group/feed-shortcut/reactions 等本轮完全不相关项。
-- **可信度=常驻静态**：tiktoken 可测、跨题稳定（3,751×3）。这是降 token 最稳的发力点，且 3 题全命中（reach=1.0），降幅不被任何子集稀释。
-- **axis=token**。文档位置：`skills/lark-im/SKILL.md` 的 `## Important Notes` 低命中小节 + `## Shortcuts` 全量表。
-- **方向张力（必须标注）**：这是 round-1 已经动过一刀的同一文件（折叠了 API Resources/权限表）。再压 Important Notes/Shortcuts 是**同向继续**，但**剩余内容大多是 identity/约束类**——删错会碰坏 015/080 已走通的 bot 身份判断。candidate-writer 取舍时这是 effect 风险点，不是 RC-1 不成立。
-
-### RC-2（token，次级抓手，080 命中、按需读取）—— `messages-send.md` 单文件偏大且内部高度冗余
-- **现象**：080 读了 `messages-send.md`，实测 **5,365 tok**——本轮所有按需 reference 里最大的单块（占 080 visible 的 24.8%）。该 reference 实测被读且**确实用上了**（080 据此发卡片成功），不是「读了没用」。
-- **从文档看为何这么大**：messages-send.md（264 行）内部「怎么选 content flag」重复表述 4 处——`## Choose The Right Content Flag`(L23–42)、`## What --markdown Really Does`(L44–92)、`## Preserving Formatting`(L94–112)、`## Common Mistakes`(L192–201)语义大量重叠；`## Commands`(L114–161) 15+ 例覆盖 image/file/video/audio/idempotency 等本轮用不到的形态。这是「单文件冗余 + 全形态罗列」，不是信息缺失。
-- **可信度=按需读取**：只在实读它的子集（reach=0.333，仅 080）里计入，压缩降幅在该子集不被稀释——但**子集只有 1 题**，证据基数小，效果需评测确认（见数据缺口）。
-- **axis=token**。文档位置：`skills/lark-im/references/lark-im-messages-send.md`。
-
-### RC-3（token，次级抓手，014+080 命中、按需读取）—— `chat-create.md` 按需读取偏大
-- **现象**：014 与 080 都读了 `chat-create.md`，实测 3,060–3,062 tok（reach=0.667）。080 据此建群成功（用上了）；014 读后因 user 授权阻断没走到建群（读了但本题没用上）。
-- **可信度=按需读取**（reach=0.667，子集 2 题）。体积本身不离群，杠杆低于 RC-2，列为更次级。
-- **axis=token**。文档位置：`skills/lark-im/references/lark-im-chat-create.md`。
-
-### RC-4（效果，无文档根因 / 本轮不可修）—— 014 的 user 授权阻断 + 跨域 contact 依赖
-- **现象**：014 需先解析「傅一铭/傅二铭」open_id，走 `contact +search-user`（**lark-contact 域，不在 candidate_modules**）：bot 身份 exit2（invalid_argument）→ `--as user` token_missing → 发起 `auth login`+qrcode → 停在扫码。判分点证据全 ✗。
-- **归因落点**：根因=沙箱不能交互扫码（环境）＋ 跨域 contact 命令不可用（非 lark-im）。**lark-im 文档侧无根因、无可修点**——这正是约束 3 的「无文档根因 / 本题不改」出口，不要为凑根因往 lark-im doc 上硬编。
-- **axis=效果**，标注**无文档根因 / 本轮不改**。effect 维持 baseline 即可，不要试图改路由让 014「修绿」（用户显式要本人身份解析联系人，改 bot 是 reward-hack）。
-
-## 命令失败热点（跨 case；失败类型由我从 timeline 命令串读出，非判决数字）
-
-| lark-cli 命令 | 失败次数 | 涉及题数 | 主要失败类型 | 指向的文档问题 |
-|---|---|---|---|---|
-| `contact +search-user` | 4 | 1 (014) | bot exit2(invalid_argument) ×2；user token_missing ×2 | **跨 lark-contact 域**，非 lark-im 内容 |
-| `auth qrcode --output 绝对路径` | 1 | 1 (014) | unsafe output path，改相对路径重试成功 | 路径约束在 lark-shared（不可改） |
-| `im +messages-search` | 2 | 1 (015) | exit2（bot 身份 + `--as user` 均 exit2） | 见下「messages-search 难用」分析 |
-| `im +chat-messages-list --page-all` | 1 | 1 (015) | exit2（无过滤 page-all） | 见下「015 token 黑洞」分析 |
-- **解读**：本轮**没有一条 lark-im 命令因「参数名/类型写错」系统性失败**。080 三条命令 0 失败；015 的失败集中在 `messages-search`（见下）。这意味着**没有 lark-im 侧的常规「报错/参数整形」工单**——与 RC-1/2/3 的 token 方向一致，本轮抓手是减体积不是补内容。
-
-### 015 的 token 黑洞（重要的新发现，round-1 完全没诊断到）
-- 015 真正的 token 大头**不是任何 lark-im doc**，而是 **block #19：一次 `Read` 工具读入 22,556 tok（占该题 visible 51.5%）**。成因链：#17 `+messages-search` exit2 → 退而求其次 #18 `+chat-messages-list --page-all`（无时间过滤）→ 输出 43.5KB 被持久化到文件 → agent `Read` 整个文件 → 22.5k tok 灌进上下文。后面又靠本地 `grep`(#27–33) 抠出「飞豆」两条。
-- **从文档角度**：`chat-messages-list.md` **本题 reach=0**（没读到），而它恰好写了 `--start/--end` 时间过滤、`--page-size`、「无 sender 排序」等能避免全量拉取的约束（L20–52）。SKILL.md 表里对该 shortcut 只写「supports time range/sort/pagination」一句、未提示「大群全量拉取会爆上下文、应先 server-side 收窄」。**这是一个真实的「该读没读 → 全量灌入」放大器**（约束 5 状态①：调用前从没读该 reference）。
-- **但这条对本轮目标是「方向张力」，不是干净的 token 抓手**：要避免全量灌入，文档侧只能**增加**收窄指引（前置或加 caution），这与「降 token」的常驻成本目标**方向相反**（见硬性约束 7 的冲突记录）。且 22.5k 黑洞是**单次工具输出**（单次输出可信度、单题、强烈依赖该群消息量），不是稳定常驻热点。**结论：列为观察项交评测裁决，不要当成 RC-1 那种干净抓手去推「前置 chat-messages-list」——很可能只增 token 不省。**
-
-## 可发现性时序（约束 5 三态；判「前置能不能救」的决定性证据）
-> 对每条相关 reference / `--help`，按相对首次失败调用的读取时序统计。`--help` 扫 Bash（本轮 3 题均未跑任何 `--help`）。
-
-| reference / `--help` | 聚合 reach | ①从没读 | ②失败后才读 | ③读了仍错/卡 | 主导态 → 改动方向 |
-|---|---|---|---|---|---|
-| `lark-shared/SKILL.md` | 1.0 | 0 | 0 | — | 三题调用前都读了；014 仍卡（环境，非内容）；不可改 |
-| `chat-create.md` | 0.667 | 0 | 0 | — | 080 调用前读→建群成功；014 调用前读→授权阻断（非 reference 错）。**非触达问题** |
-| `messages-send.md` | 0.333 | 0 | 0 | — | 080 调用前读→发卡片成功。**非触达问题** |
-| `chat-messages-list.md` | 0.0 | 1 (015) | 0 | — | ① **015 调用前从没读**→直接 `--page-all` 全量拉取→token 黑洞。触达缺口，但补它=增 token，与目标冲突（见上） |
-| `messages-search.md` | 0.0 | 1 (015) | 0 | — | ① 015 从没读 messages-search.md，直接猜 `+messages-search` ×2 → exit2。该命令 user-only（SKILL 表 L101 已注明），bot 身份必败 |
-- **结论**：本轮 effect 失败的唯一真题（014）是**状态③语义但根因是环境**（内容已触达、卡在沙箱授权+跨域），**前置/补内容救不了**。015 的两处 ① 触达缺口（chat-messages-list / messages-search 没读）确实存在，但**修它们的方向（增内容）与本轮 token 目标相反**，且 015 最终已 PASS（靠 bot + 本地 grep 兜底）——所以这两处**不是必须修的 effect 缺口，只是 token 放大器**，且修了大概率反而增 token。
-- **对 candidate-writer 的含义**：**本轮没有「该前置」的干净 case**。RC-1/2/3 都是「调用前已读、内容够用 → 减体积」的纯 token 减法，不涉及触达。不要被 015 的两处 ① 诱导去推前置——那会与目标背道而驰。
-
-## 方向冲突记录（硬性约束 7）
-- **减体积（RC-1/2/3，与 objective.direction 同向）** vs **补收窄指引（修 015 chat-messages-list 全量灌入，与 objective 反向）**：前者降常驻/按需 token，后者为省「单次工具输出」反而要**增**文档常驻 token。两者方向相反，**不可合并**。本轮目标是降 token，应取减体积一侧；015 的全量灌入作为观察项记录、不作为本轮要补的内容根因。
-
-## 差距台账复盘
-- 无（round 2，`discard-ledger.json` 为空，无已跑未采纳候选）。
-
-## 逐 case
-
-### 1 (014) [workorder=PASS / 实质 FAIL] token=34555(reported)/visible 17,364 耗时=37s 命令失败率≈5/7 维度=效果(不可修)
-- 判分点结果：3 条全 ✗——建群/拉人/发消息全未发生，卡在 `contact +search-user` 解析 open_id（user 授权阻断）。verdict=PASS 系聚合口径，按判分点证据当 FAIL 处理。
-- 命令失败：≈5/7。`contact +search-user` bot exit2 ×2、user token_missing ×2；`auth qrcode` 绝对路径 unsafe ×1（改相对路径成功）。**全部非 lark-im 命令的内容错误**。
-- 可发现性时序：调用前读了 SKILL.md(reach=1.0)+chat-create.md(3,062 tok)；失败在更上游的跨域 contact + 授权。**非 lark-im 触达问题**。
-- token 归因：SKILL.md 正文 3,751（常驻静态，21.6%）+ chat-create.md 3,062（按需，17.6%，本题没走到建群=读了没用上）+ 系统 Skill 列表注入 4,612（固定开销，不归因）。lark-cli 命令累计含多次短失败回显，单条都短、非热点。
-- 耗时归因：本题往返多（查联系人→切 contact→失败→auth status→授权→qrcode 重试）。多为授权链路 + 跨域固有串行 + 反应式重试（duration 弱信号，需多轮复现）。
-- 文档根因：效果=沙箱 user 授权 + 跨域 contact（环境，**无 lark-im 文档根因，本轮不改**）；token=SKILL.md 常驻（RC-1）+ chat-create.md 按需（RC-3）。
-
-### 2 (015) [PASS·真] token=54568(reported)/visible 43,760 耗时=2m5s 命令失败率≈3/9 维度=token
-- 判分点结果：3/3 ✓——定位群、转发「飞豆」消息、@傅六铭知会全部成功（两次 `messages-send` 均 `ok:true`）。**全程 bot 身份，无授权阻断**。
-- 命令失败：≈3/9。`+messages-search` bot exit2、`+messages-search --as user` exit2、`+chat-messages-list --page-all` exit2（无过滤）；agent 退到 `+chat-messages-list`(无 page-all) + 本地 grep 兜底成功。
-- 可发现性时序：① `messages-search.md` / `chat-messages-list.md` **调用前从没读**（reach=0），直接猜命令。messages-search 是 user-only（SKILL 表 L101 已注明）、bot 身份必败——agent 没看清就猜。
-- token 归因：**本题 token 大头不是 lark-im doc**，是 block #19 一次 `Read` 持久化文件 = **22,556 tok（51.5%，其他工具调用/返回）**，成因=`--page-all` 无过滤全量拉取→43.5KB→Read 灌入（单次输出可信度，强依赖该群消息量）。SKILL.md 正文 3,749（常驻）。lark-shared 3,749（跨 skill，不归因 lark-im）。
-- 耗时归因：本题最长(2m5s)，主因是 messages-search 连环失败→改用 page-all→大输出→多次本地 grep 抠数据的多轮往返（duration 弱信号；工具调用 16 raw32，明显高于 080，作旁证）。
-- 文档根因：token 黑洞的放大器=`chat-messages-list.md` 没被读到 + SKILL.md 表未提示大群应 server-side 收窄——但**补这条与降 token 目标相反**（方向张力，见上），列为观察项；本题已 PASS。常规 token 抓手仍是 RC-1（SKILL.md 减体积）。
-
-### 3 (080) [PASS·真] token=38009(reported)/visible 21,599 耗时=47s 命令失败率=0/3 维度=token
-- 判分点结果：3/3 ✓——`auth status` 见 bot ready→主动选 bot→建群`ok:true`→发 interactive 卡片`ok:true`。**任务完整完成，零命令失败**。
-- 命令失败：0/3。三条 lark-cli（auth status / chat-create / messages-send）全成功。
-- 可发现性时序：调用前读 SKILL.md + chat-create.md(3,060) + messages-send.md(5,365)，全部状态③（调用前已读且用上）。**无触达问题**。
-- token 归因：**本题是纯 token 抓手题**——读取 Skill 占 56.4%：messages-send.md 5,365（按需，最大单块，RC-2）+ SKILL.md 3,751（常驻，RC-1）+ chat-create.md 3,060（按需，RC-3）。三块 reference/SKILL 都实读且 RC-2 的 messages-send.md 确实用上了。系统 Skill 列表注入 4,612（固定开销，不归因）。
-- 耗时归因：47s，全部为正常建群+发卡片串行，无重试、无写后回查（无离群）。
-- 文档根因：无效果根因（已绿）；token=RC-2(messages-send.md 内部冗余) + RC-1(SKILL.md 常驻) + RC-3(chat-create.md)。**本题 token 杠杆最高且无 effect 风险**（命令全成功，减 reference 体积不碰已走通链路）。
-
-## 给 candidate-writer 的收口（不含具体改法）
-- **唯一在 T1 内可合法发力的轴是 token**，且本轮是**纯减体积**场景（无触达缺口要补、无参数错误要改）：
-  - **RC-1**（SKILL.md `## Important Notes` 低命中小节 + `## Shortcuts` 全表）：3 题全命中、常驻静态、最稳，但剩余多为 identity/约束类，删错会碰坏 015/080 已走通的 bot 身份判断——**effect 风险点**。
-  - **RC-2**（messages-send.md 内部 4 处「选 content flag」语义重叠 + 全形态 Commands）：单文件最大块、内部冗余明确，但子集只有 080 一题（reach=0.333），证据基数小、效果需评测确认。
-  - **RC-3**（chat-create.md 按需偏大）：杠杆最低，列为更次级。
-- **effect 不可在本轮 T1 内合法抬升**：014 是环境（沙箱不能扫码）+ 跨域 contact，无 lark-im 文档根因。015/080 已真 PASS。候选必须**保住 015/080 走通 bot 身份的 identity/参数说明**，降 token 时别误伤。
-- **不要推前置**：本轮没有「该前置」的干净 case。015 的两处触达缺口（chat-messages-list/messages-search 没读）虽真实存在，但修它们=增内容，与降 token 目标**方向冲突**，且 015 已 PASS——属观察项，非本轮要补的根因。
-- **缺失信息（doc_fix_hint 语气）**：SKILL.md 的 Important Notes/Shortcuts 全量罗列、本轮低命中却每题常驻；messages-send.md 同一选型规则在 4 处重复表述、Commands 罗列全部媒体形态——这类「全量/重复、低命中」内容是 token 的主要去处，且是减法（删冗余）而非加法。
-- **数据缺口**：(a) workorder 三题 verdict 全 PASS，但 014 判分点证据全 ✗——归因按判分点当 FAIL 处理，effect 实际是 2 真 PASS + 1 实质 FAIL。(b) RC-2/RC-3 子集小（messages-send.md 仅 080、chat-create.md 仅 014+080），单轮证据基数小，token 降幅需评测在子集上确认。(c) 015 的 22.5k 黑洞是单次工具输出，强依赖该群消息量，非稳定常驻热点，单题不可外推。(d) duration 三题波动大（37s/2m5s/47s），015 长尾主因是 messages-search 连环失败+大输出多轮抠数据，但单轮不足以定论，需多轮复现；工具调用数(8/16/6 model calls)可作比 wall-clock 稳的旁证。(e) 工具调用次数 session-analyze(model calls 8/16/6) 与 workorder 趋势表(R1 均值 26.3) 口径不一致，趋势表疑似含 raw 计数，旁证以 timeline 实际往返为准。

harness-opt/rounds/round-002/discard-ledger.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-002/failure-memory.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-002/module-reach.json

@@ -1,220 +0,0 @@
-{
-  "skills/lark-im/SKILL.md": {
-    "reach": 1.0,
-    "read_cases": [
-      "1",
-      "2",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "2",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": true
-  },
-  "skills/lark-im/references/lark-im-chat-create.md": {
-    "reach": 0.667,
-    "read_cases": [
-      "1",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-identity.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-messages-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-search.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-update.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-list-item.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-query-item.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-groups.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-create.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-remove.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-cancel.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-create.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-message-enrichment.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-mget.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-reply.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-resources-download.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-search.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-send.md": {
-    "reach": 0.333,
-    "read_cases": [
-      "3"
-    ],
-    "actual_cases": [
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-reactions.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-threads-messages-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  }
-}

harness-opt/rounds/round-002/review.json

@@ -1,15 +0,0 @@
-{
-  "generated_by": "lark-cli-harness:opt-reviewer",
-  "verdict": "PASS",
-  "module": "skills/lark-im/references/lark-im-messages-send.md",
-  "tier": "T1",
-  "reason": "纯结构性去重：16407→6399 字节(-61%)与策略一致；逐项核对每条承重指令(互斥规则、video-cover 必配、cwd-relative/绝对路径拒绝、markdown→post 边界、三套 <at> 语法、content 各 msg_type 样例、Safety Constraints、identity+scope 映射)均原样保留在新文档内联，删的全是重复/过度罗列(4× 选型规则、镜像 --help 的 Parameters 表、Common Mistakes、Notes、冗余 Commands)。无硬编码评测答案、未针对 080 卡片流窄化、未碰 SKILL.md 身份路由、单文件单根因。",
-  "dimensions": {
-    "reward_hack":      {"pass": true,  "evidence": "无硬编码 eval ID/答案(仅 oc_xxx/ou_xxx 等通用占位符，与原文一致)；card/interactive+bot 身份路径保留为通用指引，未按 080 卡片流做特判或窄化"},
-    "semantic_regress": {"pass": true,  "evidence": "逐条核对：互斥/video-cover/cwd-relative+绝对路径拒绝/markdown→post/三套 <at>/content 全 msg_type 样例/Safety/identity+scope 全部内联保留；仅删除的是真重复(dry-run 占位符细节、JSON wrap 示意、img_/file_ 自动识别说明)，非承重 guardrail，且运行时可观测"},
-    "token_shift":      {"pass": true,  "evidence": "真减 10008 字节常驻；--help 指针是 additive 补充(指向真实存在且 --help 已含互斥/video-cover/路径规则)，承重 gotcha 全留内联，080 不需额外调 --help 即可恢复，无运行时增读拉力。注:work-order 提的 schema im.messages.create 方法不存在，但文档本身不指向 schema，不构成运行时陷阱"},
-    "contract_break":   {"pass": true,  "evidence": "T1 文档不涉对外契约；prerequisite 链接目标存在、章节结构完整、无其他文件深链到被删 anchor(Media Input Rules/Common Mistakes 命中在 messages-reply.md 而非本文件)"},
-    "devguide":         {"pass": true,  "evidence": "符合 reference 收敛到 gotcha-only、不镜像 --help 的优化方向；同一事实只写一处，删的两类(语义回退/承重删除)均未触发——优化红线两维过关"},
-    "single_root_cause":{"pass": true,  "evidence": "commit 仅 1 文件 51 insert/208 delete，全部服务 RC-2(单文件重复表述去重)一个根因；未捆 RC-1(SKILL.md)/RC-3(chat-create)，未把无关删除以 token 对冲缝入"}
-  }
-}

harness-opt/rounds/round-002/round.json

@@ -1,380 +0,0 @@
-{
-  "round": 2,
-  "status": "admitted",
-  "parent_id": "a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e",
-  "parent_worktree": "/Users/bytedance/Projects/cli",
-  "child_worktree": "/Users/bytedance/Projects/cli",
-  "base_commit": "51f2a70e6dffeea65d928badb6207408490dc215",
-  "module": "skills/lark-im/references/lark-im-messages-send.md",
-  "candidate_modules": [
-    "skills/lark-im/SKILL.md",
-    "skills/lark-im/references/lark-im-chat-create.md",
-    "skills/lark-im/references/lark-im-chat-identity.md",
-    "skills/lark-im/references/lark-im-chat-list.md",
-    "skills/lark-im/references/lark-im-chat-messages-list.md",
-    "skills/lark-im/references/lark-im-chat-search.md",
-    "skills/lark-im/references/lark-im-chat-update.md",
-    "skills/lark-im/references/lark-im-feed-group-list-item.md",
-    "skills/lark-im/references/lark-im-feed-group-list.md",
-    "skills/lark-im/references/lark-im-feed-group-query-item.md",
-    "skills/lark-im/references/lark-im-feed-groups.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md",
-    "skills/lark-im/references/lark-im-flag-cancel.md",
-    "skills/lark-im/references/lark-im-flag-create.md",
-    "skills/lark-im/references/lark-im-flag-list.md",
-    "skills/lark-im/references/lark-im-message-enrichment.md",
-    "skills/lark-im/references/lark-im-messages-mget.md",
-    "skills/lark-im/references/lark-im-messages-reply.md",
-    "skills/lark-im/references/lark-im-messages-resources-download.md",
-    "skills/lark-im/references/lark-im-messages-search.md",
-    "skills/lark-im/references/lark-im-messages-send.md",
-    "skills/lark-im/references/lark-im-reactions.md",
-    "skills/lark-im/references/lark-im-threads-messages-list.md"
-  ],
-  "module_reach": {
-    "skills/lark-im/SKILL.md": {
-      "reach": 1.0,
-      "read_cases": [
-        "1",
-        "2",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "2",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": true
-    },
-    "skills/lark-im/references/lark-im-chat-create.md": {
-      "reach": 0.667,
-      "read_cases": [
-        "1",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-identity.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-messages-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-search.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-update.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-list-item.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-query-item.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-groups.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-cancel.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-create.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-message-enrichment.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-mget.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-reply.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-resources-download.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-search.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-send.md": {
-      "reach": 0.333,
-      "read_cases": [
-        "3"
-      ],
-      "actual_cases": [
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-reactions.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-threads-messages-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    }
-  },
-  "expected_reach": {},
-  "minibatch": [
-    "1",
-    "2",
-    "3"
-  ],
-  "pareto_cases": [
-    "1",
-    "2",
-    "3"
-  ],
-  "artifacts": {
-    "workorder": "workorder.md",
-    "diagnosis": "diagnosis.md",
-    "attribution": "attribution.json",
-    "strategy": "strategy.md",
-    "review": "review.json",
-    "trend": "trend.json"
-  },
-  "code_tip": "82a099feafb45d101116f10230ce7c2f92fbcfe5",
-  "signature": "557349b40feb359bb791749a37571d59edb7e72e",
-  "tier": "T1",
-  "intent": "consolidate 4x repeated content-flag rule + compress media enumeration & --help-mirror sections in messages-send.md (token, no capability removed)",
-  "target_axis": "token",
-  "changed_files": [
-    "skills/lark-im/references/lark-im-messages-send.md"
-  ],
-  "decision_basis": {
-    "type": "module",
-    "module": "skills/lark-im/references/lark-im-messages-send.md"
-  },
-  "decision_cases": [
-    "3"
-  ],
-  "review": {
-    "generated_by": "lark-cli-harness:opt-reviewer",
-    "verdict": "PASS",
-    "module": "skills/lark-im/references/lark-im-messages-send.md",
-    "tier": "T1",
-    "reason": "纯结构性去重：16407→6399 字节(-61%)与策略一致；逐项核对每条承重指令(互斥规则、video-cover 必配、cwd-relative/绝对路径拒绝、markdown→post 边界、三套 <at> 语法、content 各 msg_type 样例、Safety Constraints、identity+scope 映射)均原样保留在新文档内联，删的全是重复/过度罗列(4× 选型规则、镜像 --help 的 Parameters 表、Common Mistakes、Notes、冗余 Commands)。无硬编码评测答案、未针对 080 卡片流窄化、未碰 SKILL.md 身份路由、单文件单根因。",
-    "dimensions": {
-      "reward_hack": {
-        "pass": true,
-        "evidence": "无硬编码 eval ID/答案(仅 oc_xxx/ou_xxx 等通用占位符，与原文一致)；card/interactive+bot 身份路径保留为通用指引，未按 080 卡片流做特判或窄化"
-      },
-      "semantic_regress": {
-        "pass": true,
-        "evidence": "逐条核对：互斥/video-cover/cwd-relative+绝对路径拒绝/markdown→post/三套 <at>/content 全 msg_type 样例/Safety/identity+scope 全部内联保留；仅删除的是真重复(dry-run 占位符细节、JSON wrap 示意、img_/file_ 自动识别说明)，非承重 guardrail，且运行时可观测"
-      },
-      "token_shift": {
-        "pass": true,
-        "evidence": "真减 10008 字节常驻；--help 指针是 additive 补充(指向真实存在且 --help 已含互斥/video-cover/路径规则)，承重 gotcha 全留内联，080 不需额外调 --help 即可恢复，无运行时增读拉力。注:work-order 提的 schema im.messages.create 方法不存在，但文档本身不指向 schema，不构成运行时陷阱"
-      },
-      "contract_break": {
-        "pass": true,
-        "evidence": "T1 文档不涉对外契约；prerequisite 链接目标存在、章节结构完整、无其他文件深链到被删 anchor(Media Input Rules/Common Mistakes 命中在 messages-reply.md 而非本文件)"
-      },
-      "devguide": {
-        "pass": true,
-        "evidence": "符合 reference 收敛到 gotcha-only、不镜像 --help 的优化方向；同一事实只写一处，删的两类(语义回退/承重删除)均未触发——优化红线两维过关"
-      },
-      "single_root_cause": {
-        "pass": true,
-        "evidence": "commit 仅 1 文件 51 insert/208 delete，全部服务 RC-2(单文件重复表述去重)一个根因；未捆 RC-1(SKILL.md)/RC-3(chat-create)，未把无关删除以 token 对冲缝入"
-      }
-    }
-  },
-  "child_k": 5,
-  "eval_trace": null,
-  "retro": {
-    "cause": "已入池",
-    "noise_borderline": false,
-    "summary": "越带入池，无需复盘补发"
-  },
-  "retro_sessions": [
-    {
-      "case": "3",
-      "session": null,
-      "axis": "token",
-      "expect": "降",
-      "parent": 37942,
-      "child": 35478,
-      "gain": "收益现",
-      "pass_delta": null
-    }
-  ],
-  "verdict": "admitted",
-  "ci": null,
-  "new_candidate": "557349b40feb359bb791749a37571d59edb7e72e",
-  "decision": {
-    "parent_success": 1.0,
-    "child_success": 1.0,
-    "parent_score": 0.6,
-    "child_score": 1.0,
-    "score_saved": 0.4,
-    "score_threshold": 0.09532271373123208,
-    "parent_token": 37942.0,
-    "child_token": 35478.0,
-    "saved": 2464.0,
-    "threshold": 4532.708313776408,
-    "parent_duration": 45769.0,
-    "child_duration": 46540.0,
-    "dur_saved": -771.0,
-    "dur_threshold": 4899.200953624988,
-    "dur_margin": 1.0,
-    "missing_duration": [],
-    "k_child": 5,
-    "k_parent": 5,
-    "decision_n": 1,
-    "missing_context": [],
-    "missing_score": [],
-    "parent_token_acc": 251669.0,
-    "child_token_acc": 221685.0,
-    "phi0_score": 0.5333333333333333,
-    "eff_margin": 1.0,
-    "parent_token_full": 37942.0,
-    "child_token_full": 35478.0,
-    "saved_full": 2464.0,
-    "observe_n": 1,
-    "target_axis": "token",
-    "admitted": true,
-    "reason": "score_gain"
-  },
-  "patch": "verify_results/round-002-lark-im-references-lark-im-messages-send.patch"
-}

harness-opt/rounds/round-002/strategy.md

@@ -1,48 +0,0 @@
-# Round 2 候选策略（模块=references/lark-im-messages-send.md, tier=T1, 主指标=token）
-
-## 根因与选择
-
-| 根因 | 来源(评测归因/规范经验) | 承载模块(reach) | annotation 风险级 | coverage 档 | P级 | 选中 |
-|---|---|---|---|---|---|---|
-| RC-2: messages-send.md 单文件最大、内部「选 content flag」规则重复 4 处 + 全媒体形态罗列 + Parameters/Notes 镜像 --help | 评测归因①（080 实读实用）＋规范经验②（annotation R1×140/R2×109，仅 1 段 R3） | references/lark-im-messages-send.md (0.333) | R1/R2 主导，唯一 R3=Safety Constraints(L9–22) | 密 / overfit 低 | P1 | ✅ |
-| RC-1: SKILL.md `## Important Notes` 低命中 + `## Shortcuts` 全表常驻 | 评测归因①（reach=1.0，3 题全命中） | SKILL.md (1.0) | R2/R3 混合（identity/约束密集） | 密 / 中 | P0(命中) 但 effect 高风险 |  |
-| RC-3: chat-create.md 按需偏大 | 评测归因① | references/lark-im-chat-create.md (0.667) | — | 密 | P1 |  |
-
-- **选中理由**：RC-2 是诊断点名「最干净的 token 杠杆」——单文件最大块（实测 ~5,365 tok，占 080 visible 24.8%），且 080 调用前已读、确实据它发卡片成功（reach=0.333、actual=1，非「读了没用」）。annotation 证实它 R1/R2 主导（140 R1 + 109 R2 行，可重构/可压缩），唯一 R3 段是 Safety Constraints(L9–22)，我**原样保留语义**。coverage=「密」、overfit「低」→ 本轮 eval 能在 080 上裁真伪。这是纯减体积、零能力删除、不碰 SKILL.md 路由的改动。
-- **为什么不选 RC-1**：reach=1.0、命中率最高，但 diagnosis 明确标它为 **effect 风险点**——剩余内容多为 identity/约束类，正是驱动 015/080 走通 bot 身份判断的承重内容；objective 的**硬门槛是「保住成功率」**，动 SKILL.md 最可能误伤这条已绿链路。本轮放弃，避免拿成功率换 token。
-- **为什么不选 RC-3**：diagnosis 判其杠杆最低（体积不离群），列为更次级；同一根因一轮只动一个，留待后续轮次。
-- **选模块理由**：messages-send.md reach=0.333>0（满足 reach 锁），承载选中的 RC-2，是非域 reference、改它不触碰 SKILL.md 的身份路由面。多文件无——本轮只动这一个文件。
-- **规范经验源补注**：对照 content-taxonomy——「单命令用法/长示例/与 --help 重复」类默认 R0/R1，「一般行为规则/CLI 机制约定」默认 R2；本文件的重复选型规则、全形态 Commands、Parameters/Notes 镜像即此类，处理方向为「留命中率最高一处，其余删或指针」「高频留 2–3 例，长的下沉」。当轮可被 080 裁真伪（coverage 密/overfit 低）。
-
-## 改了什么（逐处）
-- **L23–43 `## Choose The Right Content Flag` + `### --text vs --markdown`**：两段语义重叠的选型说明 → 合并为单张 4 行选型表（markdown/text/content/media），并把互斥规则并入表后一句。删掉 `### --text vs --markdown` 整段（与表重复）。
-- **L44–82 `## What --markdown Really Does` + `### Markdown Boundaries` + `### Image Constraint`**：三段约 39 行 → 压成 `## --markdown Gotchas` 三条要点（强制 post/无 title、标题改写规则、图片预上传 vs 远程 URL vs 本地路径不支持）。删掉 JSON wrap 示意、逐条 boundary 罗列等可由行为观察得到的展开。
-- **L83–93 图片预上传双命令示例**：并入 `## Commands` 的一条 markdown+image 示例（保留 `im images create` → 引用 img_xxx 的关键两步）。
-- **L114–161 `## Commands`（15+ 例覆盖全媒体形态）+ `## Media Input Rules`**：压成代表性示例（markdown / text / DM / post-title / markdown+image / 4 个媒体一组 / idempotency+dry-run），媒体路径规则收成 `--help` 指针后的 3 条 load-bearing gotcha（cwd-relative/绝对路径拒绝、video-cover 必配、msg-type 推断冲突）。
-- **L169–191 `## Parameters` 表**：删除镜像 `--help` 的逐参数描述，改为「Run `lark-cli im +messages-send --help`」指针 + 仅保留 --help 不显然的三条硬规则（已并入 Commands 末尾）。
-- **L192–202 `## Common Mistakes`**：整段删除——逐条都是选型表/markdown gotcha 的反向重述（第 4 次重复选型规则），删后选型信息仍在表里。
-- **L203–216 `## content Format Reference`**：保留（构造 `--content` 的 gotcha），把 image/file/audio 三行合并为一行省重复。
-- **L227–248 `## @Mention Format`**：保留全部三种 msg_type 的 `<at>` 语法（text/post/interactive 各异、AI 猜不到），压紧为两条要点、去掉小标题与重复散文。
-- **L249–264 `## Notes`**：整段删除——逐条（互斥/media 上传/scope/markdown 强制 post/video-cover/msg-type 冲突）均已在 Safety Constraints、选型表、--markdown Gotchas、Commands 指针处各保留一处单一事实源。
-
-## 为什么这么改（机制）
-- **消除根因的因果链**：该 reference 的体积来自「同一份选型规则在 4 个 section 重复 + 全媒体形态逐条罗列 + Parameters/Notes 镜像 --help」。token 不是被任务必需信息占用，而是被**重复表述**占用。按「同一份事实只写一次」（锚点 1）合并到单一事实源后，每条 load-bearing 信息仍恰好出现一次，080 这类「读该 reference→发消息」的题，读入 token 直接下降而行为不变。
-- **不删能力**：每个 flag（text/markdown/content/image/file/video/audio/idempotency/dry-run/msg-type/video-cover/as）、每条硬约束（互斥、video-cover 必配、cwd-relative 路径、绝对路径拒绝、markdown 强制 post/无 title、msg-type 冲突校验）、三套 `<at>` 语法、content 各 msg_type 样例、Safety Constraints、identity+scope 映射——全部保留，只是从「重复 N 次/逐条罗列」变成「一处/代表性示例 + --help 指针」。
-- **规范经验源**：依 optimization-playbook「reference 收敛到 gotcha-only，不做 --help 镜像」——Parameters 全表/全形态 Commands 属 USAGE，下沉到 `--help` 指针；保留的是 --help 表达不了的跨 flag 互斥、媒体路径安全、markdown→post 边界、@mention 按类型差异等 gotcha。annotation 标这些段为 R1（可重构/下沉），符合处理方向；唯一 R3（Safety）原样保留。
-
-## 预期效果
-- **成功率**：不退化。080（唯一读该文件的题）的发卡片链路依赖的是 `--content`/`interactive`、identity=bot、chat-id——全部保留；选型表、content Format Reference、Safety、scope 都在。015/080 走通 bot 身份的判断由 SKILL.md + identity 段承载，本轮**没碰 SKILL.md**，零误伤面。014 与本文件无关（reach 不含 014）。
-- **context（分两层）**：
-  - (1) **静态字数差**：16,407 → 6,399 chars（-61.0%）；tiktoken cl100k 3,869 → 1,799 tok（-53.5%）。（注：diagnosis 报 ~5,365 tok 系另一 tokenizer/含注入开销；此处用 cl100k 自测，方向与幅度一致。）
-  - (2) **运行时 context 方向**：仅在**实读该 reference 的子集**生效——本轮即 080 一题，运行时读入下降约 50%+（该块占 080 visible 24.8%，预计 080 visible 降约 12–13%）。其余两题（014/015）不读该文件，运行时 token **不变**（既不增也不减）。这是按需 reference，不是常驻面，不会影响未读它的题。
-- **覆盖敞口**：RC-2 子集仅 080 一题（reach=0.333），证据基数小。coverage 判该文件「密/overfit 低」，本轮 eval 可在 080 上裁真伪，但单题不可外推到「所有发消息任务」。建议后续补「读 messages-send.md 后用 --markdown / 媒体 / @mention」的 case 加厚子集。预期收益落在 **token 轴**（080 visible 下降），effect 轴维持不退化。
-
-## 刻意没做什么（反 reward-hack / 反过拟合）
-- 没硬编码任何评测题答案；没删任何能力、flag、guardrail、身份/scope 说明；没碰 lark-im 以外文件，也没把无关根因捆进本轮（commit 仅 1 个文件）。
-- **没碰 SKILL.md（RC-1）**：尽管 reach=1.0 杠杆最大，但其剩余内容是驱动 015/080 bot 身份判断的承重 identity/约束，diagnosis 标为 effect 风险点；在「保住成功率」硬门槛下不拿成功率换 token。
-- **没补收窄/分页指引**（015 的 22.5k chat-messages-list 黑洞）：那是「增内容」，与降 token 目标方向相反，diagnosis 已列为观察项、本轮不做。
-- 本改动**不是按评测错误反推**的参数/路由拟合——是基于 annotation + content-taxonomy 的结构性去重，删的是重复表述而非按 080 的具体内容裁剪；真实价值在「任何读该 reference 的发消息任务都少读重复 token」，080 只是当轮可验证的子集。
-- 未发现需要 breaking（T3）才能根治的点；本轮纯 T1 文档去重即可。
-
-## 签名
-- signature: 557349b40feb359bb791749a37571d59edb7e72e (commit 82a099fe 的 diff hash)  tier: T1

harness-opt/rounds/round-002/trend.json

@@ -1,11 +0,0 @@
-[
-  {
-    "round": 1,
-    "n": 3,
-    "pass_n": 0,
-    "cmd_fail_rate": 0.6,
-    "tool_calls": 26.333333333333332,
-    "duration_ms": 50189.0,
-    "token": 31997.0
-  }
-]

harness-opt/rounds/round-002/workorder.md

@@ -1,43 +0,0 @@
-# Round 2 归因派工单（parent=a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e；模块未定，由 candidate-writer 据诊断点名）
-
-> **只读输入**——opt-attributor 读本文件，把诊断**另写** `diagnosis.md`（给 candidate-writer）+ 逐题结构化 `attribution.json`（给 dashboard）。**不要覆盖本文件**，留作派工单↔诊断的前后对比。
-> 判分点只当「什么算挂」的锚，禁止照抄 grader 药方（已从派工单剔除）。
-
-## 模块运行时可达性（选模块第一步的证据；要选须在 strategy.md 说明理由）
-> reach=**实测**触达率（域主 SKILL.md 经 Skill 工具加载、reference 经 Read，都从 trace 实测，没有恒在的面）；判决集=实测∪预期触达。**实测低但有预期触达 ⚠️=可发现性/路由根因**（本该读却没读，如没路由到该域 / 速查表漏链接 / 该前置），正该选来修——不是白烧；reach=0 且无预期 才是真白烧。 **别用「全集均摊」判 reference 价值**：判决在 reach 子集上做，压一条 reference 的降幅在它子集里不被没读它的题稀释——reach 不高(但 >0)的 reference 在自己子集上也可能越带。
-- `skills/lark-im/SKILL.md` → reach=1.0  [域主 skill·经 Skill 工具加载]；判决集(实测∪预期): ['1', '2', '3']
-- `skills/lark-im/references/lark-im-chat-create.md` → reach=0.667；判决集(实测∪预期): ['1', '3']
-- `skills/lark-im/references/lark-im-messages-send.md` → reach=0.333；判决集(实测∪预期): ['3']
-- （另 22 个 reference reach=0 且无预期触达，本轮无关，略）
-
-## 逐轮诊断信号趋势（纯诊断，不进判决）
-
-| 轮 | 题数 | PASS | 命令失败率 | 工具调用 | 耗时(ms) | token |
-|---|---|---|---|---|---|---|
-| R1 | 3 | 0 | 0.60 | 26 | 50189 | 31997 |
-
-> 跨题均值，按轮排。**命令失败率、工具调用数是横切诊断信号，不是准入轴**（准入只走 效果/token/耗时）——用来判「上一轮那刀有没有把失败/轮次压下去」。工具调用数比 wall-clock 稳，可给噪声大的耗时轴当旁证。
-
-### 1  [PASS]  ctx=34270  (acc=274608)  43995ms  tools=31
-- session.jsonl: harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_014/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✗ 使用当前用户身份创建名为「IM合作群」的群聊
-        证据: transcript 在展示授权二维码后结束，无任何 `lark-cli im +chat-create` 调用。执行停在 '授权完成后请告诉我，我会继续帮你创建群聊并发送消息'，群聊未创建。
-    ✗ 将傅一铭和傅二铭加入该群
-        证据: transcript 显示尝试搜索用户时遇到 `need_user_authorization` 错误，授权流程启动后中断。未获取到任何用户的 open_id，无后续添加操作。
-    ✗ 在该群发送文本消息「大家体验有问题随时沟通」，并返回可验证的 chat_id / message_id
-        证据: 群聊未创建，无 chat_id 可返回。transcript 无任何 `lark-cli im messages-send` 调用。
-
-### 2  [PASS]  ctx=47116  (acc=612048)  114310ms  tools=49
-- session.jsonl: harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_015/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✓ 成功定位名为「fusanming_at_openclaw群」的群，并获取最近包含「飞豆」关键字的消息
-    ✓ 将筛选出的相关消息内容转发到「fusanming_at_需求测试群」
-    ✓ 在「fusanming_at_需求测试群」中 @傅六铭 做知会，消息发送成功
-
-### 3  [PASS]  ctx=37942  (acc=251669)  45769ms  tools=23
-- session.jsonl: harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_080/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✓ 使用用户身份创建一个名为「今晚吃什么」的群，预期返回 chat_id
-    ✓ 创建一张飞书卡片，卡片内容包含「今天晚上吃什么」
-    ✓ 将该卡片发送到新建群中，预期返回 message_id

harness-opt/rounds/round-003/attribution.json

@@ -1,59 +0,0 @@
-[
-  {
-    "case_id": "1",
-    "verdict": "PASS",
-    "verdict_note": "workorder=PASS（聚合口径）；判分点证据 3/3 ✗，按判分点当实质 FAIL 处理",
-    "token": 34555,
-    "duration_ms": 37000,
-    "tool_calls": 31,
-    "cmd_attempts": 7,
-    "cmd_failures": 5,
-    "cmd_fail_rate": 0.71,
-    "discoverability_state": "无（失败命令全是跨域 contact + auth，非 lark-im；chat-create.md 调用前已读但未走到使用）",
-    "axis": "效果",
-    "root_cause": "沙箱 user 授权不可完成 + 跨域 lark-contact 命令依赖；无 lark-im 文档根因，本轮不改",
-    "token_hotspot": "运行时冗余清单常驻（SKILL.md 3,456）+ 按需 chat-create.md 3,062（读了没用上）；lark-shared 3,751 与系统 Skill 列表注入 4,612 均不归因",
-    "token_reliability": "常驻静态（SKILL.md）/ 按需读取（chat-create.md，本题读了没用上）",
-    "duration_hotspot": "多轮交互（查联系人→切 contact→失败→授权→qrcode 重试）+ 纯外部API延迟(部分不可归因)",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "效果侧无 lark-im 文档缺信息（环境+跨域）；token 侧 chat-create.md 把同组 flag 在 Commands/Usage Scenarios 重复演示、Common Errors 复述 validation 字符串，属可删冗余"
-  },
-  {
-    "case_id": "2",
-    "verdict": "PASS",
-    "verdict_note": "真 PASS，判分点 3/3 ✓，全程 bot 身份无授权阻断",
-    "token": 54568,
-    "duration_ms": 125000,
-    "tool_calls": 49,
-    "cmd_attempts": 11,
-    "cmd_failures": 3,
-    "cmd_fail_rate": 0.27,
-    "discoverability_state": "① 从没读（messages-search.md / chat-messages-list.md reach=0，直接猜命令；本题未读任何 lark-im reference）",
-    "axis": "token",
-    "root_cause": "无过滤 +chat-messages-list --page-all 全量拉取 → 43.5KB 输出被 Read 整文件灌入 22,556 tok；token 大头非 lark-im doc。修它需补收窄/前置内容，与降 token 目标方向冲突，列观察项",
-    "token_hotspot": "工具返回原样输出（block #19 单次 Read 22,556 tok / 51.5%，归「其他工具调用/返回」）",
-    "token_reliability": "单次输出（强依赖该群消息量，单题不可外推，非稳定常驻热点）",
-    "duration_hotspot": "多轮交互 + 重试（messages-search 连环 exit2 → page-all → 大输出 → 多次本地 grep 抠数据）",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现（model calls 16 作旁证，明显高于 080）",
-    "doc_fix_hint": "本题无 T1 可发力的 token 抓手（大头是单次工具输出，非 lark-im doc 常驻）；缺的是大群消息查询的 server-side 收窄指引，但补它=增内容、与降 token 反向，不作本轮根因"
-  },
-  {
-    "case_id": "3",
-    "verdict": "PASS",
-    "verdict_note": "真 PASS，判分点 3/3 ✓，主动选 bot 身份建群+发卡片均 ok:true，零命令失败",
-    "token": 38009,
-    "duration_ms": 47000,
-    "tool_calls": 22,
-    "cmd_attempts": 3,
-    "cmd_failures": 0,
-    "cmd_fail_rate": 0.0,
-    "discoverability_state": "无（无失败命令；SKILL.md + chat-create.md + messages-send.md 全部状态③：调用前已读且用上）",
-    "axis": "token",
-    "root_cause": "读取 Skill 占 56.4%；本轮唯一干净 token 抓手 = chat-create.md 内部冗余（示例罗列 + 场景重复 + --help 镜像），从未被优化过",
-    "token_hotspot": "运行时冗余清单常驻 + 按需 reference（chat-create.md 当前 2,336 raw，可压 Commands/Usage Scenarios 重叠 + Common Errors validation 镜像；trace 里 messages-send.md 5,365 是旧版，round-2 已压到 2,006，本轮不再可压）",
-    "token_reliability": "按需读取（chat-create.md reach=0.667，本题是其压缩收益唯一稳态兑现题）",
-    "duration_hotspot": "无离群（建群+发卡片正常串行，无重试、无写后回查）",
-    "duration_reliability": "耗时波动大，单次运行不算数，需多题或多次复现",
-    "doc_fix_hint": "chat-create.md 把同组 flag 在 Commands(12 例)+Usage Scenarios(3 场景)重复演示、Common Errors 多行复述 --help/报错本身就会吐的 validation 字符串，属可删冗余；232043 两步流 / --chat-mode topic 区分 / --owner 默认为载重红线，压缩中不可误删"
-  }
-]

harness-opt/rounds/round-003/case-commands.json

@@ -1,27 +0,0 @@
-{
-  "1": [
-    "auth login",
-    "auth qrcode",
-    "contact +search-user"
-  ],
-  "3": [
-    "auth login",
-    "auth qrcode",
-    "auth status",
-    "im +chat-create",
-    "im +messages-send"
-  ],
-  "2": [
-    "auth login",
-    "auth qrcode",
-    "auth status",
-    "im +chat-messages-list",
-    "im +chat-search",
-    "im +messages-mget",
-    "im +messages-search",
-    "im +messages-send",
-    "im messages forward",
-    "schema im.messages.forward",
-    "schema im.messages.search"
-  ]
-}

harness-opt/rounds/round-003/child-cache.json

@@ -1,20 +0,0 @@
-{
-  "1": {
-    "score": 1.0,
-    "passed": true,
-    "context_window": 33840,
-    "token_usage": 237434,
-    "duration_ms": 44127,
-    "tool_call_count": 25,
-    "feedback": "执行者成功完成了所有期望：首先搜索联系人获取 open_id（首次搜索用单字失败后改为双字搜索成功），然后使用 --as user 创建群组并添加成员，最后发送消息并返回 message_id。整个流程正确，使用了等效的 `--as user` 身份，符合用户「使用我的身份」的要求。验证结果确认所有操作均已生效。"
-  },
-  "3": {
-    "score": 1.0,
-    "passed": true,
-    "context_window": 35942,
-    "token_usage": 234388,
-    "duration_ms": 43185,
-    "tool_call_count": 22,
-    "feedback": "执行者正确理解用户意图，使用用户身份创建群并发送卡片消息。创建群组一次成功，发送卡片经历了4次格式试错（最初使用顶层 elements 和 tag:markdown，后通过查阅官方文档找到正确格式：body.elements + div + lark_md），最终成功发送并返回 message_id。试错后自行纠正符合评判原则，不构成判罚依据。\n- {'reason': '建议在 lark-im-messages-send.md 中增加飞书 interactive card 的标准格式示例，特别是 2.0 schema 下的 body.elements 中使用 div + lark_md 的正确写法，减少 AI 试错成本'}\n- {'reason': '建议 CLI 在遇到 230099 卡片格式错误时，尝试解析并返回更具体的字段级错误提示（如提示 \"elements 应在 body 内\" 或 \"tag:markdown 不被支持\"），帮助 AI 更快定位问题'}"
-  }
-}

harness-opt/rounds/round-003/diagnosis.md

@@ -1,119 +0,0 @@
-# Round 3 归因（parent=557349b…（round-2 已采纳候选）；候选模块见 candidate_modules，由 candidate-writer 据诊断+reach 点名）
-
-> 目标（objective.json）：**在不回退成功率的前提下降低 lark-im skill 文档的 token 成本**。effect 是硬门槛、不可退化；token 与 duration 是并列成本杆。tier=T1，仅可改 `skills/lark-im/**`。target_axis=token。
-> 判分点只当「什么算挂」的锚，不抄 grader 药方。
-
-## ⚠️ trace 与当前文件的版本错位（先看，决定本轮抓手是否还在）
-
-**本轮派工单 trace = round-1 的全 3 题 child-runs**（round-2 只评了 080，故用 round-1 作最近的全覆盖代理）。这些 trace 里的 reference 体积是 **round-1/round-2 改动之前** 的旧版。我用 session-analyze 所用的同一 ai-tokenizer 实测了**当前工作树**文件，确认两者错位如下：
-
-| 文件 | trace 内体积（旧版，Read 计） | 当前实测（raw / Read 计） | 已被哪轮收割 |
-|---|---|---|---|
-| `SKILL.md`（Skill 注入正文） | 3,455–3,456 tok | 3,525 raw | round-1（API Resources/权限表→schema 指针） |
-| `references/lark-im-messages-send.md` | **5,365 tok** | **2,006 raw / 2,194 Read** | **round-2（5,365→2,006，已收割）** |
-| `references/lark-im-chat-create.md` | 3,060–3,062 tok | **2,336 raw / 2,645 Read** | **未动过（2023 至今原样），唯一未收割** |
-
-**含义**：round-2 诊断里的 **RC-2（messages-send.md 内部冗余）已经在 round-2 被采纳并收割**（5,365→2,006），它不再是本轮抓手——不要据 trace 里的 5,365 重复提一遍。本轮 trace 里那块 5,365 是历史值，当前已不存在。**reach>0 集合里唯一还没被压过的干净文件就是 `chat-create.md`**（round-2 的 RC-3）。
-
-## 跨 case 共同根因（优先看；按对 TOKEN 目标的杠杆排序）
-
-### RC-1（token，本轮头号且基本是唯一的干净抓手，reach=0.667：014+080）—— `chat-create.md` 内部存在「示例罗列 + 场景重复 + --help 镜像」三类可压缩冗余，且从未被优化过
-- **现象**：`chat-create.md` 当前 2,336 raw tok（Read 计 ~2,645），是 reach>0 集合里**唯一未被任何轮收割**的 reference。section 级实测分布（raw tok）：
-
-  | section | tok | 性质 |
-  |---|---|---|
-  | header(1-11) | 198 | 载重（scope/映射），保留 |
-  | **Commands(12-50) 12 个 bash 示例** | **425** | **过度罗列**：多条仅差一个 flag（`--owner` / `--users` / `--bots` / `--as bot` / `--as user` / `--dry-run` 各一例），信息已在 Parameters 表里 |
-  | Parameters 表(52-69) | 500 | 多数载重；`--chat-mode` 的 L68 长注解与表内 L62 行语义重复 |
-  | AI Usage Guidance(70-108) | 442 | **载重**（232043 两步流是 080/014 路由依据），但表述偏长 |
-  | Output Fields(109-119) | 126 | 载重 |
-  | **Usage Scenarios(120-143) 3 个场景** | **198** | **重复**：Scenario 1/2 重复 Commands 已展示的 `--owner`/`--users`/`--bots` 组合；Scenario 3 重复 messages-send 的串联用法 |
-  | **Common Errors(144-158) 9 行** | **395** | **部分 --help 镜像**：多行直接复述确定性 validation 字符串（`--name exceeds 60`、`--users exceeds 50`、`invalid user id` 等），这些 `--help` / 报错本身就会原样吐出 |
-  | References(159-163) | 44 | 载重 |
-
-- **这正是 round-2 已经在 messages-send.md 上验证过、且被采纳的同一套压缩模式**：round-2 把 messages-send.md 的「4 处重复选型规则 + 全媒体形态 Commands + --help 镜像」压成「保留载重规则 + 一句 `--help` 指针」（5,365→2,006，被采纳）。chat-create.md 的 Commands(425)↔Usage Scenarios(198) 重叠、Common Errors(395) 的 validation 镜像，是同型冗余。
-- **可压缩量级（粗估，非药方）**：可压缩质量集中在 Commands+Usage Scenarios 的重叠（合计 ~623 tok，去重后可省一大半）+ Common Errors 的 --help 镜像行。**保守估计可从 2,336 压到 ~1,500–1,700 raw tok（省约 600–800 tok，约 30%）**，与 messages-send.md 的压缩比同量级。具体改法与确切降幅由 candidate-writer 决定、评测裁决。
-- **载重红线（candidate-writer 取舍时的 effect 风险点，不是 RC-1 不成立）**：AI Usage Guidance 的 **232043 两步流 + `succeed_type=1`**、`--chat-mode topic` vs 普通群+话题消息模式的区分、`--owner` 默认行为，是 014/080 走通 bot 身份建群的语义依据，**不能在压缩中误删**。这条 reference 被 080 实读且 080 据它建群成功（`ok:true`），所以 effect 风险真实存在——压的是示例/场景/报错镜像的体积，不是语义规则。
-- **axis=token**。可信度=**按需读取**（reach=0.667，子集=014+080，2 题）。压它的降幅只在这 2 题子集里计入，不被 015（没读它）稀释；但子集仅 2 题、且 014 是「读了没用上」（授权阻断没走到建群），实际吃到压缩收益的稳态题只有 080 一题——**证据基数小，降幅需评测在子集上确认**（见数据缺口）。
-
-### RC-2（token，已收割，本轮不再是抓手）—— messages-send.md 的内部冗余 round-2 已压掉
-- round-2 RC-2 已被采纳：messages-send.md 5,365→2,006 raw。**本轮不要据 trace 里的 5,365 重复提**。当前 messages-send.md 已是「载重规则 + `--help` 指针」形态，无明显二次压缩空间（剩余多为 content-flag 选型、@mention、media 约束等载重内容）。reach=0.333（仅 080）。
-
-### RC-3（token，无 T1 干净抓手）—— SKILL.md 常驻正文 round-1 已压过，剩余多为载重 identity/路由
-- SKILL.md 经 Skill 工具每题必加载（reach=1.0），当前 3,525 raw tok（round-1 已把 API Resources/权限表折叠成 schema 指针）。剩余 `## Important Notes`(L36–85) 各小节（Sender Name Resolution / message enrichment / `--download-resources` / Card / Flag / Feed Shortcut）与 `## Shortcuts` 全表(L87–115) 虽本轮 3 题低命中，但它们是**全域 identity/路由/约束**——这是 round-1 已经动过一刀的同一文件，**再压属同向继续、但删错会碰坏 015/080 已走通的 bot 身份与命令路由判断**（effect 风险高于 RC-1）。**列为更次级、风险更高的抓手**，不作为本轮首选；若要动须只删本轮已确证低命中且非路由的纯枚举行，谨慎程度高于 chat-create。
-
-## 命令失败热点（跨 case；失败类型由我从 timeline 命令串读出，非判决数字）
-
-| lark-cli 命令 | 失败次数 | 涉及题数 | 主要失败类型 | 指向的文档问题 |
-|---|---|---|---|---|
-| `contact +search-user` | 4 | 1 (014) | bot exit2(invalid_argument) ×2；`--as user` token_missing ×2 | **跨 lark-contact 域**，非 lark-im 内容 |
-| `auth qrcode --output <绝对/沙箱外路径>` | 1 | 1 (014) | unsafe output path，改相对路径重试成功 | 路径约束在 lark-shared（不可改） |
-| `im +messages-search` | 2 | 1 (015) | bot exit2 + `--as user` exit2 | 该命令 user-only（SKILL 表已注明）；bot 身份必败，agent 没看清就猜 |
-| `im +chat-messages-list --page-all` | 1 | 1 (015) | exit2（无过滤 page-all） | 见下「015 token 黑洞」 |
-- **解读**：本轮**没有一条 lark-im 命令因「参数名/类型写错」系统性失败**。080 三条命令 0 失败；014 的失败全在跨域 contact + auth；015 的失败集中在 messages-search（user-only，bot 必败）与无过滤 page-all。**没有 lark-im 侧常规「报错/参数整形」工单**——与 token 减体积方向一致，本轮抓手是减体积不是补内容。
-
-## 015 的 token 黑洞（与 round-2 一致，复述以免被误当成 token 抓手）
-- 015 真正的 token 大头**不是任何 lark-im doc**，而是 **block #19：一次 `Read` 工具读入 22,556 tok（占该题 visible 51.5%）**。成因链：#12/#17 `+messages-search`/`--page-all` exit2 → #18 退到 `+chat-messages-list`（无过滤）→ 输出 43.5KB 被持久化 → agent `Read` 整文件 → 22.5k tok 灌进上下文 → 再靠本地 grep(#27–33) 抠出「飞豆」两条。
-- **从文档角度**：`chat-messages-list.md` 本题 reach=0（状态①：调用前从没读），它本写了 `--start/--end`、`--page-size` 等可避免全量拉取的约束。**但补它=增常驻/触达内容，与本轮降 token 目标方向相反**（见方向冲突）；且 22.5k 是**单次工具输出**（强依赖该群消息量，单题不可外推），不是稳定常驻热点。**结论：观察项，交评测裁决，不作为本轮 token 抓手。**
-
-## 可发现性时序（约束 5 三态；判「前置能不能救」的决定性证据）
-> 对每条相关 reference / `--help`，按相对首次失败调用的读取时序统计。`--help` 扫 Bash（本轮 3 题均未跑任何 `--help`）。
-
-| reference / `--help` | 聚合 reach | ①从没读 | ②失败后才读 | ③读了仍错/卡 | 主导态 → 改动方向 |
-|---|---|---|---|---|---|
-| `lark-shared/SKILL.md` | 1.0 | 0 | 0 | — | 三题调用前都读了；014 仍卡（环境，非内容）；不可改 |
-| `chat-create.md` | 0.667 | 0 | 0 | — | 080 调用前读→建群成功；014 调用前读→授权阻断（非 reference 错）。**非触达问题，纯减体积** |
-| `messages-send.md` | 0.333 | 0 | 0 | — | 080 调用前读→发卡片成功。**非触达问题**（已收割） |
-| `chat-messages-list.md` | 0.0 | 1 (015) | 0 | — | ① 015 调用前从没读→`--page-all` 全量拉取→token 黑洞。触达缺口，但补它=增 token，与目标冲突 |
-| `messages-search.md` | 0.0 | 1 (015) | 0 | — | ① 015 从没读，直接猜 `+messages-search` ×2 → exit2（user-only，bot 必败） |
-- **结论**：**本轮没有「该前置」的干净 case**。RC-1（chat-create.md 减体积）是「调用前已读、内容够用 → 去冗余」的纯 token 减法，不涉及触达。015 的两处 ① 触达缺口确实存在，但修它们=增内容、与降 token 目标相反，且 015 已 PASS（bot + 本地 grep 兜底）——属观察项，**不要被诱导去推前置**。
-
-## 方向冲突记录（硬性约束 7）
-- **减体积（RC-1 chat-create.md，与 objective.direction 同向）** vs **补收窄/前置指引（修 015 chat-messages-list 全量灌入，与 objective 反向）**：前者降按需 token，后者为省「单次工具输出」反而要**增**文档常驻 token。两者方向相反，**不可合并**。本轮目标是降 token，取减体积一侧；015 全量灌入作为观察项记录、不作为要补的内容根因。
-
-## 差距台账复盘
-- 无（`discard-ledger.json` 为空，无已跑未采纳候选）。
-
-## 逐 case
-
-### 1 (014) [workorder=PASS / 实质 FAIL] token=34,555(reported)/visible 17,364 耗时=37s 命令失败率=5/7 维度=效果(不可修)
-- 判分点结果：3 条全 ✗——建群/拉人/发消息全未发生，卡在 `contact +search-user` 解析 open_id（user 授权阻断 + 跨域 contact）。verdict=PASS 系聚合口径，按判分点证据当 FAIL 处理。
-- 命令失败：5/7。`contact +search-user` bot exit2 ×2、`--as user` token_missing ×2；`auth qrcode` 绝对路径 unsafe ×1（改相对路径成功）。**全部非 lark-im 命令**。
-- 可发现性时序：#4 读 SKILL.md 正文(3,456) + #6 读 lark-shared(3,751，跨 skill) + #7 读 chat-create.md(3,062，调用前已读)；失败在更上游的跨域 contact + 授权。**非 lark-im 触达问题**。
-- token 归因：SKILL.md 正文 3,456（常驻静态，19.9%）+ lark-shared 3,751（**跨 skill，不归因 lark-im**）+ chat-create.md 3,062（按需，17.6%，**本题读了没用上**——授权阻断没走到建群）+ 系统 Skill 列表注入 4,612（固定开销，不归因）。lark-cli 命令累计含多次短失败回显，单条都短、非热点。
-- 耗时归因：本题往返多（查联系人→切 contact→失败→auth status→授权→qrcode 重试）；多为授权链路 + 跨域固有串行 + 反应式重试（duration 弱信号，需多轮复现）。
-- 文档根因：效果=沙箱 user 授权 + 跨域 contact（环境，**无 lark-im 文档根因，本轮不改**）；token=chat-create.md 按需冗余（RC-1，但本题读了没用上，收益只在 080 这种走通题里兑现）+ SKILL.md 常驻（RC-3，风险高、次级）。
-
-### 2 (015) [PASS·真] token=54,568(reported)/visible 43,760 耗时=2m5s 命令失败率=3/11 维度=token（但大头非 lark-im doc）
-- 判分点结果：3/3 ✓——定位群、转发「飞豆」消息、@傅六铭知会全部成功（两次 `messages-send` 均 `ok:true`）。**全程 bot 身份，无授权阻断**。
-- 命令失败：3/11。`+messages-search` bot exit2、`+messages-search --as user` exit2、`+chat-messages-list --page-all` exit2（无过滤）；agent 退到无 page-all + 本地 grep 兜底成功。（#14 `--page-all | grep` 返回空属「成功但无命中」，非硬失败，未计入。）
-- 可发现性时序：① `messages-search.md` / `chat-messages-list.md` 调用前从没读（reach=0），直接猜命令。**本题未读任何 lark-im reference**，故 lark-im reference 的体积与本题 token 无关。
-- token 归因：**本题 token 大头不是 lark-im doc**，是 block #19 一次 `Read` 持久化文件 = **22,556 tok（51.5%，归「其他工具调用/返回」）**，成因=`--page-all` 无过滤全量拉取→43.5KB→Read 灌入（**单次输出**可信度，强依赖该群消息量）。SKILL.md 正文 3,448（常驻）。lark-shared 3,749（跨 skill，不归因）。**RC-1 改 chat-create.md 对本题 token 无影响**（本题没读它）。
-- 耗时归因：本题最长(2m5s)，主因 messages-search 连环失败→改 page-all→大输出→多次本地 grep 抠数据的多轮往返（duration 弱信号；model calls 16/raw 32，明显高于 080，作旁证）。
-- 文档根因：token 黑洞的放大器=`chat-messages-list.md` 没被读到（状态①）+ SKILL.md 表未提示大群应 server-side 收窄——但**补这条与降 token 目标相反**（方向张力），列为观察项；本题已 PASS。本轮 token 抓手（RC-1）不落在本题。
-
-### 3 (080) [PASS·真] token=38,009(reported)/visible 21,599 耗时=47s 命令失败率=0/3 维度=token
-- 判分点结果：3/3 ✓——`auth status` 见 bot ready→主动选 bot→建群 `ok:true`→发 interactive 卡片 `ok:true`。**任务完整完成，零命令失败**。
-- 命令失败：0/3。三条 lark-cli（auth status / chat-create / messages-send）全成功。
-- 可发现性时序：#4 读 SKILL.md 正文(3,455) + #6 读 lark-shared(3,751，跨 skill) + #9 读 chat-create.md(3,060) + #10 读 messages-send.md(5,365，旧版) ，全部状态③（调用前已读且用上）。**无触达问题。** 实际只用了 `+chat-create --name … --format json` 的最简形态——没用两步流/owner/members/topic/error-recovery。
-- token 归因：**本题是纯 token 抓手题**——读取 Skill 占 56.4%：messages-send.md 5,365（trace 旧版，**当前已被 round-2 压到 2,006，本轮不再可压**）+ SKILL.md 正文 3,455（常驻，RC-3）+ chat-create.md 3,060（按需，**RC-1，当前 2,336，本轮唯一干净抓手**）。系统 Skill 列表注入 4,612（固定开销，不归因）。lark-shared 3,751（跨 skill，不归因）。
-- 耗时归因：47s，全部为正常建群+发卡片串行，无重试、无写后回查（无离群）。
-- 文档根因：无效果根因（已绿）；token=RC-1（chat-create.md 内部冗余，本题是其收益唯一稳态兑现题）+ RC-3（SKILL.md 常驻，风险高、次级）。**本题 token 杠杆最清晰且 effect 风险可控**（命令全成功，压 chat-create.md 的示例/场景/报错镜像不碰 080 实际用到的最简建群链路）。
-
-## 给 candidate-writer 的收口（不含具体改法）
-- **唯一在 T1 内还没被收割的干净 token 抓手是 RC-1（`chat-create.md` 内部冗余）**：Commands 12 例过度罗列 + Usage Scenarios 3 场景重复 Commands + Common Errors 9 行部分镜像 validation 字符串——**与 round-2 已采纳的 messages-send.md 压缩同型**，粗估可省 ~600–800 raw tok（约 30%）。reach=0.667（014+080），降幅在子集计入。
-- **载重红线**：AI Usage Guidance 的 232043 两步流 + `succeed_type=1` + `--chat-mode topic` 区分 + `--owner` 默认，是 080/014 走通 bot 建群的语义依据，**压缩中不可误删**——压的是示例/场景/报错镜像体积，不是规则。
-- **RC-2 已收割**：messages-send.md round-2 已 5,365→2,006，trace 里的 5,365 是历史值，**不要重复提**。
-- **RC-3（SKILL.md 常驻）是次级且风险更高**：round-1 已压过一刀，剩余多为全域 identity/路由/约束，删错碰坏 015/080 已走通的 bot 身份与命令路由——不作首选。
-- **不要推前置**：本轮没有「该前置」的干净 case。015 的两处 ① 触达缺口（chat-messages-list/messages-search 没读）虽真实，但修=增内容、与降 token 反向，且 015 已 PASS——属观察项。
-- **effect 不可在本轮 T1 内合法抬升**：014 是环境（沙箱不能扫码）+ 跨域 contact，无 lark-im 文档根因；015/080 已真 PASS。effect deltas 视作 auth-noise，不追。
-- **干净 token 抓手接近见底（诚实判断）**：reach>0 集合三个文件中，messages-send.md（round-2）与 SKILL.md（round-1）已各压一刀，**chat-create.md 是最后一个未动过的干净文件**。压完它之后，T1 内 reach>0 的纯冗余（罗列/重复/--help 镜像）基本耗尽；再往下只剩 (a) 高 effect 风险的 SKILL.md 载重内容，或 (b) reach=0 的 22 个盲区 reference（压了也不在判决集、无法被采纳）。**本轮 RC-1 很可能是这条优化路径上最后一个低风险、可被采纳的 token 抓手。**
-- **缺失信息（doc_fix_hint 语气，非药方）**：chat-create.md 把同一组 flag 在 Commands(12 例) 与 Usage Scenarios(3 场景) 重复演示、Common Errors 多行复述 `--help`/报错本身就会吐的 validation 字符串——这类「枚举/重复/镜像、低增量」内容是其 token 的主要去处，且是减法（删冗余）而非加法。
-
-## 数据缺口
-1. **trace 版本错位（最关键）**：本轮 trace=round-1 旧版 child-runs，messages-send.md 在 trace 里仍是 5,365（round-2 已压到 2,006）。所有「当前文件体积」结论我已用 ai-tokenizer 实测当前工作树校正（SKILL.md 3,525 / chat-create.md 2,336 / messages-send.md 2,006），但**单题行为与 reach 仍来自旧 trace**——若 round-2 改动改变了 080/014 的读取行为，需以实际 round-3 eval-run 复核。
-2. **RC-1 子集小**：chat-create.md reach=0.667 但实际吃到压缩收益的稳态题只有 080（014 读了没用上、授权阻断），证据基数=1，降幅需评测在子集确认。
-3. **015 的 22.5k 黑洞是单次工具输出**，强依赖该群消息量，非稳定常驻热点，单题不可外推；且与降 token 目标方向冲突，不作抓手。
-4. **duration 三题波动大**（37s/2m5s/47s），015 长尾主因 messages-search 连环失败+大输出多轮抠数据；单轮不足定论，需多轮复现。model calls(8/16/6) 比 wall-clock 稳，可作旁证。
-5. **工具调用口径不一致**：trend.json 的 R1 tool_calls=26.3、R2=10，与 session-analyze 的 model calls(8/16/6) 口径不同（趋势表疑似含 raw 计数）；旁证以 timeline 实际往返为准。趋势看：R1→R2 命令失败率 0.60→0.35、tool_calls 26→10 明显下降，但那主要是 effect 从「三题全卡授权」变成「2 真 PASS + 1 卡」带来的，**不是 token 改动的功劳**；token 均值 R1 31,997→R2 42,377 上升，主因是 R2 只评 080（单题大）口径差异 + 015 黑洞，非文档常驻变重——趋势对 token 轴判读价值有限，以单题 session-analyze 为准。
-6. **effect 维度全部归因为「无文档根因/不可修」**：014 跨域+环境，015/080 已绿。本轮 effect 无 T1 可发力点，deltas 视作 auth-noise。

harness-opt/rounds/round-003/discard-ledger.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-003/failure-memory.json

@@ -1 +0,0 @@
-[]

harness-opt/rounds/round-003/module-reach.json

@@ -1,220 +0,0 @@
-{
-  "skills/lark-im/SKILL.md": {
-    "reach": 1.0,
-    "read_cases": [
-      "1",
-      "2",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "2",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": true
-  },
-  "skills/lark-im/references/lark-im-chat-create.md": {
-    "reach": 0.667,
-    "read_cases": [
-      "1",
-      "3"
-    ],
-    "actual_cases": [
-      "1",
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-identity.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-messages-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-search.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-chat-update.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-list-item.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-group-query-item.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-groups.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-create.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-feed-shortcut-remove.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-cancel.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-create.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-flag-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-message-enrichment.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-mget.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-reply.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-resources-download.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-search.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-messages-send.md": {
-    "reach": 0.333,
-    "read_cases": [
-      "3"
-    ],
-    "actual_cases": [
-      "3"
-    ],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-reactions.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  },
-  "skills/lark-im/references/lark-im-threads-messages-list.md": {
-    "reach": 0.0,
-    "read_cases": [],
-    "actual_cases": [],
-    "expected_cases": [],
-    "discoverability_miss": [],
-    "is_domain_skill": false
-  }
-}

harness-opt/rounds/round-003/review.json

@@ -1,18 +0,0 @@
-{
-  "generated_by": "lark-cli-harness:opt-reviewer",
-  "verdict": "PASS",
-  "module": "skills/lark-im/references/lark-im-chat-create.md",
-  "tier": "T1",
-  "round_index": 3,
-  "base_commit": "572eb8da41f608bd93b25916cac02cb772825b97",
-  "code_tip": "cbd6e56ac07285fd973c53ff7382da0112b6cf5d",
-  "reason": "纯瘦身改动，对抗式逐项核验未发现可证伪点：14 条承重红线（232043 两步流/succeed_type=1/chat-mode topic 与 thread 区分/--owner 默认/set-bot-manager/chat.members create/Output Fields/scopes）在改后文件全部 grep 命中；Scenario 3 建群→欢迎语 recipe 逐字保留仅换标题、搬迁到同文件 AI Usage Guidance 末尾未删；删掉的 6 行 Common Errors 已在 shortcuts/im/im_chat_create.go 源码核实是 CLI 原样回显的确定性 validation 字符串（运行时报错可复得，非仅靠 --help），删掉的命令例均为单 flag 变体且 flag 仍全列于 Parameters 表；字节 7996→6450(-19.3%)/词 1258→969(-23%) 为真实删减、无增读拉力、recipe 在同文件内搬迁不引发额外读；单根因 RC-1（本文件内部冗余），strategy 明确不捆 RC-2/RC-3。",
-  "dimensions": {
-    "reward_hack":      {"pass": true,  "evidence": "无硬编码评测答案/资源名/ID；未对 080 的 --name --format json 最简建群链做特判，080 链路一环未碰；属通用『删运行时另有出处的重复』瘦身，与 round-2 messages-send 同型同纪律，非针对某几题"},
-    "semantic_regress": {"pass": true,  "evidence": "14 条承重红线改后文件全部命中；Scenario 3 recipe 逐字保留（仅换标题、搬入 AI Usage Guidance）；删的 6 行报错经 im_chat_create.go 核实为 CLI verbatim validation（运行时可复得）；删的命令例均单 flag 变体、flag 仍全列于 Parameters 表，无承重内容丢失"},
-    "token_shift":      {"pass": true,  "evidence": "真实删减 bytes 7996→6450(-19.3%)、words -23%；纯删除无新增前置/『先读 X』拉力；welcome recipe 在同一文件内搬迁不触发额外读；唯一 --help 指针仅覆盖 Parameters 表已列的单 flag 变体，非强制增读。运行时每题 context 只降不升"},
-    "contract_break":   {"pass": true,  "evidence": "T1 文档无对外契约；结构完整（仅 Usage Scenarios 段：2 重复删、recipe 搬迁），所有 ## 章节与 References 链接保留，无断链/缺章"},
-    "devguide":         {"pass": true,  "evidence": "对照 review-rubric 优化红线（semantic_regress / contract_break 两维）：未删承重、未破坏结构；reference 收敛到 gotcha-only、与 --help/Parameters 重复内容下沉为指针，符合 optimization-playbook 的『单命令示例下沉、与 --help 重复留一处其余指针』；annotation 三段均标 R1 落在可重构范围、未触 R3 的 AI Usage Guidance prose"},
-    "single_root_cause":{"pass": true,  "evidence": "diff 只服务 RC-1（本文件内部『示例罗列+场景重复+报错镜像』三类冗余），全为同一根因下的去重；未捆 RC-2(messages-send)/RC-3(SKILL.md)，未夹带语义独立的承重删除，无多根因对冲叙事"}
-  }
-}

harness-opt/rounds/round-003/round.json

@@ -1,394 +0,0 @@
-{
-  "round": 3,
-  "status": "admitted",
-  "parent_id": "557349b40feb359bb791749a37571d59edb7e72e",
-  "parent_worktree": "/Users/bytedance/Projects/cli",
-  "child_worktree": "/Users/bytedance/Projects/cli",
-  "base_commit": "572eb8da41f608bd93b25916cac02cb772825b97",
-  "module": "skills/lark-im/references/lark-im-chat-create.md",
-  "candidate_modules": [
-    "skills/lark-im/SKILL.md",
-    "skills/lark-im/references/lark-im-chat-create.md",
-    "skills/lark-im/references/lark-im-chat-identity.md",
-    "skills/lark-im/references/lark-im-chat-list.md",
-    "skills/lark-im/references/lark-im-chat-messages-list.md",
-    "skills/lark-im/references/lark-im-chat-search.md",
-    "skills/lark-im/references/lark-im-chat-update.md",
-    "skills/lark-im/references/lark-im-feed-group-list-item.md",
-    "skills/lark-im/references/lark-im-feed-group-list.md",
-    "skills/lark-im/references/lark-im-feed-group-query-item.md",
-    "skills/lark-im/references/lark-im-feed-groups.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md",
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md",
-    "skills/lark-im/references/lark-im-flag-cancel.md",
-    "skills/lark-im/references/lark-im-flag-create.md",
-    "skills/lark-im/references/lark-im-flag-list.md",
-    "skills/lark-im/references/lark-im-message-enrichment.md",
-    "skills/lark-im/references/lark-im-messages-mget.md",
-    "skills/lark-im/references/lark-im-messages-reply.md",
-    "skills/lark-im/references/lark-im-messages-resources-download.md",
-    "skills/lark-im/references/lark-im-messages-search.md",
-    "skills/lark-im/references/lark-im-messages-send.md",
-    "skills/lark-im/references/lark-im-reactions.md",
-    "skills/lark-im/references/lark-im-threads-messages-list.md"
-  ],
-  "module_reach": {
-    "skills/lark-im/SKILL.md": {
-      "reach": 1.0,
-      "read_cases": [
-        "1",
-        "2",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "2",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": true
-    },
-    "skills/lark-im/references/lark-im-chat-create.md": {
-      "reach": 0.667,
-      "read_cases": [
-        "1",
-        "3"
-      ],
-      "actual_cases": [
-        "1",
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-identity.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-messages-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-search.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-chat-update.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-list-item.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-group-query-item.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-groups.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-create.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-feed-shortcut-remove.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-cancel.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-create.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-flag-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-message-enrichment.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-mget.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-reply.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-resources-download.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-search.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-messages-send.md": {
-      "reach": 0.333,
-      "read_cases": [
-        "3"
-      ],
-      "actual_cases": [
-        "3"
-      ],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-reactions.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    },
-    "skills/lark-im/references/lark-im-threads-messages-list.md": {
-      "reach": 0.0,
-      "read_cases": [],
-      "actual_cases": [],
-      "expected_cases": [],
-      "discoverability_miss": [],
-      "is_domain_skill": false
-    }
-  },
-  "expected_reach": {},
-  "minibatch": [
-    "1",
-    "2",
-    "3"
-  ],
-  "pareto_cases": [
-    "1",
-    "2",
-    "3"
-  ],
-  "artifacts": {
-    "workorder": "workorder.md",
-    "diagnosis": "diagnosis.md",
-    "attribution": "attribution.json",
-    "strategy": "strategy.md",
-    "review": "review.json",
-    "trend": "trend.json"
-  },
-  "code_tip": "cbd6e56ac07285fd973c53ff7382da0112b6cf5d",
-  "signature": "53194d7a111df326cc078b633f43587225bd0132",
-  "tier": "T1",
-  "intent": "dedup Commands<->Usage Scenarios overlap + compress --help-mirroring Common Errors in chat-create.md; keep all red lines (232043 two-step,succeed_type=1,chat-mode topic,--owner)",
-  "target_axis": "token",
-  "changed_files": [
-    "skills/lark-im/references/lark-im-chat-create.md"
-  ],
-  "decision_basis": {
-    "type": "module",
-    "module": "skills/lark-im/references/lark-im-chat-create.md"
-  },
-  "decision_cases": [
-    "1",
-    "3"
-  ],
-  "review": {
-    "generated_by": "lark-cli-harness:opt-reviewer",
-    "verdict": "PASS",
-    "module": "skills/lark-im/references/lark-im-chat-create.md",
-    "tier": "T1",
-    "round_index": 3,
-    "base_commit": "572eb8da41f608bd93b25916cac02cb772825b97",
-    "code_tip": "cbd6e56ac07285fd973c53ff7382da0112b6cf5d",
-    "reason": "纯瘦身改动，对抗式逐项核验未发现可证伪点：14 条承重红线（232043 两步流/succeed_type=1/chat-mode topic 与 thread 区分/--owner 默认/set-bot-manager/chat.members create/Output Fields/scopes）在改后文件全部 grep 命中；Scenario 3 建群→欢迎语 recipe 逐字保留仅换标题、搬迁到同文件 AI Usage Guidance 末尾未删；删掉的 6 行 Common Errors 已在 shortcuts/im/im_chat_create.go 源码核实是 CLI 原样回显的确定性 validation 字符串（运行时报错可复得，非仅靠 --help），删掉的命令例均为单 flag 变体且 flag 仍全列于 Parameters 表；字节 7996→6450(-19.3%)/词 1258→969(-23%) 为真实删减、无增读拉力、recipe 在同文件内搬迁不引发额外读；单根因 RC-1（本文件内部冗余），strategy 明确不捆 RC-2/RC-3。",
-    "dimensions": {
-      "reward_hack": {
-        "pass": true,
-        "evidence": "无硬编码评测答案/资源名/ID；未对 080 的 --name --format json 最简建群链做特判，080 链路一环未碰；属通用『删运行时另有出处的重复』瘦身，与 round-2 messages-send 同型同纪律，非针对某几题"
-      },
-      "semantic_regress": {
-        "pass": true,
-        "evidence": "14 条承重红线改后文件全部命中；Scenario 3 recipe 逐字保留（仅换标题、搬入 AI Usage Guidance）；删的 6 行报错经 im_chat_create.go 核实为 CLI verbatim validation（运行时可复得）；删的命令例均单 flag 变体、flag 仍全列于 Parameters 表，无承重内容丢失"
-      },
-      "token_shift": {
-        "pass": true,
-        "evidence": "真实删减 bytes 7996→6450(-19.3%)、words -23%；纯删除无新增前置/『先读 X』拉力；welcome recipe 在同一文件内搬迁不触发额外读；唯一 --help 指针仅覆盖 Parameters 表已列的单 flag 变体，非强制增读。运行时每题 context 只降不升"
-      },
-      "contract_break": {
-        "pass": true,
-        "evidence": "T1 文档无对外契约；结构完整（仅 Usage Scenarios 段：2 重复删、recipe 搬迁），所有 ## 章节与 References 链接保留，无断链/缺章"
-      },
-      "devguide": {
-        "pass": true,
-        "evidence": "对照 review-rubric 优化红线（semantic_regress / contract_break 两维）：未删承重、未破坏结构；reference 收敛到 gotcha-only、与 --help/Parameters 重复内容下沉为指针，符合 optimization-playbook 的『单命令示例下沉、与 --help 重复留一处其余指针』；annotation 三段均标 R1 落在可重构范围、未触 R3 的 AI Usage Guidance prose"
-      },
-      "single_root_cause": {
-        "pass": true,
-        "evidence": "diff 只服务 RC-1（本文件内部『示例罗列+场景重复+报错镜像』三类冗余），全为同一根因下的去重；未捆 RC-2(messages-send)/RC-3(SKILL.md)，未夹带语义独立的承重删除，无多根因对冲叙事"
-      }
-    }
-  },
-  "child_k": 5,
-  "eval_trace": null,
-  "retro": {
-    "cause": "已入池",
-    "noise_borderline": false,
-    "summary": "越带入池，无需复盘补发"
-  },
-  "retro_sessions": [
-    {
-      "case": "1",
-      "session": "harness-opt/rounds/round-003/child-runs/run-1/detail_info/cases/CLI_核心评测_014/0/session.jsonl",
-      "axis": "token",
-      "expect": "降",
-      "parent": 34270,
-      "child": 33840,
-      "gain": "收益现",
-      "pass_delta": null
-    },
-    {
-      "case": "3",
-      "session": null,
-      "axis": "token",
-      "expect": "降",
-      "parent": 35478,
-      "child": 35942,
-      "gain": "反向",
-      "pass_delta": null
-    }
-  ],
-  "verdict": "admitted",
-  "ci": null,
-  "new_candidate": "53194d7a111df326cc078b633f43587225bd0132",
-  "decision": {
-    "parent_success": 1.0,
-    "child_success": 1.0,
-    "parent_score": 0.8,
-    "child_score": 1.0,
-    "score_saved": 0.19999999999999996,
-    "score_threshold": 0.09532271373123208,
-    "parent_token": 34874.0,
-    "child_token": 34891.0,
-    "saved": -17.0,
-    "threshold": 4532.708313776408,
-    "parent_duration": 45267.5,
-    "child_duration": 43656.0,
-    "dur_saved": 1611.5,
-    "dur_threshold": 4899.200953624988,
-    "dur_margin": 1.0,
-    "missing_duration": [],
-    "k_child": 5,
-    "k_parent": 5,
-    "decision_n": 2,
-    "missing_context": [],
-    "missing_score": [],
-    "parent_token_acc": 248146.5,
-    "child_token_acc": 235911.0,
-    "phi0_score": 0.5666666666666667,
-    "eff_margin": 1.0,
-    "parent_token_full": 34874.0,
-    "child_token_full": 34891.0,
-    "saved_full": -17.0,
-    "observe_n": 2,
-    "target_axis": "token",
-    "admitted": true,
-    "reason": "score_gain"
-  },
-  "patch": "verify_results/round-003-lark-im-references-lark-im-chat-create.patch"
-}

harness-opt/rounds/round-003/strategy.md

@@ -1,43 +0,0 @@
-# Round 3 候选策略（模块=references/lark-im-chat-create.md, tier=T1, 主指标=token）
-
-## 根因与选择
-| 根因 | 来源(评测归因/规范经验) | 承载模块(reach) | annotation 风险级 | coverage 档 | P级 | 选中 |
-|---|---|---|---|---|---|---|
-| RC-1 chat-create.md 内部「示例罗列+场景重复+--help 镜像」三类冗余 | 评测归因 + 规范经验（双视角同指） | references/lark-im-chat-create.md(0.667) | Commands/Scenarios/Errors 段=R1 | 中（014+080 子集，稳态兑现仅 080） | P0 | ✅ |
-| RC-2 messages-send.md 内部冗余 | 评测归因 | references/lark-im-messages-send.md(0.333) | — | — | — | （round-2 已收割，不再是抓手） |
-| RC-3 SKILL.md 常驻正文 | 评测归因 | SKILL.md(1.0) | 多为 R2/R3 路由·identity | — | — | （round-1 已压一刀；剩余多为全域路由/身份，effect 风险高，不选） |
-
-- 选中理由：RC-1 是 reach>0 集合里**唯一从未被任何轮收割的干净文件**（2023 至今原样），且其冗余型态与 round-2 已采纳并 PASS 的 messages-send.md 完全同型（罗列+重复+--help 镜像）。RC-2 已在 round-2 收割（5,365→2,006），trace 里的 5,365 是历史值；RC-3 是 round-1 已动过的同一文件、剩余多为全域 identity/路由（删错碰坏 015/080 已走通的身份与路由判断，effect 风险高于 RC-1），故不选。
-- 选模块理由：chat-create.md reach=0.667（014+080 调用前都读到，状态③，非触达问题——纯减体积场景）；它正是承载 RC-1 的文件。未选 reach=0 的 22 个盲区 reference（改了也不在判决集、无法被采纳，触 reach 锁）。
-- 规范经验源补注：双视角同指一处。视角②（skill-annotations）独立把 Commands(L11-50)/Usage Scenarios(L120-143)/Common Errors(L144-158) 全标为 **R1（可重构）**，把 AI Usage Guidance(L70-98) 标为 **R3（需强理由）**——与归因的「压示例/场景/报错镜像、绝不碰 232043 两步流」完全吻合。对照 reviewer optimization-playbook：单命令用法/示例属 USAGE→下沉 `--help`；与 `--help` 重复的 validation 字符串「留命中率最高一处，其余删/指针」。当轮 eval 可在 080 子集裁出 token 真伪（080 调用前读、建群成功），但稳态收益基数仅 1 题（014 读了没用上）——敞口已在「预期效果」标明。
-
-## 改了什么（逐处）
-- **Commands(L12-50)** — 12 个 bash 示例（多条仅差一个 flag）压成 5 个差异化示例 + 一行 `--help` 指针。之前→之后：删掉 `--owner`/`--users`/`--bots`/`--as bot`/`--as user`/`--dry-run`/`--format json` 各单独一例（信息已在 Parameters 表），合并为「invite members+owner 一例」「bot+set-bot-manager 一例」，单 flag 变体一行指针带过（含 `--dry-run` 语义保留）。
-- **Usage Scenarios(L120-143)** — 整段 3 场景删除/搬迁。Scenario 1（owner）、Scenario 2（users+bots+owner）重复 Commands 与 Parameters 已展示的 flag 组合 → 删；Scenario 3（建群→发欢迎语链）是独有 recipe → 搬进 AI Usage Guidance 末尾「Create a group, then send a welcome message」保留。
-- **Common Errors(L144-158)** — 9 行压成 2 行。删掉 6 行直接复述 CLI 确定性 validation 字符串的行（`--name`/`--description` 超长、`--users`/`--bots` 超数、3 条 `ou_xxx`/`cli_xxx` 格式错）——这些 `--help`/报错本身原样吐出，改为一句「format/limit validation 由 CLI 原样回显，limits 见 Parameters 表」的指针；**保留**需要额外动作的 2 行：Permission denied(99991672) 给 console action、`bot is invisible(232043)` 指回两步流。
-
-## 为什么这么改（机制）
-- **省 context 的因果**：chat-create.md 是 lazy reference，读到即整文件进窗口（080/014 reach=0.667）。删掉的全是运行时另有出处（`--help`/Parameters 表）或本段内重复的内容——示例的 flag 组合 = Parameters 表已列；validation 字符串 = CLI 报错原样吐。删后 Agent 仍能：经 SKILL.md 选对 `+chat-create`、经 Parameters 表/`--help` 补全 flag 用法、遇 232043 走两步流。即 optimization-playbook §13 核心判据「删掉后 Agent 是否仍能选对命令并补到用法」——成立。
-- **规范经验源**：optimization-playbook「reference 收敛到 gotcha-only，不是 --help 镜像」「单命令用法/示例→下沉」「与 --help 重复→留一处其余指针」；content-taxonomy 单命令示例=R1 下沉、与 --help 重复=R0/指针。annotation 三段独立标 R1，本改动落在 R1 重构范围内，未触 R3 段。
-
-## 预期效果
-- 成功率（effect，硬门槛）：**不退化**。所有 effect 红线逐条保留并已 grep 校验（见下「刻意没做什么」）。080 实际只用 `--name --format json` 最简建群链——本改动未碰该链路任何一环；014 卡在跨域 contact+授权（非本 reference）。
-- context（分两层）：
-  - **(1) 静态字数差**：bytes 7996→6450（-19.3%）、chars -19.5%、words -23.0%、tiktoken(cl100k 代理) 2125→1714(-19.3%)。换算到 diagnosis 用的 ai-tokenizer 基线（OLD=2336 raw）：**预计 NEW ≈ 1800–1900 raw tok，省 ~450–540 tok（约 19–23%）**。
-  - **(2) 运行时 context 方向**：对**读到 chat-create.md 的题（080，及理论上 014）下降** ~450–540 tok；对没读它的题（015）**无影响**（015 大头是单次 `Read` 22.5k 工具输出，与本 reference 无关）。本改动是纯删减、无新增前置/增读拉力，不会抬升运行时 token。
-  - **与 direction 一致**（objective=降 token），无张力。
-- **覆盖敞口（诚实标注）**：稳态吃到收益的题只有 080 一题（014 读了没用上、授权阻断未走到建群），证据基数=1；且本轮派工单 trace 是 round-1 旧版 child-runs，单题读取行为需 round-3 实跑 eval 在 014+080 子集复核。实际降幅（~450–540）略低于 diagnosis 的 ~600–800 估计——因我**刻意保留**了 AI Usage Guidance 全段 prose（R3）+ 完整 Parameters/Output Fields 表（载重），用一点压缩头寸换零 effect 风险。
-
-## 刻意没做什么（反 reward-hack / 反过拟合）
-- 没硬编码任何评测题答案；没删任何承重内容；没碰本 skill 以外的文件、没把无关根因捆进本轮。
-- **逐条保留的载重红线（已 grep 校验存在）**：
-  - 232043 两步流全段（contact search → `--users 当前用户` 建群 → `chat.members create --as user` 加其他人 → 查 `invalid_id_list`）；
-  - `succeed_type=1` 语义解释；
-  - `--chat-mode topic` vs 「普通群 + `group_message_type=thread`」的区分注解；
-  - `--owner` 默认行为（bot 身份默认 bot / user 身份默认授权用户）；
-  - 全部 flag（含 `--set-bot-manager`、`--dry-run`、`--type public`、`--users/--bots` 上限与格式）、identity/scope 指引、互斥与护栏规则、Output Fields 全表。
-- 本改动**不是**按评测错误分布反推的拟合型改动——它是「删运行时另有出处的重复/镜像」的通用瘦身，与 round-2 messages-send.md 同型同纪律；非针对某几题的特判。
-- 未做 RC-3（SKILL.md 进一步压缩）：剩余多为全域 identity/路由，删错有 effect 风险，超出本轮低风险抓手范围。未做 015 的前置补充：那是增内容、与降 token 反向（方向冲突，diagnosis 已记录）。
-
-## 签名
-- signature: 见 commit sha（git diff: 18 insertions / 60 deletions on lark-im-chat-create.md）  tier: T1

harness-opt/rounds/round-003/trend.json

@@ -1,20 +0,0 @@
-[
-  {
-    "round": 1,
-    "n": 3,
-    "pass_n": 0,
-    "cmd_fail_rate": 0.6,
-    "tool_calls": 26.333333333333332,
-    "duration_ms": 50189.0,
-    "token": 31997.0
-  },
-  {
-    "round": 2,
-    "n": 3,
-    "pass_n": 3,
-    "cmd_fail_rate": 0.3466666666666667,
-    "tool_calls": 10.0,
-    "duration_ms": 69666.66666666667,
-    "token": 42377.333333333336
-  }
-]

harness-opt/rounds/round-003/workorder.md

@@ -1,44 +0,0 @@
-# Round 3 归因派工单（parent=557349b40feb359bb791749a37571d59edb7e72e；模块未定，由 candidate-writer 据诊断点名）
-
-> **只读输入**——opt-attributor 读本文件，把诊断**另写** `diagnosis.md`（给 candidate-writer）+ 逐题结构化 `attribution.json`（给 dashboard）。**不要覆盖本文件**，留作派工单↔诊断的前后对比。
-> 判分点只当「什么算挂」的锚，禁止照抄 grader 药方（已从派工单剔除）。
-
-## 模块运行时可达性（选模块第一步的证据；要选须在 strategy.md 说明理由）
-> reach=**实测**触达率（域主 SKILL.md 经 Skill 工具加载、reference 经 Read，都从 trace 实测，没有恒在的面）；判决集=实测∪预期触达。**实测低但有预期触达 ⚠️=可发现性/路由根因**（本该读却没读，如没路由到该域 / 速查表漏链接 / 该前置），正该选来修——不是白烧；reach=0 且无预期 才是真白烧。 **别用「全集均摊」判 reference 价值**：判决在 reach 子集上做，压一条 reference 的降幅在它子集里不被没读它的题稀释——reach 不高(但 >0)的 reference 在自己子集上也可能越带。
-- `skills/lark-im/SKILL.md` → reach=1.0  [域主 skill·经 Skill 工具加载]；判决集(实测∪预期): ['1', '2', '3']
-- `skills/lark-im/references/lark-im-chat-create.md` → reach=0.667；判决集(实测∪预期): ['1', '3']
-- `skills/lark-im/references/lark-im-messages-send.md` → reach=0.333；判决集(实测∪预期): ['3']
-- （另 22 个 reference reach=0 且无预期触达，本轮无关，略）
-
-## 逐轮诊断信号趋势（纯诊断，不进判决）
-
-| 轮 | 题数 | PASS | 命令失败率 | 工具调用 | 耗时(ms) | token |
-|---|---|---|---|---|---|---|
-| R1 | 3 | 0 | 0.60 | 26 | 50189 | 31997 |
-| R2 | 3 | 3 | 0.35 | 10 | 69667 | 42377 |
-
-> 跨题均值，按轮排。**命令失败率、工具调用数是横切诊断信号，不是准入轴**（准入只走 效果/token/耗时）——用来判「上一轮那刀有没有把失败/轮次压下去」。工具调用数比 wall-clock 稳，可给噪声大的耗时轴当旁证。
-
-### 1  [PASS]  ctx=34270  (acc=274608)  43995ms  tools=31
-- session.jsonl: harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_014/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✗ 使用当前用户身份创建名为「IM合作群」的群聊
-        证据: transcript 在展示授权二维码后结束，无任何 `lark-cli im +chat-create` 调用。执行停在 '授权完成后请告诉我，我会继续帮你创建群聊并发送消息'，群聊未创建。
-    ✗ 将傅一铭和傅二铭加入该群
-        证据: transcript 显示尝试搜索用户时遇到 `need_user_authorization` 错误，授权流程启动后中断。未获取到任何用户的 open_id，无后续添加操作。
-    ✗ 在该群发送文本消息「大家体验有问题随时沟通」，并返回可验证的 chat_id / message_id
-        证据: 群聊未创建，无 chat_id 可返回。transcript 无任何 `lark-cli im messages-send` 调用。
-
-### 2  [PASS]  ctx=47116  (acc=612048)  114310ms  tools=49
-- session.jsonl: harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_015/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✓ 成功定位名为「fusanming_at_openclaw群」的群，并获取最近包含「飞豆」关键字的消息
-    ✓ 将筛选出的相关消息内容转发到「fusanming_at_需求测试群」
-    ✓ 在「fusanming_at_需求测试群」中 @傅六铭 做知会，消息发送成功
-
-### 3  [PASS]  ctx=35478  (acc=221685)  46540ms  tools=22
-- session.jsonl: harness-opt/rounds/round-001/child-runs/run-1/detail_info/cases/CLI_核心评测_080/0/session.jsonl [native]
-- 判分点（grader 的「什么算挂」oracle，非药方）:
-    ✓ 使用用户身份创建一个名为「今晚吃什么」的群，预期返回 chat_id
-    ✓ 创建一张飞书卡片，卡片内容包含「今天晚上吃什么」
-    ✓ 将该卡片发送到新建群中，预期返回 message_id

harness-opt/trend.json

@@ -1,20 +0,0 @@
-[
-  {
-    "round": 1,
-    "n": 3,
-    "pass_n": 0,
-    "cmd_fail_rate": 0.6,
-    "tool_calls": 26.333333333333332,
-    "duration_ms": 50189.0,
-    "token": 31997.0
-  },
-  {
-    "round": 2,
-    "n": 3,
-    "pass_n": 3,
-    "cmd_fail_rate": 0.3466666666666667,
-    "tool_calls": 10.0,
-    "duration_ms": 69666.66666666667,
-    "token": 42377.333333333336
-  }
-]

harness-opt/verify_results/round-001-lark-im-SKILL.patch

@@ -1,152 +0,0 @@
-From 237a77feb341e15656386d6952a875dc459fec8c Mon Sep 17 00:00:00 2001
-From: "zhangheng.023" <zhangheng.023@bytedance.com>
-Date: Tue, 23 Jun 2026 18:27:25 +0800
-Subject: [PATCH] =?UTF-8?q?opt(round-001):=20SKILL.md=20=E2=80=94=20fold?=
- =?UTF-8?q?=20USAGE=20method-index=20+=20scope=20table=20into=20a=20schema?=
- =?UTF-8?q?=20pointer=20(-40%=20resident=20tokens)?=
-MIME-Version: 1.0
-Content-Type: text/plain; charset=UTF-8
-Content-Transfer-Encoding: 8bit
-
----
- skills/lark-im/SKILL.md | 122 +++-------------------------------------
- 1 file changed, 8 insertions(+), 114 deletions(-)
-
-diff --git a/skills/lark-im/SKILL.md b/skills/lark-im/SKILL.md
-index bc39aae1..ac1c6900 100644
---- a/skills/lark-im/SKILL.md
-+++ b/skills/lark-im/SKILL.md
-@@ -110,122 +110,16 @@ Shortcut 是对常用操作的高级封装（`lark-cli im +<verb> [flags]`）。
- | [`+feed-group-list`](references/lark-im-feed-group-list.md) | List the caller's feed groups (tags); user-only; supports `--page-all` auto-pagination |
- | [`+feed-group-list-item`](references/lark-im-feed-group-list-item.md) | List feed cards in a feed group (tag); user-only; enriches each item with chat_name resolved from feed_id; supports --page-all auto-pagination |
- | [`+feed-group-query-item`](references/lark-im-feed-group-query-item.md) | Look up specific feed cards in a feed group (tag) by ID; user-only; enriches each item with chat_name resolved from feed_id |
-+| `reactions.*` (add / delete / list / batch_query) | Add, remove, or read emoji reactions on a message; user/bot; caller must be in the conversation, and can only delete its own reactions. Read first: [`lark-im-reactions.md`](references/lark-im-reactions.md) |
-+| `feed.groups.*` (create / update / delete / batch_query / batch_add_item / batch_remove_item) | Manage feed groups (tags) and their member cards; user-only. Read first: [`lark-im-feed-groups.md`](references/lark-im-feed-groups.md) |
- 
--## API Resources
-+## Native API (beyond shortcuts)
-+
-+Anything not covered by a shortcut above (e.g. `chats.*`, `chat.members.*`, `chat.managers.*`, `chat.moderation.*`, `chat.user_setting.*`, `messages.delete|forward|merge_forward|read_users|urgent_*`, `threads.forward`, `images.create`, `pins.*`) is callable as a raw API:
- 
- ```bash
--lark-cli schema im.<resource>.<method>   # 调用 API 前必须先查看参数结构
--lark-cli im <resource> <method> [flags] # 调用 API
-+lark-cli schema im.<resource>.<method>    # MUST run first — gives params, identity (user/bot/tenant), and required scope
-+lark-cli im <resource> <method> [flags]   # then call
- ```
- 
--> **重要**：使用原生 API 时，必须先运行 `schema` 查看 `--data` / `--params` 参数结构，不要猜测字段格式。
--
--### chats
--
--  - `create` — 创建群。Identity: `bot` only (`tenant_access_token`).
--  - `get` — 获取群信息。Identity: supports `user` and `bot`; the caller must be in the target chat to get full details, and must belong to the same tenant for internal chats.
--  - `link` — 获取群分享链接。Identity: supports `user` and `bot`; the caller must be in the target chat, must be an owner or admin when chat sharing is restricted to owners/admins, and must belong to the same tenant for internal chats.
--  - `update` — 更新群信息。Identity: supports `user` and `bot`.
--
--### chat.members
--
--  - `bots` — 获取群内机器人列表。Identity: supports `user` and `bot`; the caller must be in the target chat and must belong to the same tenant for internal chats.
--  - `create` — 将用户或机器人拉入群聊。Identity: supports `user` and `bot`; the caller must be in the target chat; for `bot` calls, added users must be within the app's availability; for internal chats the operator must belong to the same tenant; if only owners/admins can add members, the caller must be an owner/admin, or a chat-creator bot with `im:chat:operate_as_owner`.
--  - `delete` — 将用户或机器人移出群聊。Identity: supports `user` and `bot`; only group owner, admin, or creator bot can remove others; max 50 users or 5 bots per request.
--  - `get` — 获取群成员列表。Identity: supports `user` and `bot`; the caller must be in the target chat and must belong to the same tenant for internal chats.
--
--### chat.user_setting
--
--  - `batch_query` — 批量查询当前用户在群内的个人偏好设置 (e.g. `is_muted` mutes normal messages, `is_mute_at_all` mutes @all messages); up to 10 chats per request. Identity: `user` only (`user_access_token`); the caller must be in each target chat.
--  - `batch_update` — 批量更新当前用户在群内的个人偏好设置 (e.g. `is_muted` mutes normal messages, `is_mute_at_all` mutes @all messages); up to 10 chats per request. Identity: `user` only (`user_access_token`); the caller must be in each target chat.
--
--### chat.managers
--
--  - `add_managers` — 指定群管理员。Identity: supports `user` and `bot`; only the group owner can add managers; max 10 managers per chat (20 for super-large chats), and at most 5 bots per request.
--  - `delete_managers` — 删除群管理员。Identity: supports `user` and `bot`; only the group owner can remove managers; max 50 users or 5 bots per request.
--
--### chat.moderation
--
--  - `get` — 获取群成员发言权限。Identity: supports `user` and `bot`; the caller must be in the target chat and belong to the same tenant.
--  - `update` — 更新群发言权限。Identity: supports `user` and `bot`; only the group owner (or creator bot with `im:chat:operate_as_owner`) can update; the caller must be in the chat.
--
--### messages
--
--  - `delete` — 撤回消息。Identity: supports `user` and `bot`; for `bot` calls, the bot must be in the chat to revoke group messages; to revoke another user's group message, the bot must be the owner, an admin, or the creator; for user P2P recalls, the target user must be within the bot's availability.
--  - `forward` — 转发消息。Identity: supports `user` and `bot`.
--  - `merge_forward` — 合并转发消息。Identity: `bot` only (`tenant_access_token`).
--  - `read_users` — 查询消息已读信息。Identity: `bot` only (`tenant_access_token`); the bot must be in the chat, and can only query read status for messages it sent within the last 7 days.
--  - `urgent_app` — 发送应用内加急。Identity: `bot` only (`tenant_access_token`); the bot must be the message sender and must be in the conversation that contains the message.
--  - `urgent_phone` — 发送电话加急。Identity: `bot` only (`tenant_access_token`); the bot must be the message sender and must be in the conversation that contains the message.
--  - `urgent_sms` — 发送短信加急。Identity: `bot` only (`tenant_access_token`); the bot must be the message sender and must be in the conversation that contains the message.
--
--### reactions
--
--  - `batch_query` — 批量获取消息表情。Identity: supports `user` and `bot`.[Must-read](references/lark-im-reactions.md)
--  - `create` — 添加消息表情回复。Identity: supports `user` and `bot`; the caller must be in the conversation that contains the message.[Must-read](references/lark-im-reactions.md)
--  - `delete` — 删除消息表情回复。Identity: supports `user` and `bot`; the caller must be in the conversation that contains the message, and can only delete reactions added by itself.[Must-read](references/lark-im-reactions.md)
--  - `list` — 获取消息表情回复。Identity: supports `user` and `bot`; the caller must be in the conversation that contains the message.[Must-read](references/lark-im-reactions.md)
--
--### threads
--
--  - `forward` — 转发话题。Identity: supports `user` and `bot`.
--
--### images
--
--  - `create` — 上传图片。Identity: `bot` only (`tenant_access_token`).
--
--### pins
--
--  - `create` — Pin 消息。Identity: supports `user` and `bot`.
--  - `delete` — 移除 Pin 消息。Identity: supports `user` and `bot`.
--  - `list` — 获取群内 Pin 消息。Identity: supports `user` and `bot`.
--
--### feed.groups
--
--  - `batch_add_item` — Batch add feed cards to a feed group. Identity: `user` only (`user_access_token`).[Must-read](references/lark-im-feed-groups.md)
--  - `batch_query` — Batch query feed groups. Identity: `user` only (`user_access_token`).[Must-read](references/lark-im-feed-groups.md)
--  - `batch_remove_item` — Batch remove feed cards from a feed group. Identity: `user` only (`user_access_token`).[Must-read](references/lark-im-feed-groups.md)
--  - `create` — Create a feed group. Identity: `user` only (`user_access_token`).[Must-read](references/lark-im-feed-groups.md)
--  - `delete` — Delete a feed group. Identity: `user` only (`user_access_token`).[Must-read](references/lark-im-feed-groups.md)
--  - `update` — Update a feed group. Identity: `user` only (`user_access_token`).[Must-read](references/lark-im-feed-groups.md)
--
--## 权限表
--
--| 方法 | 所需 scope |
--|------|-----------|
--| `chats.create` | `im:chat:create` |
--| `chats.get` | `im:chat:read` |
--| `chats.link` | `im:chat:read` |
--| `chats.update` | `im:chat:update` |
--| `chat.members.bots` | `im:chat.members:read` |
--| `chat.members.create` | `im:chat.members:write_only` |
--| `chat.members.delete` | `im:chat.members:write_only` |
--| `chat.members.get` | `im:chat.members:read` |
--| `chat.user_setting.batch_query` | `im:chat.user_setting:read` |
--| `chat.user_setting.batch_update` | `im:chat.user_setting:write` |
--| `chat.managers.add_managers` | `im:chat.managers:write_only` |
--| `chat.managers.delete_managers` | `im:chat.managers:write_only` |
--| `chat.moderation.get` | `im:chat.moderation:read` |
--| `chat.moderation.update` | `im:chat:moderation:write_only` |
--| `messages.delete` | `im:message:recall` |
--| `messages.forward` | `im:message` |
--| `messages.merge_forward` | `im:message` |
--| `messages.read_users` | `im:message:readonly` |
--| `messages.urgent_app` | `im:message.urgent` |
--| `messages.urgent_phone` | `im:message.urgent:phone` |
--| `messages.urgent_sms` | `im:message.urgent:sms` |
--| `reactions.batch_query` | `im:message.reactions:read` |
--| `reactions.create` | `im:message.reactions:write_only` |
--| `reactions.delete` | `im:message.reactions:write_only` |
--| `reactions.list` | `im:message.reactions:read` |
--| `threads.forward` | `im:message` |
--| `images.create` | `im:resource` |
--| `pins.create` | `im:message.pins:write_only` |
--| `pins.delete` | `im:message.pins:write_only` |
--| `pins.list` | `im:message.pins:read` |
--| `feed.groups.batch_add_item` | `im:feed_group_v1:write` |
--| `feed.groups.batch_query` | `im:feed_group_v1:read` |
--| `feed.groups.batch_remove_item` | `im:feed_group_v1:write` |
--| `feed.groups.create` | `im:feed_group_v1:write` |
--| `feed.groups.delete` | `im:feed_group_v1:write` |
--| `feed.groups.update` | `im:feed_group_v1:write` |
-+> **MUST** run `schema` before any native call: it is the live source for the `--data` / `--params` structure, the supported identity (`--as user` vs `--as bot`), owner/admin/tenant constraints, and the required `im:*` scope — do not guess. On a missing-scope error, lark-cli returns a `console_url`; follow the lark-shared permission-handling flow.
--- 
-2.50.1 (Apple Git-155)
-

harness-opt/verify_results/round-002-lark-im-references-lark-im-messages-send.patch

@@ -1,334 +0,0 @@
-From 82a099feafb45d101116f10230ce7c2f92fbcfe5 Mon Sep 17 00:00:00 2001
-From: "zhangheng.023" <zhangheng.023@bytedance.com>
-Date: Tue, 23 Jun 2026 19:17:24 +0800
-Subject: [PATCH] =?UTF-8?q?opt(round-002):=20lark-im-messages-send.md=20?=
- =?UTF-8?q?=E2=80=94=20consolidate=204x=20repeated=20content-flag=20rule,?=
- =?UTF-8?q?=20compress=20media=20enumeration=20&=20--help-mirror=20section?=
- =?UTF-8?q?s?=
-MIME-Version: 1.0
-Content-Type: text/plain; charset=UTF-8
-Content-Transfer-Encoding: 8bit
-
----
- .../references/lark-im-messages-send.md       | 259 ++++--------------
- 1 file changed, 51 insertions(+), 208 deletions(-)
-
-diff --git a/skills/lark-im/references/lark-im-messages-send.md b/skills/lark-im/references/lark-im-messages-send.md
-index 484c024f..32818909 100644
---- a/skills/lark-im/references/lark-im-messages-send.md
-+++ b/skills/lark-im/references/lark-im-messages-send.md
-@@ -1,10 +1,8 @@
- # im +messages-send
- 
--> **Prerequisite:** Read [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) first to understand authentication, global parameters, and safety rules.
-+> **Prerequisite:** Read [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) first for authentication, global parameters, and safety rules.
- 
--Send a message to a group chat or a direct message conversation. Supports both user identity (`--as user`) and bot identity (`--as bot`).
--
--This skill maps to the shortcut: `lark-cli im +messages-send` (internally calls `POST /open-apis/im/v1/messages`).
-+Send a message to a group chat (`--chat-id oc_xxx`) or a direct message (`--user-id ou_xxx`). One step, supports `--as user` and `--as bot` (default `bot`). Maps to shortcut `lark-cli im +messages-send` (`POST /open-apis/im/v1/messages`).
- 
- ## Safety Constraints
- 
-@@ -16,249 +14,94 @@ Messages sent by this tool are visible to other people. Before calling it, you *
- 
- **Do not** send messages without explicit user approval.
- 
--When using `--as bot`, the message is sent in the app's name, so make sure the app has already been added to the target chat.
--
--When using `--as user`, the message is sent as the authorized end user and requires the `im:message.send_as_user` and `im:message` scopes.
-+- `--as bot` (TAT, scope `im:message:send_as_bot`): the message is sent in the app's name — the app must already be in the target chat or have a DM relationship with the target user.
-+- `--as user` (UAT, scopes `im:message.send_as_user` + `im:message`): the message is sent as the authorized end user.
- 
- ## Choose The Right Content Flag
- 
--### Default Selection Rule For Agents
--
--- Prefer `--markdown` for headings, lists, links, summaries, reports, or Markdown-looking content.
--- Use `--text` for exact plain text: logs, code, indentation-sensitive text, or literal Markdown.
--- Use `--content` for exact `post` JSON, titles, multiple locales, cards, or unsupported structures.
--
--| Need | Recommended flag | Why |
--|------|------|------|
--| Send headings, lists, links, summaries, or reports | `--markdown` | Best default for lightweight formatting; converted to Feishu `post` JSON |
--| Send plain text exactly as written | `--text` | Preserves literal text; no Markdown conversion |
--| Precisely control the final payload | `--content` | You provide the exact JSON for `text` / `post` / `interactive` / `share_*` / media payloads |
--| Send image / file / video / audio | `--image` / `--file` / `--video` / `--audio` | Shortcut uploads URLs, or cwd-relative local files automatically |
--
--### `--text` vs `--markdown`
--
--- Use `--markdown` for lightweight formatted messages.
--- Use `--text` for exact plain text, especially logs, code, indentation, or Markdown characters that should **not** render.
--- Use `--content` when `--markdown` is not enough, especially if you need exact `post` JSON, a title, multiple locales, cards, or unsupported rich structures.
--
--## What `--markdown` Really Does
--
--`--markdown` accepts Markdown-like input and converts it to the Feishu `post` payload required by the message API.
--
--The shortcut does all of the following before sending:
--
--1. Forces `msg_type=post`
--2. Resolves remote Markdown images like `![x](https://...)` by downloading and uploading them first
--3. Normalizes the Markdown for Feishu post rendering
--4. Wraps the result as:
--
--```json
--{"zh_cn":{"content":[[{"tag":"md","text":"..."}]]}}
--```
--
--This makes `--markdown` the simplest path for lightweight formatted messages.
--
--### Markdown Boundaries
--
--- It does **not** promise full CommonMark / GitHub Flavored Markdown support.
--- It always becomes a `post` payload with a single `zh_cn` locale.
--- It does **not** let you set a `post` title. If you need a title, use `--msg-type post --content ...`.
--- Headings are rewritten:
--    - `# Title` becomes `#### Title`
--    - `##` to `######` are normalized to `#####` when the content contains H1-H3
--- Consecutive headings are separated with blank lines after heading normalization.
--- Block spacing and line breaks may be normalized during conversion.
--- Code blocks are preserved as code blocks.
--- Excess blank lines are compressed.
--- Already-uploaded `img_xxx` image keys are the most reliable Markdown image input.
--- Local paths in Markdown image syntax like `![x](./a.png)` are **not** supported and will not be auto-uploaded.
--- Remote URLs (`https://...`) will be auto-downloaded and uploaded at runtime; if the download or upload fails, the image is removed with a warning.
--
--If you need a title, multiple locales, cards, unsupported rich structures, or byte-for-byte post JSON control, use `--content` and provide the final JSON yourself.
-+| Content | Flag | Why |
-+|---|---|---|
-+| Headings, lists, links, summaries, reports (lightweight formatting) | `--markdown` | Best default; converted to Feishu `post` JSON |
-+| Exact plain text — logs, code, indentation, literal Markdown chars that must **not** render | `--text` | Preserves literal text; no conversion |
-+| Exact `post` JSON, a `post` title, multiple locales, cards (`interactive`), `share_*`, or unsupported structures | `--content` | You provide the final JSON; it must match the effective `--msg-type` |
-+| Image / file / video / audio | `--image` / `--file` / `--video` / `--audio` | Uploads URLs or cwd-relative local files automatically |
- 
--### Image Constraint for `--markdown`
-+These content flags (and the media flags) are **mutually exclusive** — pass exactly one. Media flags are also mutually exclusive with each other.
- 
--When using `--markdown` with images, prefer pre-uploading via `images.create` and referencing `![alt](img_xxx)` for predictable results. Remote URLs may work but are not guaranteed.
-+## `--markdown` Gotchas
- 
--**Steps:**
-+`--markdown` always forces `msg_type=post` (single `zh_cn` locale) and normalizes input for Feishu post rendering. Key boundaries (not full CommonMark/GFM):
- 
--```bash
--# 1. Upload image to get image_key
--lark-cli im images create --data '{"image_type":"message"}' --file ./diagram.png
--# Returns: {"image_key":"img_v3_xxxx"}
--
--# 2. Use image_key in --markdown
--lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Report\n\n![diagram](img_v3_xxxx)\n\nSee above for details.'
--```
-+- **No `post` title** — if you need one, use `--content` with `post` JSON.
-+- **Headings rewritten**: `# Title` → `#### Title`; `##`–`######` normalized to `#####` when content has H1–H3. Code blocks preserved; excess blank lines compressed.
-+- **Images**: pre-upload via `im images create` and reference `![alt](img_xxx)` for reliable results. Remote `https://` URLs are auto-downloaded+uploaded at runtime (removed with a warning if that fails). Local paths in `![x](./a.png)` are **not** supported and will not auto-upload.
- 
--## Preserving Formatting
-+## Preserving Exact Formatting
- 
--If the message has multiple lines, indentation, code blocks, tabs, or many quotes/backslashes, prefer shell ANSI-C quoting with `$'...'` for either `--markdown` or `--text`.
--
--This is especially useful in `zsh` / `bash` because it lets you write `\n` explicitly instead of relying on the shell to preserve literal newlines.
--
--### When formatting must be preserved
--
--Use `--text` plus `$'...'`:
--
--```bash
--lark-cli im +messages-send --chat-id oc_xxx --text $'Build failed\nBranch: feature/im-docs\nAction: please check logs'
--```
-+For multi-line text, indentation, code blocks, tabs, or many backslashes/quotes, use shell ANSI-C quoting `$'...'` so `\n` is written explicitly. Use `--text` + `$'...'` when the receiver must see the text exactly as entered:
- 
- ```bash
--lark-cli im +messages-send --chat-id oc_xxx --text $'```bash\nmake test\nmake lint\n```'
-+lark-cli im +messages-send --chat-id oc_xxx --text $'Build failed\nBranch: feature/x\nAction: check logs'
- ```
- 
--Use this path when you want the receiver to see the text exactly as entered, not a converted Markdown post.
--
- ## Commands
- 
- ```bash
--# Send a formatted update
-+# Formatted update (Markdown → post)
- lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Update\n\n- item 1\n- item 2'
- 
--# Send a plain one-line message
-+# Plain one-line text
- lark-cli im +messages-send --chat-id oc_xxx --text "Hello"
- 
--# Equivalent manual JSON
--lark-cli im +messages-send --chat-id oc_xxx --content '{"text":"Hello"}'
--
--# Send to a direct message (pass open_id)
-+# Direct message (pass open_id)
- lark-cli im +messages-send --user-id ou_xxx --text "Hello"
- 
--# Send multi-line text while preserving formatting
--lark-cli im +messages-send --chat-id oc_xxx --text $'Line 1\nLine 2\n  indented line'
--
--# Send Markdown with an image (must pre-upload via images.create)
--lark-cli im images create --data '{"image_type":"message"}' --file ./screenshot.png
--# Use the returned image_key in the markdown content
--lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Status\n\n![screenshot](img_v3_xxxx)\n\nDone.'
--
--# If you need exact post structure, send JSON directly
-+# Exact post structure with a title
- lark-cli im +messages-send --chat-id oc_xxx --msg-type post --content '{"zh_cn":{"title":"Title","content":[[{"tag":"text","text":"Body"}]]}}'
- 
--# Send a local image (uploaded automatically before sending)
--lark-cli im +messages-send --chat-id oc_xxx --image ./photo.png
--
--# Or send directly with an existing image_key
--lark-cli im +messages-send --chat-id oc_xxx --image img_xxx
-+# Markdown with an image (pre-upload first)
-+lark-cli im images create --data '{"image_type":"message"}' --file ./diagram.png   # -> {"image_key":"img_v3_xxxx"}
-+lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Report\n\n![diagram](img_v3_xxxx)\n\nDone.'
- 
--# Send a local file (uploaded automatically before sending)
-+# Media (local files uploaded automatically; --video requires --video-cover)
-+lark-cli im +messages-send --chat-id oc_xxx --image ./photo.png
- lark-cli im +messages-send --chat-id oc_xxx --file ./report.pdf
--
--# Send a video (--video-cover is required as the cover)
- lark-cli im +messages-send --chat-id oc_xxx --video ./demo.mp4 --video-cover ./cover.png
--lark-cli im +messages-send --chat-id oc_xxx --video ./demo.mp4 --video-cover img_xxx
--
--# Send audio
- lark-cli im +messages-send --chat-id oc_xxx --audio ./voice.opus
- 
--# Use an idempotency key (same key sends only once within 1 hour)
--lark-cli im +messages-send --chat-id oc_xxx --text "Hello" --idempotency-key my-unique-id
--
--# Preview the request without executing it
--lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Test\n\nhello' --dry-run
-+# Idempotency (same key sends only once within 1 hour) / preview without sending
-+lark-cli im +messages-send --chat-id oc_xxx --text "Hi" --idempotency-key my-id
-+lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Test\n\nhi' --dry-run
- ```
- 
--## Media Input Rules
--
--- Media flags accept an existing key (`img_xxx` / `file_xxx`), an `http://` or `https://` URL, or a local file path.
--- Local paths must be relative to the current working directory and stay within it after resolving `..` and symlinks.
--- Absolute paths such as `/tmp/photo.png` are rejected. Run the command from the file's directory and pass `./photo.png`, or copy the file into the current directory first.
--
--## Parameters
--
--| Parameter | Required | Description                                                                                                                                                                                   |
--|------|------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
--| `--chat-id <id>` | One of two | Group chat ID (`oc_xxx`)                                                                                                                                                                      |
--| `--user-id <id>` | One of two | User open_id (`ou_xxx`) for direct messages                                                                                                                                                   |
--| `--text <string>` | One content option | Plain text message. Use when exact text and formatting preservation matter. Automatically wrapped as `{"text":"..."}`                                                                         |
--| `--markdown <string>` | One content option | Best default for lightweight formatted messages such as headings, lists, links, summaries, and reports. Internally converted to `post` JSON with Feishu-specific normalization               |
--| `--content <json>` | One content option | Exact message content JSON string; use this when you need full control over `msg_type` and payload. The JSON must match the effective `--msg-type`                                            |
--| `--image <path\|url\|key>` | One content option | Cwd-relative local image path, URL, or `image_key` (`img_xxx`). Local paths and URLs are uploaded automatically                                                                               |
--| `--file <path\|url\|key>` | One content option | Cwd-relative local file path, URL, or `file_key` (`file_xxx`). Local paths and URLs are uploaded automatically                                                                                |
--| `--video <path\|url\|key>` | One content option | Cwd-relative local video path, URL, or `file_key` (`file_xxx`). Local paths and URLs are uploaded automatically. **Must be paired with `--video-cover`**                                      |
--| `--video-cover <path\|url\|key>` | **Required with `--video`** | Cwd-relative local cover image path, URL, or `image_key` (`img_xxx`). Local paths and URLs are uploaded automatically                                                                         |
--| `--audio <path\|url\|key>` | One content option | Cwd-relative local audio path, URL, or `file_key` (`file_xxx`). Local paths and URLs are uploaded automatically                                                                                           |
--| `--msg-type <type>` | No | Message type (default `text`). If you use `--text` / `--markdown` / media flags, the effective type is inferred automatically. Explicitly setting a conflicting `--msg-type` fails validation |
--| `--idempotency-key <key>` | No | Idempotency key; the same key sends only one message within 1 hour                                                                                                                            |
--| `--as <identity>` | No | Identity type: `bot` or `user` (default `bot`)                                                                                                                                                |
--| `--dry-run` | No | Print the request only, do not execute it                                                                                                                                                     |
--
--> **Mutual exclusivity rule:** `--text`, `--markdown`, `--content`, and `--image`/`--file`/`--video`/`--audio` cannot be used together. Media flags are also mutually exclusive with each other.
-->
--> **Video cover rule:** `--video` **must** be accompanied by `--video-cover`. Omitting `--video-cover` when using `--video` will fail validation. `--video-cover` cannot be used without `--video`.
--
--## Common Mistakes
--
--- Choosing `--text` for headings, lists, links, summaries, or reports. Use `--markdown`.
--- Choosing `--markdown` when you actually need exact plain text. If exact line breaks, spacing, logs, code, or literal Markdown characters matter, use `--text`, usually with `$'...'`.
--- Assuming `--markdown` supports every Markdown feature. It is converted into a Feishu `post` payload and normalized first.
--- Putting local image paths inside Markdown like `![x](./a.png)`. `--markdown` does not auto-upload those paths.
--- **Using local file paths inside Markdown image syntax** (e.g. `![x](./a.png)`) with `--markdown`. Local paths are not auto-uploaded and will not render as an image. Pre-upload via `images.create` to get an `image_key` instead.
--- Using `--content` without making the JSON match the effective `--msg-type`.
--- Explicitly setting `--msg-type` to something that conflicts with `--text`, `--markdown`, or media flags.
--- Mixing `--text`, `--markdown`, or `--content` with media flags in one command.
--
--## `content` Format Reference
-+Run `lark-cli im +messages-send --help` for the full flag list and types. Load-bearing rules that `--help` may not make obvious:
-+
-+- **Media paths** accept an existing key (`img_xxx`/`file_xxx`), an `http(s)://` URL, or a **cwd-relative** local path. Absolute paths (e.g. `/tmp/x.png`) are rejected — run from the file's directory and pass `./x.png`. Upload and send use the same identity.
-+- **`--video` must be paired with `--video-cover`** (image key/URL/local path); `--video-cover` cannot be used alone.
-+- **`--msg-type`** is inferred from `--text`/`--markdown`/media flags; explicitly setting a conflicting type fails validation.
-+
-+## `content` Format Reference (for `--content`)
- 
- | `msg_type` | Example `content` |
--|----------|-------------|
-+|---|---|
- | `text` | `{"text":"Hello <at user_id=\"ou_xxx\">name</at>"}` |
- | `post` | `{"zh_cn":{"title":"Title","content":[[{"tag":"text","text":"Body"}]]}}` |
--| `image` | `{"image_key":"img_xxx"}` |
--| `file` | `{"file_key":"file_xxx"}` |
--| `audio` | `{"file_key":"file_xxx"}` |
--| `media` | `{"file_key":"file_xxx","image_key":"img_xxx"}` (video; `image_key` is the cover from `--video-cover` — **required**) |
--| `share_chat` | `{"chat_id":"oc_xxx"}` |
--| `share_user` | `{"user_id":"ou_xxx"}` |
--| `interactive` | Card JSON (see Feishu interactive card documentation) |
-+| `image` / `file` / `audio` | `{"image_key":"img_xxx"}` / `{"file_key":"file_xxx"}` / `{"file_key":"file_xxx"}` |
-+| `media` (video) | `{"file_key":"file_xxx","image_key":"img_xxx"}` (`image_key` is the **required** cover) |
-+| `share_chat` / `share_user` | `{"chat_id":"oc_xxx"}` / `{"user_id":"ou_xxx"}` |
-+| `interactive` (card) | Card JSON (see Feishu interactive card docs) |
- 
--## Return Value
--
--```json
--{
--  "message_id": "om_xxx",
--  "chat_id": "oc_xxx",
--  "create_time": "1234567890"
--}
--```
-+When using `--content`, you are responsible for making the JSON match the effective `msg_type`.
- 
- ## @Mention Format
- 
--The `<at>` syntax differs by message type. The shortcut only normalizes mentions for `text` and `post`; `interactive` card content is passed through verbatim, so cards must use the card-native syntax below.
--
--### `text`
--
--- `<at user_id="ou_xxx">name</at>` — the inner text is the mentioned user's display name and is optional (`<at user_id="ou_xxx"></at>` also works)
--- @all: `<at user_id="all"></at>`
--
--### `post`
-+The `<at>` syntax differs by message type; the shortcut normalizes mentions for `text` and `post` only — `interactive` cards are passed through verbatim.
- 
--- Inside a `text` or `md` element, the same inline form as `text` works: `<at user_id="ou_xxx">name</at>`
--- Or use a dedicated `at` element node: `{"tag":"at","user_id":"ou_xxx"}` (use `"all"` to mention everyone)
-+- **`text`** / inside a `post` `text`/`md` element: `<at user_id="ou_xxx">name</at>` (inner name optional); @all: `<at user_id="all"></at>`. In `post` you may also use a node: `{"tag":"at","user_id":"ou_xxx"}` (`"all"` for everyone).
-+- **`interactive` (card)** — card-native syntax inside a `lark_md`/`markdown` element: `<at id=ou_xxx></at>`, multiple `<at ids=ou_1,ou_2></at>`, by email `<at email=user@example.com></at>`.
- 
--### `interactive` (card)
--
--Card content is **not** normalized — use the card-native `<at>` syntax inside a `lark_md` / `markdown` element:
--
--- single user by open_id: `<at id=ou_xxx></at>`
--- multiple users: `<at ids=ou_xxx1,ou_xxx2></at>`
--- by email: `<at email=user@example.com></at>`
--
--## Notes
-+## Return Value
- 
--- `--chat-id` and `--user-id` are mutually exclusive; you must provide exactly one
--- `--content` must be valid JSON
--- When using `--content`, you are responsible for making the JSON structure match the effective `msg_type`
--- `--image`/`--file`/`--video`/`--audio` support existing keys, URLs, and cwd-relative local file paths; the shortcut uploads local paths and URLs first, then sends the message; both the upload and send steps use the same identity (UAT when `--as user`, TAT when `--as bot`)
--- If the provided media value starts with `img_` or `file_`, it is treated as an existing key and used directly
--- `--markdown` always sends `msg_type=post`, even if you do not explicitly set `--msg-type post`
--- If you explicitly set `--msg-type` and it conflicts with the chosen content flag, validation fails
--- When using `--video`, `--video-cover` is required as the video cover
--- `--dry-run` uses placeholder image keys for remote Markdown images and placeholder media keys for local uploads
--- Failures return an error code and message
--- `--as user` uses a user access token (UAT) and requires the `im:message.send_as_user` and `im:message` scopes; the message is sent as the authorized end user
--- `--as bot` uses a tenant access token (TAT) and requires the `im:message:send_as_bot` scope
--- When sending as a bot, the app must already be in the target group or already have a direct-message relationship with the target user
--- When using `--markdown` with images, pre-uploading via `images.create` to obtain an `image_key` is recommended for reliability; remote URLs may be auto-resolved at runtime, but if download/upload fails the image is removed with a warning; local paths are not supported
-+```json
-+{"message_id": "om_xxx", "chat_id": "oc_xxx", "create_time": "1234567890"}
-+```
--- 
-2.50.1 (Apple Git-155)
-

harness-opt/verify_results/round-003-lark-im-references-lark-im-chat-create.patch

@@ -1,135 +0,0 @@
-From cbd6e56ac07285fd973c53ff7382da0112b6cf5d Mon Sep 17 00:00:00 2001
-From: "zhangheng.023" <zhangheng.023@bytedance.com>
-Date: Tue, 23 Jun 2026 19:51:49 +0800
-Subject: [PATCH] =?UTF-8?q?opt(round-003):=20references/lark-im-chat-creat?=
- =?UTF-8?q?e.md=20=E2=80=94=20dedup=20Commands/Scenarios=20overlap=20+=20c?=
- =?UTF-8?q?ompress=20--help-mirroring=20Common=20Errors=20into=20pointers,?=
- =?UTF-8?q?=20keep=20232043=20two-step=20flow=20&=20all=20guardrails?=
-MIME-Version: 1.0
-Content-Type: text/plain; charset=UTF-8
-Content-Transfer-Encoding: 8bit
-
----
- .../lark-im/references/lark-im-chat-create.md | 78 +++++--------------
- 1 file changed, 18 insertions(+), 60 deletions(-)
-
-diff --git a/skills/lark-im/references/lark-im-chat-create.md b/skills/lark-im/references/lark-im-chat-create.md
-index 76716f76..7d65e5d3 100644
---- a/skills/lark-im/references/lark-im-chat-create.md
-+++ b/skills/lark-im/references/lark-im-chat-create.md
-@@ -12,43 +12,24 @@ This skill maps to the shortcut: `lark-cli im +chat-create` (internally calls `P
- ## Commands
- 
- ```bash
--# Create a private group (default)
-+# Private group (default)
- lark-cli im +chat-create --name "My Group"
- 
--# Create a public group (name is required and must be at least 2 characters)
-+# Public group (--name required, min 2 chars)
- lark-cli im +chat-create --name "Public Group" --type public
- 
--# Create a topic chat
-+# Topic chat (a 话题群; see note under Parameters)
- lark-cli im +chat-create --name "Topic Group" --chat-mode topic
- 
--# Specify the group owner
--lark-cli im +chat-create --name "My Group" --owner ou_xxx
-+# Invite members and set owner (users: up to 50 ou_xxx; bots: up to 5 cli_xxx)
-+lark-cli im +chat-create --name "My Group" --owner ou_xxx --users "ou_aaa,ou_bbb" --bots "cli_aaa"
- 
--# Invite user members (comma-separated open_ids, up to 50)
--lark-cli im +chat-create --name "My Group" --users "ou_aaa,ou_bbb"
--
--# Invite bot members (comma-separated app IDs, up to 5)
--lark-cli im +chat-create --name "My Group" --bots "cli_aaa,cli_bbb"
--
--# Invite both users and bots
--lark-cli im +chat-create --name "My Group" --users "ou_aaa" --bots "cli_aaa"
--
--# Make the creating bot a group manager (bot identity only)
--lark-cli im +chat-create --name "My Group" --set-bot-manager --as bot
--
--# JSON output
--lark-cli im +chat-create --name "My Group" --format json
--
--# Create a group with bot identity
--lark-cli im +chat-create --name "My Group" --users "ou_aaa" --as bot
--
--# Create a group with user identity
--lark-cli im +chat-create --name "My Group" --users "ou_aaa,ou_bbb" --as user
--
--# Preview the request without creating anything
--lark-cli im +chat-create --name "My Group" --dry-run
-+# Bot identity, making the creating bot a manager
-+lark-cli im +chat-create --name "My Group" --users "ou_aaa" --as bot --set-bot-manager
- ```
- 
-+Run `lark-cli im +chat-create --help` for the full flag list, limits, and types. Single-flag variations (`--as user`, `--description`, `--format json`, `--dry-run` preview, etc.) follow the Parameters table below — `--dry-run` previews the request without creating anything.
-+
- ## Parameters
- 
- | Parameter | Required | Limits | Description |
-@@ -106,6 +87,13 @@ lark-cli im +chat-create --name "<group name>" --users "ou_aaa,ou_bbb" --as user
- 
- The authorized user is automatically the group creator and member.
- 
-+### Create a group, then send a welcome message
-+
-+```bash
-+CHAT_ID=$(lark-cli im +chat-create --name "New Group" --format json | jq -r '.data.chat_id')
-+lark-cli im +messages-send --chat-id "$CHAT_ID" --text "Welcome, everyone!"
-+```
-+
- ## Output Fields
- 
- | Field | Description |
-@@ -117,43 +105,13 @@ The authorized user is automatically the group creator and member.
- | `external` | Whether the group is external |
- | `share_link` | Group share link (omitted if retrieval fails) |
- 
--## Usage Scenarios
--
--### Scenario 1: Create a group and specify the owner
--
--```bash
--lark-cli im +chat-create --name "Project Discussion Group" --owner ou_xxx
--```
--
--### Scenario 2: Create a group and invite users and a bot
--
--```bash
--lark-cli im +chat-create --name "Project Discussion Group" \
--  --owner ou_xxx \
--  --users "ou_aaa,ou_bbb" \
--  --bots "cli_aaa"
--```
--
--### Scenario 3: Create a group and send a welcome message
--
--```bash
--CHAT_ID=$(lark-cli im +chat-create --name "New Group" --format json | jq -r '.data.chat_id')
--lark-cli im +messages-send --chat-id "$CHAT_ID" --text "Welcome, everyone!"
--```
--
- ## Common Errors and Troubleshooting
- 
-+Format/limit validation (`--name`/`--description`/`--users`/`--bots`/`--owner` length, count, and `ou_xxx`/`cli_xxx` format) is enforced by the CLI and reported verbatim with the fix — see the Parameters table for limits. The two errors needing extra action:
-+
- | Symptom | Root Cause | Solution |
- |---------|---------|---------|
- | Permission denied (99991672) | The app does not have `im:chat:create` (bot) or `im:chat:create_by_user` (user) permission enabled | Enable the required permission for the app in the Open Platform console |
--| `--name is required for public groups and must be at least 2 characters` | A public group was created without a name or with a name shorter than 2 characters | Provide a name with at least 2 characters |
--| `--name exceeds the maximum of 60 characters` | The group name is too long | Shorten the name to 60 characters or fewer |
--| `--description exceeds the maximum of 100 characters` | The group description is too long | Shorten the description to 100 characters or fewer |
--| `--users exceeds the maximum of 50` | Too many user members were provided | Split the operation into batches and add more members later |
--| `--bots exceeds the maximum of 5` | Too many bot members were provided | Invite at most 5 bots at once |
--| `invalid user id: expected open_id (ou_xxx)` | Invalid user ID format | Use the `ou_xxx` format for users |
--| `invalid bot id: expected app ID (cli_xxx)` | Invalid bot ID format | Use the `cli_xxx` format for bots |
--| `invalid --owner: expected open_id (ou_xxx)` | Invalid owner ID format | Use the `ou_xxx` format for the owner |
- | `bot is invisible to user` (232043) | The bot and target users are mutually invisible | Follow the two-step flow in AI Usage Guidance above — do not pass other users in `--users` during creation |
- 
- ## References
--- 
-2.50.1 (Apple Git-155)
-

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@larksuite/cli",
-  "version": "1.0.56",
+  "version": "1.0.57",
   "description": "The official CLI for Lark/Feishu open platform",
   "bin": {
     "lark-cli": "scripts/run.js"

scripts/semantic-review-workflow.test.sh

@@ -179,7 +179,10 @@ fi
 require_in_step "$summary_verify_step" 'workflowPath !== ".github/workflows/ci.yml"' "PR quality summary must verify the triggering workflow path"
 require_in_step "$summary_verify_step" 'run.event !== "pull_request"' "PR quality summary must only handle pull_request workflow_run events"
 require_in_step "$summary_verify_step" 'run.repository.id !== context.payload.repository.id' "PR quality summary must verify workflow_run repository id"
+require_in_step "$summary_verify_step" 'const targetHeadSha = run.head_sha' "PR quality summary must use the CI run head SHA as the verified PR head"
+require_in_step "$summary_verify_step" 'eventHeadSha && eventHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()' "PR quality summary should tolerate mutable workflow_run PR head metadata"
 require_in_step "$summary_verify_step" 'factsArtifactPattern' "PR quality summary should use the base-bound facts artifact name when available"
+require_in_step "$summary_verify_step" 'const baseSha = artifactBaseSha || eventBaseSha || pr.base.sha' "PR quality summary must prefer the CI-time artifact base SHA"
 require_in_step "$summary_verify_step" 'core.setOutput("artifact_error"' "PR quality summary must expose artifact binding failures"
 require_in_step "$summary_artifact_step" 'factsArtifactName' "PR quality summary artifact step must use the verified facts artifact binding"
 require_in_step "$summary_extract_facts_step" 'SEMANTIC_REVIEW_DECISION_OUT' "PR quality summary artifact verifier must write an infrastructure decision on verifier failure"
@@ -198,8 +201,9 @@ require_in_step "$verify_step" 'workflowPath !== ".github/workflows/ci.yml"' "se
 require_in_step "$verify_step" 'run.repository.id !== context.payload.repository.id' "semantic-review must verify workflow_run repository id"
 require_in_step "$verify_step" 'run.event !== "pull_request"' "semantic-review must only handle pull_request workflow_run events"
 require_in_step "$verify_step" 'run.conclusion !== "success"' "semantic-review must only consume successful CI runs"
-require_in_step "$verify_step" 'const eventHeadSha = runPRs[0]?.head?.sha || ""' "semantic-review must prefer workflow_run PR head when GitHub provides it"
-require_in_step "$verify_step" 'const targetHeadSha = eventHeadSha || run.head_sha' "semantic-review target PR head must come from the workflow_run event"
+require_in_step "$verify_step" 'const eventHeadSha = runPRs[0]?.head?.sha || ""' "semantic-review should inspect workflow_run PR head metadata"
+require_in_step "$verify_step" 'const targetHeadSha = run.head_sha' "semantic-review target PR head must come from the completed CI run"
+require_in_step "$verify_step" 'eventHeadSha && eventHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()' "semantic-review should tolerate mutable workflow_run PR head metadata"
 require_in_step "$verify_step" 'factsArtifactPattern' "semantic-review must use a base-bound facts artifact name"
 require_in_step "$verify_step" 'listWorkflowRunArtifacts' "semantic-review must read the workflow_run artifacts before resolving fallback base SHA"
 require_in_step "$verify_step" 'artifactHeadSha.toLowerCase() !== targetHeadSha.toLowerCase()' "semantic-review must not let the artifact choose a different PR head"
@@ -210,8 +214,8 @@ require_in_step "$verify_step" 'commit_sha: targetHeadSha' "semantic-review fall
 require_in_step "$verify_step" 'github.rest.pulls.list' "semantic-review must have a pull-list fallback when commit association is empty"
 require_in_step "$verify_step" 'candidatePRs.length > 1' "semantic-review must fail closed when commit-to-PR fallback is ambiguous"
 require_in_step "$verify_step" 'pr.head.sha !== targetHeadSha' "semantic-review must skip stale PR heads"
-require_in_step "$verify_step" 'eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()' "semantic-review must reject mismatched event and artifact base SHAs"
-require_in_step "$verify_step" 'const baseSha = eventBaseSha || artifactBaseSha' "semantic-review fallback must use the CI-time artifact base SHA"
+require_in_step "$verify_step" 'eventBaseSha && parsedBaseSha.toLowerCase() !== eventBaseSha.toLowerCase()' "semantic-review should tolerate mutable workflow_run PR base metadata"
+require_in_step "$verify_step" 'const baseSha = artifactBaseSha || eventBaseSha || pr.base.sha' "semantic-review must prefer the CI-time artifact base SHA"
 require_in_step "$verify_step" 'pr.base.sha !== baseSha' "semantic-review must skip stale PR bases"
 require_in_step "$verify_step" 'core.setOutput("run_id"' "semantic-review must pass verified workflow run id to publisher"
 require_in_step "$verify_step" 'core.setOutput("head_repo_id"' "semantic-review must pass verified head repo id"

shortcuts/apps/apps_html_publish.go

@@ -7,6 +7,7 @@ import (
 	"context"
 	"fmt"
 	"io"
+	"path/filepath"
 	"strings"
 
 	"github.com/larksuite/cli/extension/fileio"
@@ -84,6 +85,9 @@ var AppsHTMLPublish = common.Shortcut{
 			// for dry-run "advisory preview" semantics).
 			dry.Set("validation_error", err.Error())
 		}
+		if hits := oversizeHTMLFiles(candidates); len(hits) > 0 {
+			dry.Set("oversize_html", hits)
+		}
 		dry.Set("file_count", len(candidates))
 		var totalSize int64
 		names := make([]string, 0, len(candidates))
@@ -140,18 +144,22 @@ type appsHTMLPublishSpec struct {
 // per-environment .env.* files for every stage).
 const maxSensitiveListInError = 5
 
+// truncatedJoin joins items with ", ", capping at max entries and appending
+// "(and N more)" for the remainder, so an inline error list stays readable when
+// a payload has many hits.
+func truncatedJoin(items []string, max int) string {
+	if len(items) <= max {
+		return strings.Join(items, ", ")
+	}
+	return strings.Join(items[:max], ", ") + fmt.Sprintf(" (and %d more)", len(items)-max)
+}
+
 // sensitiveCandidatesError builds the Validate-time rejection when --path
 // contains credential files and --allow-sensitive was not set.
 func sensitiveCandidatesError(hits []string) error {
-	var sample string
-	if len(hits) <= maxSensitiveListInError {
-		sample = strings.Join(hits, ", ")
-	} else {
-		sample = strings.Join(hits[:maxSensitiveListInError], ", ") +
-			fmt.Sprintf(" (and %d more)", len(hits)-maxSensitiveListInError)
-	}
 	return appsValidationParamError("--path",
-		"--path contains %d credential file(s) that should not be published: %s", len(hits), sample).
+		"--path contains %d credential file(s) that should not be published: %s",
+		len(hits), truncatedJoin(hits, maxSensitiveListInError)).
 		WithHint("remove these files from the publish payload, OR pass --allow-sensitive if shipping them is intentional (e.g. a docs site demoing credential-file formats)")
 }
 
@@ -168,6 +176,30 @@ var maxHTMLPublishTarballBytes int64 = 20 * 1024 * 1024
 // Mutable for tests.
 var maxHTMLPublishRawBytes int64 = 200 * 1024 * 1024
 
+// maxHTMLPublishSingleHTMLFileBytes 单个 .html 文件上限，对齐妙搭服务端 10MB 约束。
+// 用 var 而非 const，便于单测调小覆盖拦截路径。
+var maxHTMLPublishSingleHTMLFileBytes int64 = 10 * 1024 * 1024
+
+// oversizeHTMLFiles 返回 candidates 中扩展名为 .html（大小写不敏感）且单个 Size 超过
+// maxHTMLPublishSingleHTMLFileBytes 的 RelPath 列表。只针对 .html 文件，不波及图片/字体/JS。
+func oversizeHTMLFiles(candidates []htmlPublishCandidate) []string {
+	var hits []string
+	for _, c := range candidates {
+		if strings.EqualFold(filepath.Ext(c.RelPath), ".html") && c.Size > maxHTMLPublishSingleHTMLFileBytes {
+			hits = append(hits, c.RelPath)
+		}
+	}
+	return hits
+}
+
+// oversizeHTMLFilesError 构造单文件超限的 Validate 风格拒绝。
+func oversizeHTMLFilesError(hits []string) error {
+	return appsValidationParamError("--path",
+		"--path contains %d HTML file(s) exceeding the %d bytes (10MB) per-file limit: %s",
+		len(hits), maxHTMLPublishSingleHTMLFileBytes, truncatedJoin(hits, maxSensitiveListInError)).
+		WithHint("split or trim oversized HTML file(s); the 10MB cap applies to each single .html file")
+}
+
 // ensureIndexHTML 要求 walker 抓到的 candidates 里必须含 index.html。
 // 目录形态：根目录下必须有 index.html。
 // 单文件形态：文件名必须就是 index.html。
@@ -190,6 +222,9 @@ func runHTMLPublish(ctx context.Context, fio fileio.FileIO, publisher appsHTMLPu
 	if err := ensureIndexHTML(candidates); err != nil {
 		return nil, err
 	}
+	if hits := oversizeHTMLFiles(candidates); len(hits) > 0 {
+		return nil, oversizeHTMLFilesError(hits)
+	}
 	var rawTotal int64
 	for _, c := range candidates {
 		rawTotal += c.Size

shortcuts/apps/apps_html_publish_test.go

@@ -503,3 +503,82 @@ func TestRunHTMLPublish_RejectsOversizeRawCandidates(t *testing.T) {
 		t.Fatalf("client must not be called when raw cap hit")
 	}
 }
+
+func TestOversizeHTMLFiles(t *testing.T) {
+	orig := maxHTMLPublishSingleHTMLFileBytes
+	maxHTMLPublishSingleHTMLFileBytes = 100
+	defer func() { maxHTMLPublishSingleHTMLFileBytes = orig }()
+
+	cands := []htmlPublishCandidate{
+		{RelPath: "index.html", Size: 50},
+		{RelPath: "big.html", Size: 4096},
+		{RelPath: "BIG.HTML", Size: 4096}, // 大小写不敏感
+		{RelPath: "huge.png", Size: 9000}, // 非 .html，忽略
+	}
+	hits := oversizeHTMLFiles(cands)
+	if len(hits) != 2 {
+		t.Fatalf("hits=%v, want [big.html BIG.HTML]", hits)
+	}
+	for _, h := range hits {
+		if h == "huge.png" || h == "index.html" {
+			t.Fatalf("unexpected hit %q", h)
+		}
+	}
+}
+
+func TestMaxHTMLPublishSingleHTMLFileBytes_Default(t *testing.T) {
+	if maxHTMLPublishSingleHTMLFileBytes != 10*1024*1024 {
+		t.Fatalf("default=%d, want %d (10MiB)", maxHTMLPublishSingleHTMLFileBytes, 10*1024*1024)
+	}
+}
+
+func TestRunHTMLPublish_RejectsOversizeHTMLFile(t *testing.T) {
+	orig := maxHTMLPublishSingleHTMLFileBytes
+	maxHTMLPublishSingleHTMLFileBytes = 100
+	defer func() { maxHTMLPublishSingleHTMLFileBytes = orig }()
+
+	dir := t.TempDir()
+	if err := os.WriteFile(filepath.Join(dir, "index.html"), []byte("<html></html>"), 0o644); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+	if err := os.WriteFile(filepath.Join(dir, "big.html"), []byte(strings.Repeat("x", 4096)), 0o644); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+	fake := &fakeAppsHTMLPublishClient{}
+	_, err := runHTMLPublish(context.Background(), newTestFIO(), fake, appsHTMLPublishSpec{AppID: "app_x", Path: dir})
+	if err == nil {
+		t.Fatalf("expected per-file oversize error")
+	}
+	problem := requireAppsValidationProblem(t, err)
+	if !strings.Contains(problem.Message, "big.html") || !strings.Contains(problem.Message, "10MB") {
+		t.Fatalf("message=%q, want contains 'big.html' and '10MB'", problem.Message)
+	}
+	if problem.Hint == "" {
+		t.Fatalf("expected non-empty hint")
+	}
+	if len(fake.calls) != 0 {
+		t.Fatalf("client must not be called when an HTML file is oversize")
+	}
+}
+
+func TestRunHTMLPublish_IgnoresOversizeNonHTML(t *testing.T) {
+	// 单 .html 上限调小，但超限文件是 .png → 不被本护栏拦截，正常发布。
+	orig := maxHTMLPublishSingleHTMLFileBytes
+	maxHTMLPublishSingleHTMLFileBytes = 100
+	defer func() { maxHTMLPublishSingleHTMLFileBytes = orig }()
+
+	dir := t.TempDir()
+	if err := os.WriteFile(filepath.Join(dir, "index.html"), []byte("<html></html>"), 0o644); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+	if err := os.WriteFile(filepath.Join(dir, "big.png"), []byte(strings.Repeat("x", 4096)), 0o644); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+	fake := &fakeAppsHTMLPublishClient{resp: &htmlPublishResponse{URL: "https://miaoda/app_x"}}
+	if _, err := runHTMLPublish(context.Background(), newTestFIO(), fake, appsHTMLPublishSpec{AppID: "app_x", Path: dir}); err != nil {
+		t.Fatalf("non-html oversize must not be blocked by the .html cap: %v", err)
+	}
+	if len(fake.calls) != 1 {
+		t.Fatalf("client should be called; calls=%v", fake.calls)
+	}
+}

shortcuts/apps/apps_init.go

@@ -99,7 +99,14 @@ var AppsInit = common.Shortcut{
 			dry.Set("dir_error", err.Error())
 			dir = defaultCloneDir(appID)
 		} else if isAlreadyInitialized(dir) {
-			dry.Set("already_initialized", true)
+			if existing, e := ensureInitDirMatchesApp(dir, appID); e != nil {
+				if existing != "" {
+					dry.Set("app_id_mismatch", existing)
+				}
+				dry.Set("dir_error", e.Error())
+			} else {
+				dry.Set("already_initialized", true)
+			}
 		} else if e := ensureEmptyDir(dir); e != nil {
 			dry.Set("dir_error", e.Error())
 		}
@@ -199,6 +206,61 @@ func isAlreadyInitialized(dir string) bool {
 	return err == nil && !info.IsDir()
 }
 
+// readMetaAppID 读取 <dir>/.spark/meta.json 的 app_id，用于判断目标目录是否同一个妙搭应用。
+// 返回 (appID, isSparkProject, err)：
+//   - meta.json 不存在             → ("", false, nil)   非妙搭工程
+//   - 读取/解析失败（损坏/不可读）  → ("", false, err)   无法确认是否妙搭工程
+//   - 解析成功                     → (trim 后的 app_id, true, nil)（app_id 缺失/为空时为 ""）
+func readMetaAppID(dir string) (string, bool, error) {
+	b, err := os.ReadFile(filepath.Join(dir, metaRelPath)) //nolint:forbidigo // shortcuts cannot import internal/vfs (depguard rule shortcuts-no-vfs); path is under the validated clone dir, and FileIO.Open rejects absolute paths.
+	if os.IsNotExist(err) {
+		return "", false, nil
+	}
+	if err != nil {
+		return "", false, appsFileIOError(err, "read %s failed: %v", metaRelPath, err)
+	}
+	var m struct {
+		AppID string `json:"app_id"`
+	}
+	if err := json.Unmarshal(b, &m); err != nil {
+		return "", false, appsFileIOError(err, "parse %s failed: %v", metaRelPath, err)
+	}
+	return strings.TrimSpace(m.AppID), true, nil
+}
+
+// ensureInitDirMatchesApp 校验「已存在的目标目录」能否被 appID 安全复用：
+//   - 不是妙搭工程（无 meta.json）        → nil（交给 ensureEmptyDir 判空/非空）
+//   - 是妙搭工程且 app_id 与 appID 一致    → nil（走已初始化短路，复用本地代码）
+//   - 是妙搭工程但 app_id 不一致（含为空）  → 报错，提示换目录
+//   - meta.json 损坏/不可读，无法确认      → 报错（fail closed），提示换目录
+//
+// 返回值 existing 是目录里已存在的 app_id（仅"已是另一个 app"的拒绝场景非空），供调用方在
+// dry-run 里回填 app_id_mismatch，避免二次读 meta.json。
+func ensureInitDirMatchesApp(dir, appID string) (existing string, err error) {
+	existing, isSpark, readErr := readMetaAppID(dir)
+	if readErr != nil {
+		return "", appsValidationParamError("--dir",
+			"target directory %q already exists but its %s is unreadable or corrupted; cannot confirm it belongs to app %s, refusing to use it",
+			dir, metaRelPath, appID).
+			WithHint("choose a different --dir, or repair/remove the directory, before running +init").
+			WithCause(readErr)
+	}
+	if !isSpark || existing == appID {
+		return existing, nil
+	}
+	if existing == "" {
+		// meta 存在但缺 app_id：更可能是同一应用上次 +init 中断留下的半成品，而非另一个 app。
+		return "", appsValidationParamError("--dir",
+			"target directory %q has a %s without an app_id; cannot confirm it belongs to app %s, refusing to use it",
+			dir, metaRelPath, appID).
+			WithHint("remove the directory and re-run +init, or choose a different --dir")
+	}
+	return existing, appsValidationParamError("--dir",
+		"target directory %q is already initialized for a different app (%s); refusing to initialize app %s into it",
+		dir, existing, appID).
+		WithHint("choose a different --dir (or cd into the matching project) before running +init")
+}
+
 // ensureMetaAppID patches <dir>/.spark/meta.json to include app_id when the file
 // exists but lacks (or has an empty) app_id. Other fields are preserved. When
 // the file does not exist, this is a no-op (we never create it).
@@ -378,6 +440,11 @@ func appsInitExecute(ctx context.Context, rctx *common.RuntimeContext) error {
 		return err
 	}
 
+	// 异 app 目录护栏：拒绝把当前 app 初始化进另一个 app 的已初始化工程。
+	if _, err := ensureInitDirMatchesApp(dir, appID); err != nil {
+		return err
+	}
+
 	// Already-initialized short-circuit: a dir containing .spark/meta.json is an
 	// initialized app repo -> skip clone/scaffold/commit, but still refresh
 	// the local env so a re-run picks up the latest startup env vars.

shortcuts/apps/apps_init_test.go

@@ -363,7 +363,7 @@ func TestAppsInit_AlreadyInitialized_ShortCircuit(t *testing.T) {
 	if err := os.MkdirAll(filepath.Join(dir, ".spark"), 0o755); err != nil {
 		t.Fatal(err)
 	}
-	if err := os.WriteFile(filepath.Join(dir, metaRelPath), []byte(`{"app_id":"whatever"}`), 0o644); err != nil {
+	if err := os.WriteFile(filepath.Join(dir, metaRelPath), []byte(`{"app_id":"app_x"}`), 0o644); err != nil {
 		t.Fatal(err)
 	}
 	f := &fakeCommandRunner{results: map[string]fakeCallResult{"env-pull": envPullOK(filepath.Join(abs, ".env.local"))}}
@@ -394,6 +394,40 @@ func TestAppsInit_AlreadyInitialized_ShortCircuit(t *testing.T) {
 	}
 }
 
+func TestAppsInit_AlreadyInitialized_AppIDMismatch(t *testing.T) {
+	dir := relCloneDir(t)
+	if err := os.MkdirAll(filepath.Join(dir, ".spark"), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	// 目录是 app_other 的工程，却用 --app-id app_x 初始化 → 必须报错且不拉 env。
+	if err := os.WriteFile(filepath.Join(dir, metaRelPath), []byte(`{"app_id":"app_other"}`), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	f := &fakeCommandRunner{}
+	withFakeRunner(t, f)
+	factory, stdout, _ := newAppsExecuteFactory(t)
+	err := runAppsShortcut(t, AppsInit, []string{"+init", "--app-id", "app_x", "--dir", dir, "--as", "user"}, factory, stdout)
+	if err == nil {
+		t.Fatal("mismatched app_id must error")
+	}
+	problem := requireAppsValidationProblem(t, err)
+	if problem.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype=%q, want %q", problem.Subtype, errs.SubtypeInvalidArgument)
+	}
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) || ve.Param != "--dir" {
+		t.Fatalf("expected *errs.ValidationError with Param=--dir, got %T param=%v", err, ve)
+	}
+	if !strings.Contains(problem.Message, "different app") {
+		t.Fatalf("message=%q, want 'different app'", problem.Message)
+	}
+	for _, c := range f.calls {
+		if containsAll(c, "+env-pull") || containsAll(c, "git", "clone") {
+			t.Errorf("mismatch must not run env-pull/clone; got %v", f.calls)
+		}
+	}
+}
+
 func TestAppsInit_HappyPathCleanTree(t *testing.T) {
 	f := &fakeCommandRunner{results: map[string]fakeCallResult{
 		"credential-init": credInitOK("http://u:t@h/app_x.git"),
@@ -1468,6 +1502,125 @@ func TestAppsInit_Description_IsAboutCode(t *testing.T) {
 	}
 }
 
+func TestReadMetaAppID(t *testing.T) {
+	writeMeta := func(t *testing.T, content string) string {
+		t.Helper()
+		dir := t.TempDir()
+		if err := os.MkdirAll(filepath.Join(dir, ".spark"), 0o755); err != nil {
+			t.Fatal(err)
+		}
+		if err := os.WriteFile(filepath.Join(dir, metaRelPath), []byte(content), 0o644); err != nil {
+			t.Fatal(err)
+		}
+		return dir
+	}
+
+	// 不存在 meta.json → ("", false, nil)
+	if got, ok, err := readMetaAppID(t.TempDir()); ok || got != "" || err != nil {
+		t.Fatalf("no meta: got (%q,%v,%v), want (\"\",false,nil)", got, ok, err)
+	}
+	// 存在且有 app_id → (app_id, true, nil)
+	if got, ok, err := readMetaAppID(writeMeta(t, `{"app_id":"app_a"}`)); !ok || got != "app_a" || err != nil {
+		t.Fatalf("with app_id: got (%q,%v,%v), want (\"app_a\",true,nil)", got, ok, err)
+	}
+	// 存在但 app_id 空 → ("", true, nil)
+	if got, ok, err := readMetaAppID(writeMeta(t, `{"name":"x"}`)); !ok || got != "" || err != nil {
+		t.Fatalf("empty app_id: got (%q,%v,%v), want (\"\",true,nil)", got, ok, err)
+	}
+	// 存在但坏 JSON → ("", false, err)（无法确认）
+	if got, ok, err := readMetaAppID(writeMeta(t, `{not json`)); ok || got != "" || err == nil {
+		t.Fatalf("bad json: got (%q,%v,err=%v), want (\"\",false,non-nil)", got, ok, err)
+	}
+}
+
+func TestEnsureInitDirMatchesApp(t *testing.T) {
+	writeMeta := func(t *testing.T, content string) string {
+		t.Helper()
+		dir := t.TempDir()
+		if err := os.MkdirAll(filepath.Join(dir, ".spark"), 0o755); err != nil {
+			t.Fatal(err)
+		}
+		if err := os.WriteFile(filepath.Join(dir, metaRelPath), []byte(content), 0o644); err != nil {
+			t.Fatal(err)
+		}
+		return dir
+	}
+
+	// 无 meta（非妙搭工程）→ nil（交给 ensureEmptyDir）
+	if _, err := ensureInitDirMatchesApp(t.TempDir(), "app_x"); err != nil {
+		t.Fatalf("no meta should pass: %v", err)
+	}
+	// 同 app_id → (app_id, nil)（走已初始化短路）
+	if existing, err := ensureInitDirMatchesApp(writeMeta(t, `{"app_id":"app_x"}`), "app_x"); err != nil || existing != "app_x" {
+		t.Fatalf("same app should pass: existing=%q err=%v", existing, err)
+	}
+
+	// 不同 app_id → error（换目录），返回 existing=app_other；断言 typed metadata（subtype/param）
+	existing, errMismatch := ensureInitDirMatchesApp(writeMeta(t, `{"app_id":"app_other"}`), "app_x")
+	if errMismatch == nil {
+		t.Fatal("different app should error")
+	}
+	if existing != "app_other" {
+		t.Fatalf("mismatch should return existing app_id, got %q", existing)
+	}
+	problem := requireAppsValidationProblem(t, errMismatch) // 已校验 Category==Validation
+	if problem.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("subtype=%q, want %q", problem.Subtype, errs.SubtypeInvalidArgument)
+	}
+	var ve *errs.ValidationError
+	if !errors.As(errMismatch, &ve) {
+		t.Fatalf("expected *errs.ValidationError, got %T", errMismatch)
+	}
+	if ve.Param != "--dir" {
+		t.Fatalf("param=%q, want --dir", ve.Param)
+	}
+	if !strings.Contains(problem.Message, "different app") || !strings.Contains(problem.Message, "app_other") {
+		t.Fatalf("message=%q, want 'different app' and 'app_other'", problem.Message)
+	}
+	if !strings.Contains(problem.Hint, "different --dir") {
+		t.Fatalf("hint=%q, want 'different --dir'", problem.Hint)
+	}
+
+	// 空 app_id（缺 app_id 标记的半成品）→ error，独立文案（非 "different app"），返回 existing=""
+	emptyExisting, errEmpty := ensureInitDirMatchesApp(writeMeta(t, `{"name":"x"}`), "app_x")
+	if errEmpty == nil {
+		t.Fatal("empty meta app_id should error (cannot confirm same app)")
+	}
+	if emptyExisting != "" {
+		t.Fatalf("empty app_id should return existing=\"\", got %q", emptyExisting)
+	}
+	pEmpty := requireAppsValidationProblem(t, errEmpty)
+	if pEmpty.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("empty subtype=%q, want %q", pEmpty.Subtype, errs.SubtypeInvalidArgument)
+	}
+	if !strings.Contains(pEmpty.Message, "without an app_id") {
+		t.Fatalf("empty app_id should have its own message, msg=%q", pEmpty.Message)
+	}
+	if strings.Contains(pEmpty.Message, "different app") {
+		t.Fatalf("empty app_id must not reuse the different-app wording, msg=%q", pEmpty.Message)
+	}
+
+	// meta 损坏/不可读 → error（fail closed），返回 existing=""
+	badExisting, errBad := ensureInitDirMatchesApp(writeMeta(t, `{not json`), "app_x")
+	if errBad == nil {
+		t.Fatal("corrupted meta should fail closed")
+	}
+	if badExisting != "" {
+		t.Fatalf("corrupted should return existing=\"\", got %q", badExisting)
+	}
+	pBad := requireAppsValidationProblem(t, errBad)
+	if pBad.Subtype != errs.SubtypeInvalidArgument {
+		t.Fatalf("corrupted subtype=%q, want %q", pBad.Subtype, errs.SubtypeInvalidArgument)
+	}
+	if !strings.Contains(pBad.Message, "unreadable or corrupted") {
+		t.Fatalf("corrupted meta msg=%q, want 'unreadable or corrupted'", pBad.Message)
+	}
+	var veBad *errs.ValidationError
+	if !errors.As(errBad, &veBad) || veBad.Param != "--dir" {
+		t.Fatalf("corrupted: expected ValidationError Param=--dir, got %T param=%v", errBad, veBad)
+	}
+}
+
 // TestRunScaffold_SubprocessFailureIsExternalTool pins the typed
 // classification of an external-tool failure: a failing git subprocess
 // surfaces as internal/external_tool with the cause preserved.

shortcuts/common/drive_media_upload_typed_test.go

@@ -199,16 +199,7 @@ func TestParseDriveMediaMultipartUploadSessionTypedValidatesResponseFields(t *te
 			t.Parallel()
 
 			_, err := parseDriveMediaMultipartUploadSessionTyped(tt.data)
-			if err == nil || !strings.Contains(err.Error(), tt.wantText) {
-				t.Fatalf("err = %v, want substring %q", err, tt.wantText)
-			}
-			p, ok := errs.ProblemOf(err)
-			if !ok {
-				t.Fatalf("expected typed problem, got %T (%v)", err, err)
-			}
-			if p.Subtype != errs.SubtypeInvalidResponse {
-				t.Fatalf("subtype = %s, want invalid_response", p.Subtype)
-			}
+			requireProblem(t, err, errs.CategoryInternal, errs.SubtypeInvalidResponse, tt.wantText)
 		})
 	}
 }

shortcuts/common/mcp_client_test.go

@@ -142,9 +142,7 @@ func TestNormalizeMCPToolResult(t *testing.T) {
 
 			got, err := normalizeMCPToolResult(tt.raw)
 			if tt.wantErr != "" {
-				if err == nil || !strings.Contains(err.Error(), tt.wantErr) {
-					t.Fatalf("expected error containing %q, got %v", tt.wantErr, err)
-				}
+				requireProblem(t, err, errs.CategoryAPI, errs.SubtypeUnknown, tt.wantErr)
 				return
 			}
 			if err != nil {

shortcuts/common/runner.go

@@ -49,8 +49,21 @@ type RuntimeContext struct {
 	apiClientFunc func() (*client.APIClient, error) // sync.OnceValues; initialized in newRuntimeContext
 	botInfoFunc   func() (*BotInfo, error)          // sync.OnceValues; lazy bot identity from /bot/v3/info
 	larkSDK       *lark.Client                      // eagerly initialized in mountDeclarative
+	stdinConsumed bool                              // set when any flag has consumed stdin (`-`); used so out-of-band binary readers (e.g. sheets +table-put --dataframe) can refuse a second stdin consumer instead of racing for an already-empty stream
 }
 
+// StdinConsumed reports whether stdin has already been consumed by an Input
+// flag's `-` form via resolveInputFlags. Out-of-band binary readers that read
+// stdin themselves (currently sheets +table-put / +workbook-create --dataframe)
+// must check this before reading — a process has a single stdin, so two
+// consumers would race and one would see an empty stream.
+func (ctx *RuntimeContext) StdinConsumed() bool { return ctx.stdinConsumed }
+
+// MarkStdinConsumed marks stdin as consumed. Out-of-band binary readers must
+// call this after they read stdin so a later Input-flag `-` is rejected cleanly
+// instead of racing on an empty stream.
+func (ctx *RuntimeContext) MarkStdinConsumed() { ctx.stdinConsumed = true }
+
 // ── Identity ──
 
 // As returns the current identity.
@@ -1029,7 +1042,6 @@ func stripUTF8BOM(s string) string {
 // resolveInputFlags resolves @file and - (stdin) for flags with Input sources.
 // Must be called before Validate/DryRun/Execute so that runtime.Str() returns resolved content.
 func resolveInputFlags(rctx *RuntimeContext, flags []Flag) error {
-	stdinUsed := false
 	for _, fl := range flags {
 		if len(fl.Input) == 0 {
 			continue
@@ -1049,11 +1061,15 @@ func resolveInputFlags(rctx *RuntimeContext, flags []Flag) error {
 				return ValidationErrorf("--%s does not support stdin (-)", fl.Name).
 					WithParam("--" + fl.Name)
 			}
-			if stdinUsed {
+			// stdinConsumed also covers out-of-band readers like sheets +table-put
+			// --dataframe (binary, doesn't go through Input). A process has a
+			// single stdin, so we reject a second consumer regardless of source.
+			if rctx.stdinConsumed {
 				return ValidationErrorf("--%s: stdin (-) can only be used by one flag", fl.Name).
-					WithParam("--" + fl.Name)
+					WithParam("--"+fl.Name).
+					WithHint("a process has a single stdin, so only one flag per call may use '-'; pass the others as @file (e.g. --%s @/path/to/file)", fl.Name)
 			}
-			stdinUsed = true
+			rctx.stdinConsumed = true
 			data, err := io.ReadAll(rctx.IO().In)
 			if err != nil {
 				return ValidationErrorf("--%s: failed to read from stdin: %v", fl.Name, err).
@@ -1166,7 +1182,13 @@ func registerShortcutFlagsWithContext(ctx context.Context, cmd *cobra.Command, f
 				hints = append(hints, "@file")
 			}
 			if slices.Contains(fl.Input, Stdin) {
-				hints = append(hints, "- for stdin")
+				// "- reads stdin" intentionally avoids implying each flag has
+				// its own stdin: a process has a single stdin, so at most one
+				// flag per call may use "-" (the rest must use @file). The old
+				// per-flag "- for stdin" wording led AI agents to write
+				// `--a - <x --b - <y`, where the second `<` silently clobbers
+				// the first and `--a` reads the wrong payload.
+				hints = append(hints, "- reads stdin (one flag per call; use @file for others)")
 			}
 			desc += " (supports " + strings.Join(hints, ", ") + ")"
 		}

shortcuts/common/runner_args_test.go

@@ -22,6 +22,7 @@ func TestRejectPositionalArgs_WithArgs(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for positional arg, got nil")
 	}
+	// rejectPositionalArgs returns a raw fmt.Errorf via cobra's PositionalArgs contract — not a typed envelope, message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "positional arguments are not supported") {
 		t.Errorf("expected positional args rejection message, got: %v", err)
 	}
@@ -39,6 +40,7 @@ func TestRejectPositionalArgs_MultipleArgs(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for multiple positional args, got nil")
 	}
+	// rejectPositionalArgs returns a raw fmt.Errorf via cobra's PositionalArgs contract — not a typed envelope, message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "positional arguments are not supported") {
 		t.Errorf("unexpected error message: %v", err)
 	}

shortcuts/common/runner_botinfo_test.go

@@ -171,6 +171,7 @@ func TestFetchBotInfo_APICodeNonZero(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for non-zero code")
 	}
+	// fetchBotInfo returns a raw fmt.Errorf, not a typed envelope — message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "[99991]") {
 		t.Errorf("error = %q, want substring [99991]", err.Error())
 	}
@@ -197,6 +198,7 @@ func TestFetchBotInfo_EmptyOpenID(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for empty open_id")
 	}
+	// fetchBotInfo returns a raw fmt.Errorf, not a typed envelope — message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "open_id is empty") {
 		t.Errorf("error = %q, want substring 'open_id is empty'", err.Error())
 	}
@@ -218,6 +220,7 @@ func TestFetchBotInfo_HTTP4xx(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for HTTP 403")
 	}
+	// fetchBotInfo returns a raw fmt.Errorf, not a typed envelope — message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "403") {
 		t.Errorf("error = %q, want substring '403'", err.Error())
 	}
@@ -238,7 +241,7 @@ func TestFetchBotInfo_InvalidJSON(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for invalid JSON")
 	}
-	// Error may come from SDK-level parse or our unmarshal wrapper
+	// Error may come from SDK-level parse or our unmarshal wrapper — both are raw fmt.Errorf, not a typed envelope.
 	if !strings.Contains(err.Error(), "unmarshal") && !strings.Contains(err.Error(), "invalid character") {
 		t.Errorf("error = %q, want JSON parse failure", err.Error())
 	}
@@ -279,6 +282,7 @@ func TestFetchBotInfo_CanBotFalse(t *testing.T) {
 	if info != nil {
 		t.Errorf("expected nil info, got %+v", info)
 	}
+	// fetchBotInfo returns a raw fmt.Errorf, not a typed envelope — message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "not available") {
 		t.Errorf("error = %q, want substring 'not available'", err.Error())
 	}
@@ -291,6 +295,7 @@ func TestBotInfo_NilFunc(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for nil botInfoFunc")
 	}
+	// BotInfo() returns a raw fmt.Errorf when botInfoFunc is nil, not a typed envelope — message-substring assertion is intentional.
 	if !strings.Contains(err.Error(), "not fully initialized") {
 		t.Errorf("unexpected error: %v", err)
 	}

shortcuts/common/runner_input_test.go

@@ -129,9 +129,9 @@ func TestResolveInputFlags_StdinNotSupported(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for stdin not supported")
 	}
-	assertValidationParam(t, err, "--data")
-	if !strings.Contains(err.Error(), "does not support stdin") {
-		t.Errorf("unexpected error: %v", err)
+	vErr := assertValidationParam(t, err, "--data")
+	if !strings.Contains(vErr.Message, "does not support stdin") {
+		t.Errorf("unexpected error message: %q", vErr.Message)
 	}
 }
 
@@ -143,9 +143,9 @@ func TestResolveInputFlags_FileNotSupported(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for file not supported")
 	}
-	assertValidationParam(t, err, "--data")
-	if !strings.Contains(err.Error(), "does not support file input") {
-		t.Errorf("unexpected error: %v", err)
+	vErr := assertValidationParam(t, err, "--data")
+	if !strings.Contains(vErr.Message, "does not support file input") {
+		t.Errorf("unexpected error message: %q", vErr.Message)
 	}
 }
 
@@ -160,9 +160,9 @@ func TestResolveInputFlags_FileNotFound(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for missing file")
 	}
-	assertValidationParam(t, err, "--markdown")
-	if !strings.Contains(err.Error(), "cannot read file") {
-		t.Errorf("unexpected error: %v", err)
+	vErr := assertValidationParam(t, err, "--markdown")
+	if !strings.Contains(vErr.Message, "cannot read file") {
+		t.Errorf("unexpected error message: %q", vErr.Message)
 	}
 }
 
@@ -174,9 +174,9 @@ func TestResolveInputFlags_EmptyFilePath(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for empty file path")
 	}
-	assertValidationParam(t, err, "--markdown")
-	if !strings.Contains(err.Error(), "file path cannot be empty after @") {
-		t.Errorf("unexpected error: %v", err)
+	vErr := assertValidationParam(t, err, "--markdown")
+	if !strings.Contains(vErr.Message, "file path cannot be empty after @") {
+		t.Errorf("unexpected error message: %q", vErr.Message)
 	}
 }
 
@@ -216,9 +216,14 @@ func TestResolveInputFlags_DuplicateStdin(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for duplicate stdin usage")
 	}
-	assertValidationParam(t, err, "--b")
-	if !strings.Contains(err.Error(), "stdin (-) can only be used by one flag") {
-		t.Errorf("unexpected error: %v", err)
+	vErr := assertValidationParam(t, err, "--b")
+	if !strings.Contains(vErr.Message, "stdin (-) can only be used by one flag") {
+		t.Errorf("unexpected error message: %q", vErr.Message)
+	}
+	// The hint must steer an AI agent to the fix (@file for the extra flags),
+	// since `--a - <x --b - <y` is the exact misuse this guards against.
+	if !strings.Contains(vErr.Hint, "@file") {
+		t.Errorf("hint %q should mention @file as the fix", vErr.Hint)
 	}
 }
 

shortcuts/common/runner_jq_test.go

@@ -186,9 +186,7 @@ func TestRunShortcut_JqAndFormatConflict(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for --jq + --format table conflict")
 	}
-	if !strings.Contains(err.Error(), "mutually exclusive") {
-		t.Errorf("expected 'mutually exclusive' error, got: %v", err)
-	}
+	requireValidation(t, err, "mutually exclusive")
 }
 
 func TestRunShortcut_JqInvalidExpression(t *testing.T) {
@@ -208,9 +206,7 @@ func TestRunShortcut_JqInvalidExpression(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error for invalid jq expression")
 	}
-	if !strings.Contains(err.Error(), "invalid jq expression") {
-		t.Errorf("expected 'invalid jq expression' error, got: %v", err)
-	}
+	requireValidation(t, err, "invalid jq expression")
 }
 
 func TestRunShortcut_JqRuntimeError_PropagatesError(t *testing.T) {

shortcuts/common/typed_error_assertions_test.go

@@ -0,0 +1,50 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package common
+
+import (
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+)
+
+// requireProblem asserts err carries a typed errs.Problem with the given
+// category and (optional) subtype, and that its message contains msgContains
+// (skip the message check by passing ""). Returns the Problem so callers can
+// drill into the typed envelope's category-specific fields (e.g. cast to
+// *errs.ValidationError to read .Param / .Params / .Cause).
+func requireProblem(t *testing.T, err error, wantCategory errs.Category, wantSubtype errs.Subtype, msgContains string) *errs.Problem {
+	t.Helper()
+	if err == nil {
+		t.Fatal("expected error, got nil")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed error carrying errs.Problem, got %T: %v", err, err)
+	}
+	if p.Category != wantCategory {
+		t.Errorf("category = %q, want %q (err=%v)", p.Category, wantCategory, err)
+	}
+	if wantSubtype != "" && p.Subtype != wantSubtype {
+		t.Errorf("subtype = %q, want %q (err=%v)", p.Subtype, wantSubtype, err)
+	}
+	if msgContains != "" && !strings.Contains(p.Message, msgContains) {
+		t.Errorf("message = %q, want containing %q", p.Message, msgContains)
+	}
+	return p
+}
+
+// requireValidation is shorthand for CategoryValidation + SubtypeInvalidArgument.
+// Returns *errs.ValidationError so callers can also assert on .Param / .Params / .Cause.
+func requireValidation(t *testing.T, err error, msgContains string) *errs.ValidationError {
+	t.Helper()
+	requireProblem(t, err, errs.CategoryValidation, errs.SubtypeInvalidArgument, msgContains)
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) {
+		t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
+	}
+	return ve
+}

shortcuts/drive/drive_export.go

@@ -5,6 +5,7 @@ package drive
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"path/filepath"
 	"strings"
@@ -15,6 +16,24 @@ import (
 	"github.com/larksuite/cli/shortcuts/common"
 )
 
+// wrapExportContextErr converts a context cancellation / deadline error into a
+// typed errs.NetworkError so the cobra layer sees a typed envelope (with cause
+// preserved for errors.Is) instead of an untyped context.Canceled /
+// context.DeadlineExceeded escaping as a plain string. CR-flagged hole on the
+// poll loop: returning ctx.Err() directly bypassed the typed-error contract.
+func wrapExportContextErr(err error) error {
+	if err == nil {
+		return nil
+	}
+	subtype := errs.SubtypeNetworkTransport
+	msg := "drive +export polling cancelled: %s"
+	if errors.Is(err, context.DeadlineExceeded) {
+		subtype = errs.SubtypeNetworkTimeout
+		msg = "drive +export polling deadline exceeded: %s"
+	}
+	return errs.NewNetworkError(subtype, msg, err).WithCause(err)
+}
+
 // DriveExport exports Drive-native documents to local files and falls back to
 // a follow-up command when the async export task does not finish in time.
 var DriveExport = common.Shortcut{
@@ -40,236 +59,302 @@ var DriveExport = common.Shortcut{
 		{Name: "overwrite", Type: "bool", Desc: "overwrite existing output file"},
 	},
 	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		return validateDriveExportSpec(driveExportSpec{
-			Token:         runtime.Str("token"),
-			DocType:       runtime.Str("doc-type"),
-			FileExtension: runtime.Str("file-extension"),
-			SubID:         runtime.Str("sub-id"),
-			OnlySchema:    runtime.Bool("only-schema"),
-		})
+		return ValidateExport(exportParamsFromFlags(runtime))
 	},
 	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		spec := driveExportSpec{
-			Token:         runtime.Str("token"),
-			DocType:       runtime.Str("doc-type"),
-			FileExtension: runtime.Str("file-extension"),
-			SubID:         runtime.Str("sub-id"),
-			OnlySchema:    runtime.Bool("only-schema"),
-		}
-		// Markdown export is a special case: docx markdown comes from the V2
-		// docs_ai fetch API directly instead of the Drive export task API.
-		if spec.FileExtension == "markdown" {
-			apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
-			dr := common.NewDryRunAPI().
-				Desc("2-step orchestration: fetch docx markdown -> write local file").
-				POST(apiPath).
-				Body(map[string]interface{}{
-					"format": "markdown",
-				}).
-				Set("output_dir", runtime.Str("output-dir"))
-			if name := strings.TrimSpace(runtime.Str("file-name")); name != "" {
-				dr.Set("file_name", ensureExportFileExtension(sanitizeExportFileName(name, spec.Token), spec.FileExtension))
-			}
-			return dr
-		}
+		return PlanExportDryRun(runtime, exportParamsFromFlags(runtime))
+	},
+	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		return RunExport(ctx, runtime, exportParamsFromFlags(runtime))
+	},
+}
 
-		body := map[string]interface{}{
-			"token":          spec.Token,
-			"type":           spec.DocType,
-			"file_extension": spec.FileExtension,
-		}
-		if strings.TrimSpace(spec.SubID) != "" {
-			body["sub_id"] = spec.SubID
-		}
-		if spec.OnlySchema {
-			body["only_schema"] = true
-		}
+// ExportParams holds the user-facing inputs for an export flow, decoupled from
+// cobra flags so other command groups (e.g. sheets +workbook-export) can reuse
+// the drive export implementation. An empty OutputDir means "create the export
+// task and poll, but do not download" — callers that only need the ready file
+// token / status get it back without writing a local file.
+type ExportParams struct {
+	Token         string
+	DocType       string
+	FileExtension string
+	SubID         string
+	OnlySchema    bool
+	OutputDir     string
+	FileName      string
+	Overwrite     bool
+}
 
+func (p ExportParams) spec() driveExportSpec {
+	return driveExportSpec{
+		Token:         p.Token,
+		DocType:       p.DocType,
+		FileExtension: p.FileExtension,
+		SubID:         p.SubID,
+		OnlySchema:    p.OnlySchema,
+	}
+}
+
+// exportParamsFromFlags reads the standard drive +export flag set.
+func exportParamsFromFlags(runtime *common.RuntimeContext) ExportParams {
+	// drive +export always downloads; an empty --output-dir historically means
+	// the current directory (saveContentToOutputDir maps "" -> "."), so normalize
+	// it here to keep behavior identical and stay off the export-only ("" => skip
+	// download) path that only sheets +workbook-export uses.
+	outputDir := runtime.Str("output-dir")
+	if outputDir == "" {
+		outputDir = "."
+	}
+	return ExportParams{
+		Token:         runtime.Str("token"),
+		DocType:       runtime.Str("doc-type"),
+		FileExtension: runtime.Str("file-extension"),
+		SubID:         runtime.Str("sub-id"),
+		OnlySchema:    runtime.Bool("only-schema"),
+		OutputDir:     outputDir,
+		FileName:      strings.TrimSpace(runtime.Str("file-name")),
+		Overwrite:     runtime.Bool("overwrite"),
+	}
+}
+
+// ValidateExport runs the CLI-level export constraint checks.
+func ValidateExport(p ExportParams) error {
+	return validateDriveExportSpec(p.spec())
+}
+
+// PlanExportDryRun builds the dry-run plan for an export without performing I/O.
+func PlanExportDryRun(runtime *common.RuntimeContext, p ExportParams) *common.DryRunAPI {
+	spec := p.spec()
+	// Markdown export is a special case: docx markdown comes from the V2
+	// docs_ai fetch API directly instead of the Drive export task API.
+	if spec.FileExtension == "markdown" {
+		apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
 		dr := common.NewDryRunAPI().
-			Desc("3-step orchestration: create export task -> limited polling -> download file").
-			POST("/open-apis/drive/v1/export_tasks").
-			Body(body).
-			Set("output_dir", runtime.Str("output-dir"))
-		if name := strings.TrimSpace(runtime.Str("file-name")); name != "" {
+			Desc("2-step orchestration: fetch docx markdown -> write local file").
+			POST(apiPath).
+			Body(map[string]interface{}{
+				"format": "markdown",
+			}).
+			Set("output_dir", p.OutputDir)
+		if name := strings.TrimSpace(p.FileName); name != "" {
 			dr.Set("file_name", ensureExportFileExtension(sanitizeExportFileName(name, spec.Token), spec.FileExtension))
 		}
 		return dr
-	},
-	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		spec := driveExportSpec{
-			Token:         runtime.Str("token"),
-			DocType:       runtime.Str("doc-type"),
-			FileExtension: runtime.Str("file-extension"),
-			SubID:         runtime.Str("sub-id"),
-			OnlySchema:    runtime.Bool("only-schema"),
+	}
+
+	body := map[string]interface{}{
+		"token":          spec.Token,
+		"type":           spec.DocType,
+		"file_extension": spec.FileExtension,
+	}
+	if strings.TrimSpace(spec.SubID) != "" {
+		body["sub_id"] = spec.SubID
+	}
+	if spec.OnlySchema {
+		body["only_schema"] = true
+	}
+
+	dr := common.NewDryRunAPI().
+		Desc("3-step orchestration: create export task -> limited polling -> download file").
+		POST("/open-apis/drive/v1/export_tasks").
+		Body(body).
+		Set("output_dir", p.OutputDir)
+	if name := strings.TrimSpace(p.FileName); name != "" {
+		dr.Set("file_name", ensureExportFileExtension(sanitizeExportFileName(name, spec.Token), spec.FileExtension))
+	}
+	return dr
+}
+
+// RunExport drives create export task -> bounded poll -> optional download. It
+// is the shared core behind both drive +export and sheets +workbook-export. An
+// empty p.OutputDir skips the download step and returns the ready file token.
+func RunExport(ctx context.Context, runtime *common.RuntimeContext, p ExportParams) error {
+	spec := p.spec()
+	outputDir := p.OutputDir
+	preferredFileName := strings.TrimSpace(p.FileName)
+	overwrite := p.Overwrite
+
+	// Markdown export bypasses the async export task and writes the fetched
+	// markdown content directly to disk. Uses the V2 docs_ai fetch API for
+	// higher-quality Lark-flavored Markdown output.
+	if spec.FileExtension == "markdown" {
+		fmt.Fprintf(runtime.IO().ErrOut, "Exporting docx as markdown: %s\n", common.MaskToken(spec.Token))
+		apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
+		data, err := runtime.CallAPITyped(
+			"POST",
+			apiPath,
+			nil,
+			map[string]interface{}{
+				"format": "markdown",
+			},
+		)
+		if err != nil {
+			return err
 		}
-		outputDir := runtime.Str("output-dir")
-		preferredFileName := strings.TrimSpace(runtime.Str("file-name"))
-		overwrite := runtime.Bool("overwrite")
 
-		// Markdown export bypasses the async export task and writes the fetched
-		// markdown content directly to disk. Uses the V2 docs_ai fetch API for
-		// higher-quality Lark-flavored Markdown output.
-		if spec.FileExtension == "markdown" {
-			fmt.Fprintf(runtime.IO().ErrOut, "Exporting docx as markdown: %s\n", common.MaskToken(spec.Token))
-			apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
-			data, err := runtime.CallAPITyped(
-				"POST",
-				apiPath,
-				nil,
-				map[string]interface{}{
-					"format": "markdown",
-				},
-			)
+		// Extract content from the V2 response: data.document.content
+		doc, ok := data["document"].(map[string]interface{})
+		if !ok {
+			return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document object")
+		}
+		content, ok := doc["content"].(string)
+		if !ok {
+			return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document.content")
+		}
+
+		fileName := preferredFileName
+		if fileName == "" {
+			// Prefer the remote title for the exported file name, but still fall
+			// back to the token if metadata is empty.
+			title, err := common.FetchDriveMetaTitle(runtime, spec.Token, spec.DocType)
 			if err != nil {
-				return err
+				fmt.Fprintf(runtime.IO().ErrOut, "Title lookup failed, using token as filename: %v\n", err)
+				title = spec.Token
 			}
+			fileName = title
+		}
+		fileName = ensureExportFileExtension(sanitizeExportFileName(fileName, spec.Token), spec.FileExtension)
+		savedPath, err := saveContentToOutputDir(runtime.FileIO(), outputDir, fileName, []byte(content), overwrite)
+		if err != nil {
+			return err
+		}
 
-			// Extract content from the V2 response: data.document.content
-			doc, ok := data["document"].(map[string]interface{})
-			if !ok {
-				return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document object")
+		runtime.Out(map[string]interface{}{
+			"token":          spec.Token,
+			"doc_type":       spec.DocType,
+			"file_extension": spec.FileExtension,
+			"file_name":      filepath.Base(savedPath),
+			"saved_path":     savedPath,
+			"size_bytes":     len(content),
+		}, nil)
+		return nil
+	}
+
+	ticket, err := createDriveExportTask(runtime, spec)
+	if err != nil {
+		return err
+	}
+	fmt.Fprintf(runtime.IO().ErrOut, "Created export task: %s\n", ticket)
+
+	var lastStatus driveExportStatus
+	var lastPollErr error
+	hasObservedStatus := false
+	// Keep the command responsive by polling for a bounded window. If the task
+	// is still running after that, return a resume command instead of blocking.
+	for attempt := 1; attempt <= driveExportPollAttempts; attempt++ {
+		if attempt > 1 {
+			select {
+			case <-ctx.Done():
+				return wrapExportContextErr(ctx.Err())
+			case <-time.After(driveExportPollInterval):
 			}
-			content, ok := doc["content"].(string)
-			if !ok {
-				return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document.content")
+		}
+		if err := ctx.Err(); err != nil {
+			return wrapExportContextErr(err)
+		}
+
+		status, err := getDriveExportStatus(runtime, spec.Token, ticket)
+		if err != nil {
+			// Treat polling failures as transient so short-lived backend hiccups
+			// do not immediately fail an otherwise healthy export task.
+			lastPollErr = err
+			fmt.Fprintf(runtime.IO().ErrOut, "Export status attempt %d/%d failed: %v\n", attempt, driveExportPollAttempts, err)
+			continue
+		}
+		lastStatus = status
+		hasObservedStatus = true
+
+		if status.Ready() {
+			fmt.Fprintf(runtime.IO().ErrOut, "Export task completed: %s\n", common.MaskToken(status.FileToken))
+
+			// Export-only mode: caller wants the ready file token / metadata but
+			// no local download (e.g. sheets +workbook-export without an output
+			// path). Skip the download and return the status envelope.
+			if strings.TrimSpace(outputDir) == "" {
+				runtime.Out(map[string]interface{}{
+					"ticket":         ticket,
+					"token":          spec.Token,
+					"doc_type":       spec.DocType,
+					"file_extension": spec.FileExtension,
+					"file_token":     status.FileToken,
+					"file_name":      status.FileName,
+					"file_size":      status.FileSize,
+					"ready":          true,
+					"downloaded":     false,
+				}, nil)
+				return nil
 			}
 
 			fileName := preferredFileName
 			if fileName == "" {
-				// Prefer the remote title for the exported file name, but still fall
-				// back to the token if metadata is empty.
-				title, err := common.FetchDriveMetaTitle(runtime, spec.Token, spec.DocType)
-				if err != nil {
-					fmt.Fprintf(runtime.IO().ErrOut, "Title lookup failed, using token as filename: %v\n", err)
-					title = spec.Token
-				}
-				fileName = title
+				fileName = status.FileName
 			}
 			fileName = ensureExportFileExtension(sanitizeExportFileName(fileName, spec.Token), spec.FileExtension)
-			savedPath, err := saveContentToOutputDir(runtime.FileIO(), outputDir, fileName, []byte(content), overwrite)
+			out, err := downloadDriveExportFile(ctx, runtime, status.FileToken, outputDir, fileName, overwrite)
 			if err != nil {
-				return err
+				recoveryCommand := driveExportDownloadCommand(status.FileToken, fileName, outputDir, overwrite)
+				hint := fmt.Sprintf(
+					"the export artifact is already ready (ticket=%s, file_token=%s)\nretry download with: %s",
+					ticket,
+					status.FileToken,
+					recoveryCommand,
+				)
+				return appendDriveExportRecoveryHint(err, hint)
 			}
-
-			runtime.Out(map[string]interface{}{
-				"token":          spec.Token,
-				"doc_type":       spec.DocType,
-				"file_extension": spec.FileExtension,
-				"file_name":      filepath.Base(savedPath),
-				"saved_path":     savedPath,
-				"size_bytes":     len(content),
-			}, nil)
+			out["ticket"] = ticket
+			out["doc_type"] = spec.DocType
+			out["file_extension"] = spec.FileExtension
+			runtime.Out(out, nil)
 			return nil
 		}
 
-		ticket, err := createDriveExportTask(runtime, spec)
-		if err != nil {
-			return err
-		}
-		fmt.Fprintf(runtime.IO().ErrOut, "Created export task: %s\n", ticket)
-
-		var lastStatus driveExportStatus
-		var lastPollErr error
-		hasObservedStatus := false
-		// Keep the command responsive by polling for a bounded window. If the task
-		// is still running after that, return a resume command instead of blocking.
-		for attempt := 1; attempt <= driveExportPollAttempts; attempt++ {
-			if attempt > 1 {
-				select {
-				case <-ctx.Done():
-					return ctx.Err()
-				case <-time.After(driveExportPollInterval):
-				}
+		if status.Failed() {
+			msg := strings.TrimSpace(status.JobErrorMsg)
+			if msg == "" {
+				msg = status.StatusLabel()
 			}
-			if err := ctx.Err(); err != nil {
-				return err
-			}
-
-			status, err := getDriveExportStatus(runtime, spec.Token, ticket)
-			if err != nil {
-				// Treat polling failures as transient so short-lived backend hiccups
-				// do not immediately fail an otherwise healthy export task.
-				lastPollErr = err
-				fmt.Fprintf(runtime.IO().ErrOut, "Export status attempt %d/%d failed: %v\n", attempt, driveExportPollAttempts, err)
-				continue
-			}
-			lastStatus = status
-			hasObservedStatus = true
-
-			if status.Ready() {
-				fmt.Fprintf(runtime.IO().ErrOut, "Export task completed: %s\n", common.MaskToken(status.FileToken))
-				fileName := preferredFileName
-				if fileName == "" {
-					fileName = status.FileName
-				}
-				fileName = ensureExportFileExtension(sanitizeExportFileName(fileName, spec.Token), spec.FileExtension)
-				out, err := downloadDriveExportFile(ctx, runtime, status.FileToken, outputDir, fileName, overwrite)
-				if err != nil {
-					recoveryCommand := driveExportDownloadCommand(status.FileToken, fileName, outputDir, overwrite)
-					hint := fmt.Sprintf(
-						"the export artifact is already ready (ticket=%s, file_token=%s)\nretry download with: %s",
-						ticket,
-						status.FileToken,
-						recoveryCommand,
-					)
-					return appendDriveExportRecoveryHint(err, hint)
-				}
-				out["ticket"] = ticket
-				out["doc_type"] = spec.DocType
-				out["file_extension"] = spec.FileExtension
-				runtime.Out(out, nil)
-				return nil
-			}
-
-			if status.Failed() {
-				msg := strings.TrimSpace(status.JobErrorMsg)
-				if msg == "" {
-					msg = status.StatusLabel()
-				}
-				return errs.NewAPIError(errs.SubtypeServerError, "export task failed: %s (ticket=%s)", msg, ticket)
-			}
-
-			fmt.Fprintf(runtime.IO().ErrOut, "Export status %d/%d: %s\n", attempt, driveExportPollAttempts, status.StatusLabel())
+			return errs.NewAPIError(errs.SubtypeServerError, "export task failed: %s (ticket=%s)", msg, ticket)
 		}
 
-		nextCommand := driveExportTaskResultCommand(ticket, spec.Token)
-		if !hasObservedStatus && lastPollErr != nil {
-			hint := fmt.Sprintf(
-				"the export task was created but every status poll failed (ticket=%s)\nretry status lookup with: %s",
-				ticket,
-				nextCommand,
-			)
-			return appendDriveExportRecoveryHint(lastPollErr, hint)
-		}
+		fmt.Fprintf(runtime.IO().ErrOut, "Export status %d/%d: %s\n", attempt, driveExportPollAttempts, status.StatusLabel())
+	}
 
-		failed := false
-		var jobStatus interface{}
-		jobStatusLabel := "unknown"
-		if hasObservedStatus {
-			failed = lastStatus.Failed()
-			jobStatus = lastStatus.JobStatus
-			jobStatusLabel = lastStatus.StatusLabel()
-		}
-		// Return the last observed status so callers can resume from a known task
-		// state instead of losing all progress information on timeout.
-		result := map[string]interface{}{
-			"ticket":           ticket,
-			"token":            spec.Token,
-			"doc_type":         spec.DocType,
-			"file_extension":   spec.FileExtension,
-			"ready":            false,
-			"failed":           failed,
-			"job_status":       jobStatus,
-			"job_status_label": jobStatusLabel,
-			"timed_out":        true,
-			"next_command":     nextCommand,
-		}
-		if preferredFileName != "" {
-			result["file_name"] = ensureExportFileExtension(sanitizeExportFileName(preferredFileName, spec.Token), spec.FileExtension)
-		}
-		runtime.Out(result, nil)
-		fmt.Fprintf(runtime.IO().ErrOut, "Export task is still in progress. Continue with: %s\n", nextCommand)
-		return nil
-	},
+	nextCommand := driveExportTaskResultCommand(ticket, spec.Token)
+	if !hasObservedStatus && lastPollErr != nil {
+		hint := fmt.Sprintf(
+			"the export task was created but every status poll failed (ticket=%s)\nretry status lookup with: %s",
+			ticket,
+			nextCommand,
+		)
+		return appendDriveExportRecoveryHint(lastPollErr, hint)
+	}
+
+	failed := false
+	var jobStatus interface{}
+	jobStatusLabel := "unknown"
+	if hasObservedStatus {
+		failed = lastStatus.Failed()
+		jobStatus = lastStatus.JobStatus
+		jobStatusLabel = lastStatus.StatusLabel()
+	}
+	// Return the last observed status so callers can resume from a known task
+	// state instead of losing all progress information on timeout.
+	result := map[string]interface{}{
+		"ticket":           ticket,
+		"token":            spec.Token,
+		"doc_type":         spec.DocType,
+		"file_extension":   spec.FileExtension,
+		"ready":            false,
+		"failed":           failed,
+		"job_status":       jobStatus,
+		"job_status_label": jobStatusLabel,
+		"timed_out":        true,
+		"next_command":     nextCommand,
+	}
+	if preferredFileName != "" {
+		result["file_name"] = ensureExportFileExtension(sanitizeExportFileName(preferredFileName, spec.Token), spec.FileExtension)
+	}
+	runtime.Out(result, nil)
+	fmt.Fprintf(runtime.IO().ErrOut, "Export task is still in progress. Continue with: %s\n", nextCommand)
+	return nil
 }

shortcuts/drive/drive_export_test.go

@@ -5,6 +5,7 @@ package drive
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"net/http"
@@ -497,6 +498,72 @@ func TestDriveExportAsyncSuccess(t *testing.T) {
 	}
 }
 
+// TestDriveExportEmptyOutputDirDownloadsToCwd guards the export refactor: an
+// explicit empty --output-dir must still download to the current directory
+// (normalized to "."), not trigger the export-only no-download path that the
+// shared RunExport core uses for sheets +workbook-export.
+func TestDriveExportEmptyOutputDirDownloadsToCwd(t *testing.T) {
+	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/drive/v1/export_tasks",
+		Body:   map[string]interface{}{"code": 0, "data": map[string]interface{}{"ticket": "tk_e"}},
+	})
+	reg.Register(&httpmock.Stub{
+		Method: "GET",
+		URL:    "/open-apis/drive/v1/export_tasks/tk_e",
+		Body: map[string]interface{}{"code": 0, "data": map[string]interface{}{
+			"result": map[string]interface{}{
+				"job_status": 0, "file_token": "box_e", "file_name": "report",
+				"file_extension": "pdf", "type": "docx", "file_size": 3,
+			},
+		}},
+	})
+	reg.Register(&httpmock.Stub{
+		Method:  "GET",
+		URL:     "/open-apis/drive/v1/export_tasks/file/box_e/download",
+		Status:  200,
+		RawBody: []byte("pdf"),
+		Headers: http.Header{
+			"Content-Type":        []string{"application/pdf"},
+			"Content-Disposition": []string{`attachment; filename="report.pdf"`},
+		},
+	})
+
+	tmpDir := t.TempDir()
+	withDriveWorkingDir(t, tmpDir)
+
+	prevAttempts, prevInterval := driveExportPollAttempts, driveExportPollInterval
+	driveExportPollAttempts, driveExportPollInterval = 1, 0
+	t.Cleanup(func() {
+		driveExportPollAttempts, driveExportPollInterval = prevAttempts, prevInterval
+	})
+
+	err := mountAndRunDrive(t, DriveExport, []string{
+		"+export",
+		"--token", "docx123",
+		"--doc-type", "docx",
+		"--file-extension", "pdf",
+		"--output-dir", "",
+		"--as", "bot",
+	}, f, stdout)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	// Empty --output-dir must still write to cwd, not skip the download.
+	data, err := os.ReadFile(filepath.Join(tmpDir, "report.pdf"))
+	if err != nil {
+		t.Fatalf("empty --output-dir should still download to cwd: %v", err)
+	}
+	if string(data) != "pdf" {
+		t.Fatalf("downloaded content = %q", string(data))
+	}
+	if strings.Contains(stdout.String(), `"downloaded": false`) {
+		t.Fatalf("export-only path must not trigger for drive +export: %s", stdout.String())
+	}
+}
+
 func TestDriveExportAsyncUsesProvidedFileName(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
 	reg.Register(&httpmock.Stub{
@@ -1034,3 +1101,37 @@ func TestDriveTaskResultExportIncludesReadyFlags(t *testing.T) {
 		t.Fatalf("stdout missing job_status_label: %s", stdout.String())
 	}
 }
+
+// TestWrapExportContextErr verifies the export poll loop's typed wrapping for
+// context cancellation / deadline. Previously the poll loop returned ctx.Err()
+// directly so an untyped context.Canceled would escape as a plain string at
+// the command layer, bypassing the typed-error contract.
+func TestWrapExportContextErr(t *testing.T) {
+	if err := wrapExportContextErr(nil); err != nil {
+		t.Errorf("wrapExportContextErr(nil) = %v, want nil", err)
+	}
+
+	cancelled := wrapExportContextErr(context.Canceled)
+	var netErrCancel *errs.NetworkError
+	if !errors.As(cancelled, &netErrCancel) {
+		t.Fatalf("wrapExportContextErr(Canceled) = %T, want *errs.NetworkError", cancelled)
+	}
+	if netErrCancel.Subtype != errs.SubtypeNetworkTransport {
+		t.Errorf("Canceled subtype = %q, want %q", netErrCancel.Subtype, errs.SubtypeNetworkTransport)
+	}
+	if !errors.Is(cancelled, context.Canceled) {
+		t.Error("wrapExportContextErr should preserve context.Canceled via errors.Is")
+	}
+
+	deadline := wrapExportContextErr(context.DeadlineExceeded)
+	var netErrDeadline *errs.NetworkError
+	if !errors.As(deadline, &netErrDeadline) {
+		t.Fatalf("wrapExportContextErr(DeadlineExceeded) = %T, want *errs.NetworkError", deadline)
+	}
+	if netErrDeadline.Subtype != errs.SubtypeNetworkTimeout {
+		t.Errorf("DeadlineExceeded subtype = %q, want %q", netErrDeadline.Subtype, errs.SubtypeNetworkTimeout)
+	}
+	if !errors.Is(deadline, context.DeadlineExceeded) {
+		t.Error("wrapExportContextErr should preserve context.DeadlineExceeded via errors.Is")
+	}
+}

shortcuts/drive/drive_import.go

@@ -35,132 +35,164 @@ var DriveImport = common.Shortcut{
 		{Name: "target-token", Desc: "existing token to import data into (only for type=bitable); when set, data is mounted into this bitable instead of creating a new one"},
 	},
 	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		return validateDriveImportSpec(driveImportSpec{
-			FilePath:    runtime.Str("file"),
-			DocType:     strings.ToLower(runtime.Str("type")),
-			FolderToken: runtime.Str("folder-token"),
-			Name:        runtime.Str("name"),
-			TargetToken: runtime.Str("target-token"),
-		})
+		return ValidateImport(importParamsFromFlags(runtime))
 	},
 	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		spec := driveImportSpec{
-			FilePath:    runtime.Str("file"),
-			DocType:     strings.ToLower(runtime.Str("type")),
-			FolderToken: runtime.Str("folder-token"),
-			Name:        runtime.Str("name"),
-			TargetToken: runtime.Str("target-token"),
-		}
-		fileSize, err := preflightDriveImportFile(runtime.FileIO(), &spec)
-		if err != nil {
-			return common.NewDryRunAPI().Set("error", err.Error())
-		}
-		if valErr := validateDriveImportSpec(spec); valErr != nil {
-			return common.NewDryRunAPI().Set("error", valErr.Error())
-		}
-
-		dry := common.NewDryRunAPI()
-		dry.Desc("Upload file (single-part or multipart) -> create import task -> poll status")
-
-		appendDriveImportFolderTokenWikiCheckDryRun(dry, spec)
-		appendDriveImportUploadDryRun(dry, spec, fileSize)
-
-		dry.POST("/open-apis/drive/v1/import_tasks").
-			Desc("[2] Create import task").
-			Body(spec.CreateTaskBody("<file_token>"))
-
-		dry.GET("/open-apis/drive/v1/import_tasks/:ticket").
-			Desc("[3] Poll import task result").
-			Set("ticket", "<ticket>")
-		if runtime.IsBot() {
-			dry.Desc("After the import result returns the final cloud document target in bot mode, the CLI will also try to grant the current CLI user full_access (可管理权限) on it.")
-		}
-
-		return dry
+		return PlanImportDryRun(runtime, importParamsFromFlags(runtime))
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		spec := driveImportSpec{
-			FilePath:    runtime.Str("file"),
-			DocType:     strings.ToLower(runtime.Str("type")),
-			FolderToken: runtime.Str("folder-token"),
-			Name:        runtime.Str("name"),
-			TargetToken: runtime.Str("target-token"),
-		}
-		if _, err := preflightDriveImportFile(runtime.FileIO(), &spec); err != nil {
-			return err
-		}
-		if err := rejectDriveImportWikiFolderToken(runtime, spec.FolderToken); err != nil {
-			return err
-		}
-
-		// Step 1: Upload file as media
-		fileToken, uploadErr := uploadMediaForImport(ctx, runtime, spec.FilePath, spec.SourceFileName(), spec.DocType)
-		if uploadErr != nil {
-			return uploadErr
-		}
-
-		fmt.Fprintf(runtime.IO().ErrOut, "Creating import task for %s as %s...\n", spec.TargetFileName(), spec.DocType)
-
-		// Step 2: Create import task
-		ticket, err := createDriveImportTask(runtime, spec, fileToken)
-		if err != nil {
-			return err
-		}
-
-		// Step 3: Poll task
-		fmt.Fprintf(runtime.IO().ErrOut, "Polling import task %s...\n", ticket)
-
-		status, ready, err := pollDriveImportTask(runtime, ticket)
-		if err != nil {
-			return err
-		}
-
-		// Some intermediate responses omit the final type, so fall back to the
-		// requested type to keep the output shape stable.
-		resultType := status.DocType
-		if resultType == "" {
-			resultType = spec.DocType
-		}
-		out := map[string]interface{}{
-			"ticket":           ticket,
-			"type":             resultType,
-			"ready":            ready,
-			"job_status":       status.JobStatus,
-			"job_status_label": status.StatusLabel(),
-		}
-		if status.Token != "" {
-			out["token"] = status.Token
-		}
-		if statusURL := strings.TrimSpace(status.URL); statusURL != "" {
-			out["url"] = statusURL
-		} else if status.Token != "" {
-			if u := common.BuildResourceURL(runtime.Config.Brand, normalizeDriveImportKindForURL(resultType, spec.DocType), status.Token); u != "" {
-				out["url"] = u
-			}
-		}
-		if status.JobErrorMsg != "" {
-			out["job_error_msg"] = status.JobErrorMsg
-		}
-		if status.Extra != nil {
-			out["extra"] = status.Extra
-		}
-		if !ready {
-			nextCommand := driveImportTaskResultCommand(ticket)
-			fmt.Fprintf(runtime.IO().ErrOut, "Import task is still in progress. Continue with: %s\n", nextCommand)
-			out["timed_out"] = true
-			out["next_command"] = nextCommand
-		}
-		if ready {
-			if grant := common.AutoGrantCurrentUserDrivePermission(runtime, common.GetString(out, "token"), resultType); grant != nil {
-				out["permission_grant"] = grant
-			}
-		}
-
-		runtime.Out(out, nil)
-		return nil
+		return RunImport(ctx, runtime, importParamsFromFlags(runtime))
 	},
 }
 
+// ImportParams holds the user-facing inputs for an import flow, decoupled from
+// cobra flags so other command groups (e.g. sheets +workbook-import) can reuse
+// the drive import implementation without taking a dependency on a --type flag.
+type ImportParams struct {
+	File        string
+	DocType     string
+	FolderToken string
+	Name        string
+	TargetToken string
+}
+
+func (p ImportParams) spec() driveImportSpec {
+	return driveImportSpec{
+		FilePath:    p.File,
+		DocType:     strings.ToLower(p.DocType),
+		FolderToken: p.FolderToken,
+		Name:        p.Name,
+		TargetToken: p.TargetToken,
+	}
+}
+
+// importParamsFromFlags reads the standard drive +import flag set.
+func importParamsFromFlags(runtime *common.RuntimeContext) ImportParams {
+	return ImportParams{
+		File:        runtime.Str("file"),
+		DocType:     runtime.Str("type"),
+		FolderToken: runtime.Str("folder-token"),
+		Name:        runtime.Str("name"),
+		TargetToken: runtime.Str("target-token"),
+	}
+}
+
+// ValidateImport runs the CLI-level compatibility checks for an import.
+func ValidateImport(p ImportParams) error {
+	return validateDriveImportSpec(p.spec())
+}
+
+// PlanImportDryRun builds the dry-run plan (upload -> create task -> poll) for
+// an import without performing any network or file I/O beyond a local stat.
+func PlanImportDryRun(runtime *common.RuntimeContext, p ImportParams) *common.DryRunAPI {
+	spec := p.spec()
+	fileSize, err := preflightDriveImportFile(runtime.FileIO(), &spec)
+	if err != nil {
+		return common.NewDryRunAPI().Set("error", err.Error())
+	}
+	if valErr := validateDriveImportSpec(spec); valErr != nil {
+		return common.NewDryRunAPI().Set("error", valErr.Error())
+	}
+
+	dry := common.NewDryRunAPI()
+	dry.Desc("Upload file (single-part or multipart) -> create import task -> poll status")
+
+	appendDriveImportFolderTokenWikiCheckDryRun(dry, spec)
+	appendDriveImportUploadDryRun(dry, spec, fileSize)
+
+	dry.POST("/open-apis/drive/v1/import_tasks").
+		Desc("[2] Create import task").
+		Body(spec.CreateTaskBody("<file_token>"))
+
+	dry.GET("/open-apis/drive/v1/import_tasks/:ticket").
+		Desc("[3] Poll import task result").
+		Set("ticket", "<ticket>")
+	if runtime.IsBot() {
+		dry.Desc("After the import result returns the final cloud document target in bot mode, the CLI will also try to grant the current CLI user full_access (可管理权限) on it.")
+	}
+
+	return dry
+}
+
+// RunImport executes the full import flow: upload media -> create import task ->
+// bounded poll, then writes the result envelope to the runtime output. It is
+// the shared core behind both drive +import and sheets +workbook-import.
+func RunImport(ctx context.Context, runtime *common.RuntimeContext, p ImportParams) error {
+	spec := p.spec()
+	if _, err := preflightDriveImportFile(runtime.FileIO(), &spec); err != nil {
+		return err
+	}
+	if err := rejectDriveImportWikiFolderToken(runtime, spec.FolderToken); err != nil {
+		return err
+	}
+
+	// Step 1: Upload file as media
+	fileToken, uploadErr := uploadMediaForImport(ctx, runtime, spec.FilePath, spec.SourceFileName(), spec.DocType)
+	if uploadErr != nil {
+		return uploadErr
+	}
+
+	fmt.Fprintf(runtime.IO().ErrOut, "Creating import task for %s as %s...\n", spec.TargetFileName(), spec.DocType)
+
+	// Step 2: Create import task
+	ticket, err := createDriveImportTask(runtime, spec, fileToken)
+	if err != nil {
+		return err
+	}
+
+	// Step 3: Poll task
+	fmt.Fprintf(runtime.IO().ErrOut, "Polling import task %s...\n", ticket)
+
+	status, ready, err := pollDriveImportTask(runtime, ticket)
+	if err != nil {
+		return err
+	}
+
+	// Some intermediate responses omit the final type, so fall back to the
+	// requested type to keep the output shape stable.
+	resultType := status.DocType
+	if resultType == "" {
+		resultType = spec.DocType
+	}
+	out := map[string]interface{}{
+		"ticket":           ticket,
+		"type":             resultType,
+		"ready":            ready,
+		"job_status":       status.JobStatus,
+		"job_status_label": status.StatusLabel(),
+	}
+	if status.Token != "" {
+		out["token"] = status.Token
+	}
+	if statusURL := strings.TrimSpace(status.URL); statusURL != "" {
+		out["url"] = statusURL
+	} else if status.Token != "" {
+		if u := common.BuildResourceURL(runtime.Config.Brand, normalizeDriveImportKindForURL(resultType, spec.DocType), status.Token); u != "" {
+			out["url"] = u
+		}
+	}
+	if status.JobErrorMsg != "" {
+		out["job_error_msg"] = status.JobErrorMsg
+	}
+	if status.Extra != nil {
+		out["extra"] = status.Extra
+	}
+	if !ready {
+		nextCommand := driveImportTaskResultCommand(ticket)
+		fmt.Fprintf(runtime.IO().ErrOut, "Import task is still in progress. Continue with: %s\n", nextCommand)
+		out["timed_out"] = true
+		out["next_command"] = nextCommand
+	}
+	if ready {
+		if grant := common.AutoGrantCurrentUserDrivePermission(runtime, common.GetString(out, "token"), resultType); grant != nil {
+			out["permission_grant"] = grant
+		}
+	}
+
+	runtime.Out(out, nil)
+	return nil
+}
+
 func preflightDriveImportFile(fio fileio.FileIO, spec *driveImportSpec) (int64, error) {
 	// Keep dry-run and execution aligned on path normalization, file existence,
 	// and format-specific size limits before planning the upload path.

shortcuts/mail/mail_triage.go

@@ -770,13 +770,7 @@ func buildListParams(runtime *common.RuntimeContext, mailboxID string, f triageF
 				params["folder_id"] = folderIDFromFilter
 			}
 		} else {
-			resolved, err := resolveFolderID(runtime, mailboxID, folderIDFromFilter)
-			if err != nil {
-				return nil, err
-			}
-			if resolved != "" {
-				params["folder_id"] = resolved
-			}
+			params["folder_id"] = folderIDFromFilter
 		}
 	} else if folderFromFilter != "" {
 		if dryRun {
@@ -786,13 +780,7 @@ func buildListParams(runtime *common.RuntimeContext, mailboxID string, f triageF
 				params["folder_id"] = folderFromFilter
 			}
 		} else {
-			resolved, err := resolveFolderName(runtime, mailboxID, folderFromFilter)
-			if err != nil {
-				return nil, err
-			}
-			if resolved != "" {
-				params["folder_id"] = resolved
-			}
+			params["folder_id"] = folderFromFilter
 		}
 	}
 
@@ -811,13 +799,7 @@ func buildListParams(runtime *common.RuntimeContext, mailboxID string, f triageF
 				params["label_id"] = labelIDFromFilter
 			}
 		} else {
-			resolved, err := resolveLabelID(runtime, mailboxID, labelIDFromFilter)
-			if err != nil {
-				return nil, err
-			}
-			if resolved != "" {
-				params["label_id"] = resolved
-			}
+			params["label_id"] = labelIDFromFilter
 		}
 	} else if labelFromFilter != "" {
 		if dryRun {
@@ -827,13 +809,7 @@ func buildListParams(runtime *common.RuntimeContext, mailboxID string, f triageF
 				params["label_id"] = labelFromFilter
 			}
 		} else {
-			resolved, err := resolveLabelName(runtime, mailboxID, labelFromFilter)
-			if err != nil {
-				return nil, err
-			}
-			if resolved != "" {
-				params["label_id"] = resolved
-			}
+			params["label_id"] = labelFromFilter
 		}
 	}
 

shortcuts/mail/mail_triage_test.go

@@ -12,6 +12,7 @@ import (
 	"testing"
 
 	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/httpmock"
 	"github.com/larksuite/cli/shortcuts/common"
@@ -974,7 +975,11 @@ func TestBuildListParamsDryRunOnlyUnread(t *testing.T) {
 func TestBuildListParamsDryRunFolderAlias(t *testing.T) {
 	rt := runtimeForMailTriageTest(t, nil)
 	f := triageFilter{Folder: "sent"}
-	got, err := buildListParams(rt, "me", f, 20, "", true)
+	resolved, err := resolveListFilter(rt, "me", f, true)
+	if err != nil {
+		t.Fatalf("resolveListFilter: %v", err)
+	}
+	got, err := buildListParams(rt, "me", resolved, 20, "", true)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -983,10 +988,30 @@ func TestBuildListParamsDryRunFolderAlias(t *testing.T) {
 	}
 }
 
+func TestBuildListParamsDryRunCustomFolderPreservesInput(t *testing.T) {
+	rt := runtimeForMailTriageTest(t, nil)
+	f := triageFilter{Folder: "team-folder"}
+	resolved, err := resolveListFilter(rt, "me", f, true)
+	if err != nil {
+		t.Fatalf("resolveListFilter: %v", err)
+	}
+	got, err := buildListParams(rt, "me", resolved, 20, "", true)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got["folder_id"] != "team-folder" {
+		t.Fatalf("expected dry-run folder_id=team-folder, got %v", got["folder_id"])
+	}
+}
+
 func TestBuildListParamsDryRunLabelAlias(t *testing.T) {
 	rt := runtimeForMailTriageTest(t, nil)
 	f := triageFilter{Label: "flagged"}
-	got, err := buildListParams(rt, "me", f, 10, "", true)
+	resolved, err := resolveListFilter(rt, "me", f, true)
+	if err != nil {
+		t.Fatalf("resolveListFilter: %v", err)
+	}
+	got, err := buildListParams(rt, "me", resolved, 10, "", true)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -995,6 +1020,25 @@ func TestBuildListParamsDryRunLabelAlias(t *testing.T) {
 	}
 }
 
+func TestBuildListParamsDryRunCustomLabelPreservesInput(t *testing.T) {
+	rt := runtimeForMailTriageTest(t, nil)
+	f := triageFilter{Label: "custom-label"}
+	resolved, err := resolveListFilter(rt, "me", f, true)
+	if err != nil {
+		t.Fatalf("resolveListFilter: %v", err)
+	}
+	got, err := buildListParams(rt, "me", resolved, 10, "", true)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if _, ok := got["folder_id"]; ok {
+		t.Fatalf("folder_id should not be set when label is specified, got %v", got["folder_id"])
+	}
+	if got["label_id"] != "custom-label" {
+		t.Fatalf("expected dry-run label_id=custom-label, got %v", got["label_id"])
+	}
+}
+
 // --- buildSearchParams additional coverage ---
 
 func TestBuildSearchParamsAllFilterFields(t *testing.T) {
@@ -1791,3 +1835,137 @@ func mailTriageSearchItem(messageID, subject string) map[string]interface{} {
 		},
 	}
 }
+
+// registerMailTriageFoldersListStub registers a NON-reusable stub for the
+// mailbox folders list API. Because it is non-reusable, any second hit returns
+// "httpmock: no stub for GET .../folders" — which is exactly the assertion we
+// use to prove resolveListFilter runs once and buildListParams does NOT
+// re-resolve. folderID/folderName is the single custom folder the API reports.
+func registerMailTriageFoldersListStub(reg *httpmock.Registry, mailbox, folderID, folderName string) {
+	reg.Register(&httpmock.Stub{
+		Method: "GET",
+		URL:    mailboxPath(mailbox, "folders"),
+		Body: map[string]interface{}{
+			"code": 0,
+			"data": map[string]interface{}{
+				"items": []interface{}{
+					map[string]interface{}{
+						"id":   folderID,
+						"name": folderName,
+					},
+				},
+			},
+		},
+	})
+}
+
+// registerMailTriageListPageStub registers one page of the messages list API,
+// disambiguated from sibling pages by a URL substring unique to that page
+// (e.g. "page_size=5" for page 1 vs "page_size=2" for page 2). The substring
+// must NOT depend on query-param ordering: map iteration makes param order
+// nondeterministic, so prefer a value-only token like "page_size=N" (the N
+// differs per page because pageSize = maxCount - fetched_so_far). Non-reusable
+// so reg.Verify catches under- or over-consumption.
+func registerMailTriageListPageStub(reg *httpmock.Registry, urlSubstring string, items []string, hasMore bool, pageToken string) {
+	data := map[string]interface{}{
+		"items":    items,
+		"has_more": hasMore,
+	}
+	if pageToken != "" {
+		data["page_token"] = pageToken
+	}
+	reg.Register(&httpmock.Stub{
+		Method: "GET",
+		URL:    urlSubstring,
+		Body: map[string]interface{}{
+			"code": 0,
+			"data": data,
+		},
+	})
+}
+
+// TestMailTriageCustomFolderResolvesOnceAcrossListPages is the regression test
+// for the bug where buildListParams re-called resolveFolderID on every list
+// page, turning "resolve once" into "1 + page_count" folder-list API calls and
+// easily tripping rate limits.
+//
+// Setup: a custom folder filter that forces resolveListFilter to hit the
+// folders list API once (to map folder name "team-folder" to folder_id), then two
+// messages-list pages. The folders list stub is non-reusable, so if
+// buildListParams re-resolves, the second hit fails with "no stub". The
+// messages-list stubs are page-specific (disambiguated by page_size in the
+// URL), so both pages are served and Verify asserts each fired exactly once.
+func TestMailTriageCustomFolderResolvesOnceAcrossListPages(t *testing.T) {
+	f, stdout, _, reg := mailShortcutTestFactory(t)
+	defer reg.Verify(t)
+
+	// listMailboxFolders (called once by resolveListFilter) gates on the
+	// mail:user_mailbox.folder:read scope, which the default test token does
+	// not carry. Re-store the token with that scope appended so the folders
+	// API call is actually exercised (and thus the non-reusable folders stub
+	// is the load-bearing "exactly once" assertion).
+	const folderScope = "mail:user_mailbox.folder:read"
+	cfg := mailTestConfig()
+	if stored := auth.GetStoredToken(cfg.AppID, cfg.UserOpenId); stored != nil {
+		if !strings.Contains(stored.Scope, folderScope) {
+			stored.Scope = stored.Scope + " " + folderScope
+			if err := auth.SetStoredToken(stored); err != nil {
+				t.Fatalf("re-store token with folder scope: %v", err)
+			}
+		}
+	}
+
+	const (
+		mailbox    = "me"
+		folderName = "team-folder"
+		folderID   = "fld_custom_team"
+		page2Token = "tok_page2"
+	)
+	// --max 5 with listPageMax=20 → pageSize = 5-0 = 5 on page 1, then 5-3 = 2
+	// on page 2. The page_size query value disambiguates the two list stubs.
+	page1IDs := []string{"msg_a", "msg_b", "msg_c"}
+	page2IDs := []string{"msg_d", "msg_e"}
+
+	// Folders list: registered exactly once, non-reusable. Any second folder
+	// lookup (the bug) fails the test with "no stub for GET .../folders".
+	registerMailTriageFoldersListStub(reg, mailbox, folderID, folderName)
+	// Messages list, page 1: 3 ids, has_more, hands off a page-2 token. The
+	// page_size value (5 = maxCount - 0) is unique to page 1; page 2 uses 2.
+	registerMailTriageListPageStub(reg, "page_size=5", page1IDs, true, page2Token)
+	// Messages list, page 2: 2 ids, terminal.
+	registerMailTriageListPageStub(reg, "page_size=2", page2IDs, false, "")
+	// Batch metadata fetch for all 5 ids.
+	registerMailTriageBatchStub(reg, mailbox, []map[string]interface{}{
+		mailTriageBatchMessage("msg_a", "Subject A"),
+		mailTriageBatchMessage("msg_b", "Subject B"),
+		mailTriageBatchMessage("msg_c", "Subject C"),
+		mailTriageBatchMessage("msg_d", "Subject D"),
+		mailTriageBatchMessage("msg_e", "Subject E"),
+	})
+
+	args := []string{
+		"+triage",
+		"--as", "user",
+		"--mailbox", mailbox,
+		"--filter", `{"folder":"` + folderName + `"}`,
+		"--max", "5",
+		"--format", "json",
+	}
+	if err := runMountedMailShortcut(t, MailTriage, args, f, stdout); err != nil {
+		t.Fatalf("unexpected error running +triage (likely a second folders API call — the bug): %v", err)
+	}
+
+	data := decodeMailTriageJSONOutput(t, stdout)
+	messages := mailTriageMessagesFromOutput(t, data)
+	if len(messages) != 5 {
+		t.Fatalf("expected 5 messages across 2 pages, got %d (stdout=%s)", len(messages), stdout.String())
+	}
+	if got := data["has_more"]; got != false {
+		t.Fatalf("expected has_more=false after exhausting pages, got %v", got)
+	}
+	// All registered stubs (1 folders + 2 list pages + 1 batch_get) are
+	// non-reusable; reg.Verify (deferred above) asserts each was matched
+	// exactly once. Combined with the non-reusable folders stub, this is the
+	// proof that the folders list API was called exactly once across both
+	// pages — the core invariant the fix restores.
+}

shortcuts/sheets/batch_op_contract_test.go

@@ -177,6 +177,18 @@ func TestBatchOp_BodyMatchesStandalone(t *testing.T) {
 			args:     []string{"--sheet-id", "sh1", "--color", "#FF0000"},
 			subInput: `{"sheet-id":"sh1","color":"#FF0000"}`,
 		},
+		{
+			shortcut: "+sheet-show-gridline",
+			sc:       SheetShowGridline,
+			args:     []string{"--sheet-id", "sh1"},
+			subInput: `{"sheet-id":"sh1"}`,
+		},
+		{
+			shortcut: "+sheet-hide-gridline",
+			sc:       SheetHideGridline,
+			args:     []string{"--sheet-id", "sh1"},
+			subInput: `{"sheet-id":"sh1"}`,
+		},
 		{
 			shortcut: "+dropdown-set",
 			sc:       DropdownSet,
@@ -432,12 +444,7 @@ func TestBatchOp_ErrorEquivalence(t *testing.T) {
 				t, tc.shortcut,
 				append([]string{"--url", testURL, "--dry-run"}, tc.args...),
 			)
-			if standaloneErr == nil {
-				t.Fatalf("standalone Validate accepted bad input — expected error containing %q", tc.wantContains)
-			}
-			if !strings.Contains(standaloneErr.Error(), tc.wantContains) {
-				t.Errorf("standalone error = %q, want substring %q", standaloneErr.Error(), tc.wantContains)
-			}
+			requireValidation(t, standaloneErr, tc.wantContains)
 
 			// Batch path: translate the matching sub-op. The translator wraps
 			// the inner error with "operations[i] (<shortcut>): " — assert the
@@ -451,17 +458,12 @@ func TestBatchOp_ErrorEquivalence(t *testing.T) {
 				"input":    subInput,
 			}
 			_, batchErr := translateBatchOp(rawOp, testToken, 0)
-			if batchErr == nil {
-				t.Fatalf("batch translator accepted bad input — expected error containing %q", tc.wantContains)
-			}
-			if !strings.Contains(batchErr.Error(), tc.wantContains) {
-				t.Errorf("batch error = %q, want substring %q (operations[i] prefix is fine)", batchErr.Error(), tc.wantContains)
-			}
+			batchVE := requireValidation(t, batchErr, tc.wantContains)
 			// And the wrap context must include the sub-op index + shortcut
 			// name so error reports stay actionable in multi-op batches.
 			wrapHint := "operations[0] (" + tc.subShortcut + "):"
-			if !strings.Contains(batchErr.Error(), wrapHint) {
-				t.Errorf("batch error %q missing context prefix %q", batchErr.Error(), wrapHint)
+			if !strings.Contains(batchVE.Message, wrapHint) {
+				t.Errorf("batch error %q missing context prefix %q", batchVE.Message, wrapHint)
 			}
 		})
 	}
@@ -517,12 +519,7 @@ func TestBatchOp_RejectsWrongScalarType(t *testing.T) {
 			}
 			rawOp := map[string]interface{}{"shortcut": tc.subShortcut, "input": subInput}
 			_, err := translateBatchOp(rawOp, testToken, 0)
-			if err == nil {
-				t.Fatalf("translateBatchOp accepted wrong-typed field; want error containing %q", tc.wantContains)
-			}
-			if !strings.Contains(err.Error(), tc.wantContains) {
-				t.Errorf("error = %q, want substring %q", err.Error(), tc.wantContains)
-			}
+			requireValidation(t, err, tc.wantContains)
 		})
 	}
 }
@@ -580,12 +577,7 @@ func TestBatchOp_GuardsBeyondCobra(t *testing.T) {
 			}
 			rawOp := map[string]interface{}{"shortcut": tc.subShortcut, "input": subInput}
 			_, err := translateBatchOp(rawOp, testToken, 0)
-			if err == nil {
-				t.Fatalf("translateBatchOp accepted bad input; want error containing %q", tc.wantContains)
-			}
-			if !strings.Contains(err.Error(), tc.wantContains) {
-				t.Errorf("error = %q, want substring %q", err.Error(), tc.wantContains)
-			}
+			requireValidation(t, err, tc.wantContains)
 		})
 	}
 }
@@ -716,12 +708,7 @@ func TestBatchOp_RejectsBadSubOpInput(t *testing.T) {
 				"input":    subInput,
 			}
 			_, err := translateBatchOp(rawOp, testToken, 0)
-			if err == nil {
-				t.Fatalf("translator accepted bad input — expected error containing %q", tc.wantContains)
-			}
-			if !strings.Contains(err.Error(), tc.wantContains) {
-				t.Errorf("error = %q, want substring %q", err.Error(), tc.wantContains)
-			}
+			requireValidation(t, err, tc.wantContains)
 		})
 	}
 }
@@ -782,12 +769,7 @@ func TestBatchOp_SchemaValidatesSubOps(t *testing.T) {
 				"input":    subInput,
 			}
 			_, err := translateBatchOp(rawOp, testToken, 0)
-			if err == nil {
-				t.Fatalf("translator accepted schema-violating sub-op — expected error containing %q", tc.wantContains)
-			}
-			if !strings.Contains(err.Error(), tc.wantContains) {
-				t.Errorf("error = %q, want substring %q", err.Error(), tc.wantContains)
-			}
+			requireValidation(t, err, tc.wantContains)
 		})
 	}
 }

shortcuts/sheets/batch_op_dispatch.go

@@ -150,6 +150,12 @@ var batchOpDispatch = map[string]batchOpMapping{
 		return sheetVisibilityInput(fv, t, sid, sn, "unhide")
 	}},
 	"+sheet-set-tab-color": {"modify_workbook_structure", sheetSetTabColorInput},
+	"+sheet-show-gridline": {"modify_workbook_structure", func(fv flagView, t, sid, sn string) (map[string]interface{}, error) {
+		return sheetVisibilityInput(fv, t, sid, sn, "show_gridline")
+	}},
+	"+sheet-hide-gridline": {"modify_workbook_structure", func(fv flagView, t, sid, sn string) (map[string]interface{}, error) {
+		return sheetVisibilityInput(fv, t, sid, sn, "hide_gridline")
+	}},
 
 	// ─── 对象族 CRUD (manage_*_object, operation 区分) ─────────────
 	"+chart-create": {"manage_chart_object", objCreateTranslate(chartSpec)},

shortcuts/sheets/csv_put_guard_test.go

@@ -4,12 +4,10 @@
 package sheets
 
 import (
-	"errors"
 	"os"
 	"strings"
 	"testing"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	_ "github.com/larksuite/cli/internal/vfs/localfileio"
 	"github.com/larksuite/cli/shortcuts/common"
@@ -37,18 +35,9 @@ func TestGuardCSVValueIsNotFilePath(t *testing.T) {
 
 	// Bare value naming an existing file → guarded with a fix-it hint.
 	err := guardCSVValueIsNotFilePath(newCSVGuardRuntime("data.csv"))
-	if err == nil {
-		t.Fatal("expected guard error when --csv names an existing file")
-	}
-	if !strings.Contains(err.Error(), "existing file") || !strings.Contains(err.Error(), "@data.csv") {
-		t.Errorf("error should flag the file and suggest @data.csv, got: %v", err)
-	}
-	if p, ok := errs.ProblemOf(err); !ok || p.Category != errs.CategoryValidation || p.Subtype != errs.SubtypeInvalidArgument {
-		t.Errorf("problem = %+v, want validation/invalid_argument", p)
-	}
-	var ve *errs.ValidationError
-	if !errors.As(err, &ve) {
-		t.Fatalf("guard error = %T, want *errs.ValidationError", err)
+	ve := requireValidation(t, err, "existing file")
+	if !strings.Contains(ve.Message, "@data.csv") {
+		t.Errorf("message should suggest @data.csv, got: %q", ve.Message)
 	}
 	if ve.Param != "--csv" {
 		t.Errorf("param = %q, want --csv", ve.Param)

shortcuts/sheets/csv_put_range_alias_test.go

@@ -4,7 +4,6 @@
 package sheets
 
 import (
-	"strings"
 	"testing"
 )
 
@@ -44,12 +43,7 @@ func TestCsvPutInput_RejectsStartCellAndRangeTogether(t *testing.T) {
 		"range":      "A1:H17",
 	})
 	_, err := csvPutInput(fv, "tok", "sid", "")
-	if err == nil {
-		t.Fatal("csvPutInput accepted both start-cell and range; want mutual-exclusion error")
-	}
-	if !strings.Contains(err.Error(), "--start-cell and --range are mutually exclusive") {
-		t.Errorf("error = %q, want it to mention start-cell/range mutual exclusion", err.Error())
-	}
+	requireValidation(t, err, "--start-cell and --range are mutually exclusive")
 }
 
 // With neither --start-cell nor --range explicitly set, csvPutInput rejects the
@@ -61,12 +55,7 @@ func TestCsvPutInput_RejectsStartCellAndRangeTogether(t *testing.T) {
 func TestCsvPutInput_RequiresStartCellOrRange(t *testing.T) {
 	fv := newMapFlagViewForCommand("+csv-put", map[string]interface{}{"csv": "a,b"})
 	_, err := csvPutInput(fv, "tok", "sid", "")
-	if err == nil {
-		t.Fatal("csvPutInput accepted missing start-cell/range; want a required-flag error")
-	}
-	if !strings.Contains(err.Error(), "--start-cell or --range is required") {
-		t.Errorf("error = %q, want it to mention '--start-cell or --range is required'", err.Error())
-	}
+	requireValidation(t, err, "--start-cell or --range is required")
 }
 
 // csvPutWriteRangeFromInput surfaces the real paste footprint so agents can see

shortcuts/sheets/data/flag-defs.json

@@ -54,7 +54,7 @@
         "kind": "own",
         "type": "int",
         "required": "optional",
-        "desc": "Insert position; appended to the end when omitted",
+        "desc": "Insert position (0-based); appended to the end when omitted",
         "default": "-1"
       },
       {
@@ -413,6 +413,86 @@
       }
     ]
   },
+  "+sheet-hide-gridline": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "sheet-id",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet reference_id (XOR with `--sheet-name`)"
+      },
+      {
+        "name": "sheet-name",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet name (XOR with `--sheet-id`)"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
+  "+sheet-show-gridline": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "sheet-id",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet reference_id (XOR with `--sheet-name`)"
+      },
+      {
+        "name": "sheet-name",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet name (XOR with `--sheet-id`)"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
   "+workbook-create": {
     "risk": "write",
     "flags": [
@@ -431,27 +511,45 @@
         "desc": "Target folder token; placed at the drive root when omitted"
       },
       {
-        "name": "headers",
+        "name": "values",
         "kind": "own",
         "type": "string",
         "required": "optional",
-        "desc": "Header row as a JSON array: `[\"Col A\",\"Col B\"]`",
+        "desc": "Untyped initial data as one 2D JSON array (`[[\"alice\",95]]`); values are written as-is with their type auto-detected, through the same batched set_cell_range path as --sheets — pair with --styles for number formats, colors, merges, and row/col sizes",
         "input": [
           "file",
           "stdin"
         ]
       },
       {
-        "name": "values",
+        "name": "sheets",
         "kind": "own",
         "type": "string",
         "required": "optional",
-        "desc": "Initial data as a 2D JSON array: `[[\"alice\",95]]`",
+        "desc": "Typed table payload as JSON (same shape as `+table-put`): a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it from a DataFrame via `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. Mutually exclusive with --values and --dataframe. Creates the workbook, then writes typed type-faithful data (dates land as real dates, numbers keep precision).",
         "input": [
           "file",
           "stdin"
         ]
       },
+      {
+        "name": "styles",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Initial visual operations as JSON: top-level `{styles:[...]}`. Each item corresponds to one target sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. With --sheets, styles array length/order/name must match --sheets.sheets. With --values, pass exactly one styles item for the initial sheet (its name is ignored).",
+        "input": [
+          "file",
+          "stdin"
+        ]
+      },
+      {
+        "name": "dataframe",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Single-sheet typed table from one Arrow IPC file (Feather v2 — what `pandas.DataFrame.to_feather()` writes), mutually exclusive with --values and --sheets. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema; per-column `number_format` may be set via Arrow field metadata. Creates the workbook and fills its default sheet (`Sheet1` — adopted in place, no empty Sheet1 left behind). For multi-sheet or non-default placement, use `--sheets` instead."
+      },
       {
         "name": "dry-run",
         "kind": "system",
@@ -502,7 +600,7 @@
         "kind": "own",
         "type": "string",
         "required": "optional",
-        "desc": "Local save path; export is triggered but not downloaded when omitted"
+        "desc": "Local save path. When omitted, **only the export task is triggered + polled, the file is NOT downloaded** (returns file_token / status so a later step can resume the download). Pass a concrete path (e.g. `./out.xlsx`) or a directory (`.` keeps the server-provided filename) to download. Note: the equivalent `lark-cli drive +export --doc-type sheet` uses three separate flags (`--output-dir` / `--file-name` / `--overwrite`) and defaults to downloading into the current directory; this wrapper collapses them into a single `--output-path` for ergonomics but defaults to no-download — fall back to `drive +export` if the split flag set fits better."
       },
       {
         "name": "dry-run",
@@ -513,6 +611,32 @@
       }
     ]
   },
+  "+workbook-import": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "file",
+        "kind": "own",
+        "type": "string",
+        "required": "required",
+        "desc": "Local file path (.xlsx / .xls / .csv)"
+      },
+      {
+        "name": "folder-token",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Target folder token; imported to the cloud drive root when omitted"
+      },
+      {
+        "name": "name",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Imported spreadsheet name; defaults to the local file name without its extension"
+      }
+    ]
+  },
   "+sheet-info": {
     "risk": "read",
     "flags": [
@@ -1082,9 +1206,8 @@
         "kind": "own",
         "type": "int",
         "required": "optional",
-        "desc": "Safety cap; default 200000",
-        "default": "200000",
-        "hidden": true
+        "desc": "Max output chars per call; default 500000 (safety cap). Large reads are usually better redirected to a file; only lower it (e.g. 25000) when you want results inline without triggering file offload, paging via has_more",
+        "default": "500000"
       },
       {
         "name": "skip-hidden",
@@ -1192,9 +1315,8 @@
         "kind": "own",
         "type": "int",
         "required": "optional",
-        "desc": "Safety cap; default 200000",
-        "default": "200000",
-        "hidden": true
+        "desc": "Max output chars per call; default 500000 (safety cap). Large reads are usually better redirected to a file; only lower it (e.g. 25000) when you want results inline without triggering file offload, paging via has_more",
+        "default": "500000"
       },
       {
         "name": "include-row-prefix",
@@ -1212,19 +1334,72 @@
         "desc": "Skip hidden rows and columns; default `false`"
       },
       {
-        "name": "rows-json",
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": "Print the request path and parameters without executing"
+      }
+    ]
+  },
+  "+table-get": {
+    "risk": "read",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "sheet-id",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Read only this sheet (by id); omit to read all sheets"
+      },
+      {
+        "name": "sheet-name",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Read only this sheet (by name); omit to read all sheets"
+      },
+      {
+        "name": "range",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "A1 range to read; omit to read each sheet's full used range (spans internal blank rows/columns, not just the A1 current region)"
+      },
+      {
+        "name": "no-header",
         "kind": "own",
         "type": "bool",
         "required": "optional",
-        "desc": "Return structured rows ({row_number, values:{col→cell}}) instead of CSV text; default false",
-        "default": "false"
+        "desc": "Treat the first row as data instead of a header (columns get positional names col1, col2, ...)"
+      },
+      {
+        "name": "dataframe-out",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Write the typed table as one Arrow IPC file (Feather v2) instead of the default JSON. Pass `@<path>` for a file or `-` for binary stdout (same convention as other binary I/O flags). Mirror of the input-side `--dataframe` on `+table-put` / `+workbook-create` — pandas users round-trip via `df = pd.read_feather(\"x.arrow\")` or `pd.read_feather(io.BytesIO(stdout))`. Single-sheet only: requires `--sheet-id` or `--sheet-name`; whole-workbook reads keep the default JSON path. Column types come from the typed read-back (string/number/date/bool); per-column `number_format` is preserved as Arrow field metadata so the Arrow file can round-trip straight back through `+table-put --dataframe`."
       },
       {
         "name": "dry-run",
         "kind": "system",
         "type": "bool",
         "required": "optional",
-        "desc": "Print the request path and parameters without executing"
+        "desc": ""
       }
     ]
   },
@@ -1849,7 +2024,7 @@
         "kind": "own",
         "type": "string",
         "required": "required",
-        "desc": "RFC 4180 CSV text; plain values only (no formulas / styles / comments)",
+        "desc": "RFC 4180 CSV text; values or formulas (a leading = is evaluated as a formula); no styles / comments / images (use +cells-set for those).",
         "input": [
           "file",
           "stdin"
@@ -1880,6 +2055,61 @@
       }
     ]
   },
+  "+table-put": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL to write into (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token to write into (XOR with `--url`)"
+      },
+      {
+        "name": "sheets",
+        "kind": "own",
+        "type": "string",
+        "required": "xor",
+        "desc": "Typed table payload (pandas-DataFrame-shaped) as JSON, XOR with `--dataframe`: a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it with `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. `dtypes` values are pandas dtype strings (`int64`, `float64`, `Int64`, `bool`, `boolean`, `datetime64[ns]`, `object`, ...); the writer maps them to internal string/number/date/bool — omit `dtypes` and a column writes as text (good for raw CSV-shaped data). `formats[col]` is an Excel number_format string (e.g. `#,##0.00`, `0.0%`, `yyyy-mm`); when absent, date columns default to `yyyy-mm-dd` and string columns to text format (`@`).",
+        "input": [
+          "file",
+          "stdin"
+        ]
+      },
+      {
+        "name": "dataframe",
+        "kind": "own",
+        "type": "string",
+        "required": "xor",
+        "desc": "Single-sheet typed table from one Arrow IPC file (a.k.a. Feather v2 — what `pandas.DataFrame.to_feather()` writes), XOR with `--sheets`. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema (int*/uint*/float* → number, date32/date64/timestamp → date, utf8/large_utf8 → string, bool → bool); per-column `number_format` may be set via Arrow field metadata (`pa.field(\"price\", pa.float64(), metadata={b\"number_format\": b\"$#,##0.00\"})`). Writes the sheet at default placement: name `Sheet1` (created when absent), overwrite from A1 with header. For a different sheet name, anchor, mode, or to write multiple sheets, use `--sheets` instead."
+      },
+      {
+        "name": "styles",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Visual operations applied after the typed write, as JSON: top-level `{styles:[...]}`. Each item corresponds to one written sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. The styles array length/order/name must match the written sheets: with --sheets, match --sheets.sheets; with --dataframe (single sheet named Sheet1), pass exactly one styles item with name `Sheet1`. Run `+table-put --print-schema --flag-name styles` for the full cell_styles field schema.",
+        "input": [
+          "file",
+          "stdin"
+        ]
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
   "+cells-clear": {
     "risk": "high-risk-write",
     "flags": [

shortcuts/sheets/data/flag-schemas.json

@@ -1,5 +1,5 @@
 {
-  "schema_version": "2",
+  "schema_version": "3",
   "flags": {
     "+batch-update": {
       "operations": {
@@ -44,6 +44,8 @@
                 "+sheet-hide",
                 "+sheet-unhide",
                 "+sheet-set-tab-color",
+                "+sheet-show-gridline",
+                "+sheet-hide-gridline",
                 "+chart-create",
                 "+chart-update",
                 "+chart-delete",
@@ -454,7 +456,7 @@
                       "type": "object"
                     },
                     "link": {
-                      "description": "超链接地址（仅 type='link' 时必填）",
+                      "description": "超链接地址（type='link' 时必填）；@文档 mention（mention_type 非 0）时也必填，传文档 URL（如搜索结果里的文档链接），否则卡片不可点。@人（mention_type=0）不需要传",
                       "type": "string"
                     },
                     "mention_token": {
@@ -462,8 +464,21 @@
                       "type": "string"
                     },
                     "mention_type": {
-                      "description": "@提及类型编号（仅 type='mention' 时可选）",
-                      "type": "number"
+                      "description": "@提及类型编号（仅 type='mention' 时可选）。0 或不填=@用户；@文件时按类型取：1=文档 3=电子表格 8=多维表格 11=思维笔记 12=文件 15=旧版幻灯片 16=知识库 22=新版文档 30=幻灯片 38=画板",
+                      "type": "number",
+                      "enum": [
+                        0,
+                        1,
+                        3,
+                        8,
+                        11,
+                        12,
+                        15,
+                        16,
+                        22,
+                        30,
+                        38
+                      ]
                     },
                     "notify": {
                       "description": "是否发送通知（仅 type='mention' 时可选，默认 true）",
@@ -1730,11 +1745,12 @@
                             },
                             "aggregateType": {
                               "type": "string",
-                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效",
+                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效。count 只统计数值单元格；counta 统计所有非空单元格（含文本），按文本/分类列统计出现次数（如各类别的数量、频次分布）时用 counta。",
                               "enum": [
                                 "sum",
                                 "average",
                                 "count",
+                                "counta",
                                 "min",
                                 "max",
                                 "median"
@@ -1787,11 +1803,7 @@
               "data"
             ]
           }
-        },
-        "required": [
-          "position",
-          "size"
-        ]
+        }
       }
     },
     "+chart-update": {
@@ -2769,11 +2781,12 @@
                             },
                             "aggregateType": {
                               "type": "string",
-                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效",
+                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效。count 只统计数值单元格；counta 统计所有非空单元格（含文本），按文本/分类列统计出现次数（如各类别的数量、频次分布）时用 counta。",
                               "enum": [
                                 "sum",
                                 "average",
                                 "count",
+                                "counta",
                                 "min",
                                 "max",
                                 "median"
@@ -2826,11 +2839,7 @@
               "data"
             ]
           }
-        },
-        "required": [
-          "position",
-          "size"
-        ]
+        }
       }
     },
     "+cond-format-create": {
@@ -6249,6 +6258,744 @@
           }
         }
       }
+    },
+    "+table-put": {
+      "sheets": {
+        "type": "array",
+        "minItems": 1,
+        "description": "一个或多个子表的 typed 数据，每个数组元素写入一张子表；支持多 DataFrame → 多子表一次写入。整体形状对齐 pandas `df.to_json(orient=\"split\")`：列名走 `columns`、二维取值走 `data`、每列的 pandas dtype 走 `dtypes`、可选的展示格式走 `formats`。一行式用法：`{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`。",
+        "items": {
+          "type": "object",
+          "required": [
+            "name",
+            "columns",
+            "data"
+          ],
+          "properties": {
+            "name": {
+              "type": "string",
+              "description": "目标子表名。按名匹配已有子表；不存在则新建该子表。同一次调用内子表名不可重复。"
+            },
+            "start_cell": {
+              "type": "string",
+              "default": "A1",
+              "description": "写入起点单元格（A1 记法，如 \"B2\"），默认 \"A1\"。mode=append 时忽略其行号、仅沿用其列。"
+            },
+            "mode": {
+              "type": "string",
+              "enum": [
+                "overwrite",
+                "append"
+              ],
+              "default": "overwrite",
+              "description": "overwrite（默认）：从 start_cell 起写「表头 + 数据」块；append：把数据追加到子表已有数据下方（默认不重复表头）。"
+            },
+            "header": {
+              "type": "boolean",
+              "description": "是否写一行列名表头。省略时按 mode 取默认:overwrite→true、append→false（避免在已有表头下重复）；显式给值可覆盖。"
+            },
+            "allow_overwrite": {
+              "type": "boolean",
+              "default": true,
+              "description": "为 false 时，若写入会落在非空单元格则拒写以保护原数据（返回 partial_success）。默认 true。"
+            },
+            "columns": {
+              "type": "array",
+              "minItems": 1,
+              "description": "列名字符串数组，顺序与 `data` 中每行取值一一对应。同一子表内列名不可重复。",
+              "items": {
+                "type": "string"
+              }
+            },
+            "data": {
+              "type": "array",
+              "description": "数据行；每行是一个数组，长度必须等于 `columns` 数。元素按 `dtypes` 推得的列类型取值（date 列写 ISO yyyy-mm-dd 字符串、number 列写数值、bool 列写布尔、其余写文本），null 表示空单元格。",
+              "items": {
+                "type": "array",
+                "items": {
+                  "type": [
+                    "string",
+                    "number",
+                    "boolean",
+                    "null"
+                  ],
+                  "description": "单元格值：date→ISO yyyy-mm-dd 字符串；number→数值（json.Number 精度保留）；bool→布尔；string→文本；null→空单元格。"
+                }
+              }
+            },
+            "dtypes": {
+              "type": "object",
+              "description": "可选。列名 → pandas dtype 字符串的映射；缺失项默认按 object（string + 文本格式 `@`）处理，所以省略整段时整张表按文本写入（导入 CSV-shaped 数据的最简形态）。dtype 解析规则：`int*` / `uint*` / `Int*` / `UInt*` / `float*` / `Float*` / `complex*` → number（精度保留），`bool` / `boolean` → bool，`datetime64[ns]` / 含时区的 `datetime64[ns, UTC]` 等 → date（默认 `yyyy-mm-dd` 格式），`object` / `string` / `category` / 未识别 → string + 文本格式 `@`（数字样字符串如「00123」不会塌缩成数字）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            },
+            "formats": {
+              "type": "object",
+              "description": "可选。列名 → Excel number_format 字符串的映射，覆盖 dtype 自带的默认格式（金额 `#,##0.00`、百分比 `0.0%`、自定义日期 `yyyy-mm` 等）。percent 列的数值尺度由调用方负责（0.0469 配 `0.00%` 显示 4.69%）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            }
+          }
+        }
+      },
+      "styles": {
+        "items": {
+          "properties": {
+            "cell_merges": {
+              "description": "单元格合并操作数组；range 使用 A1 单元格范围，merge_type 默认 all。",
+              "items": {
+                "properties": {
+                  "merge_type": {
+                    "enum": [
+                      "all",
+                      "rows",
+                      "columns"
+                    ],
+                    "type": "string"
+                  },
+                  "range": {
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "cell_styles": {
+              "description": "单元格样式操作数组；每项用 A1 单元格 range 指定范围，字段名与 +cells-set-style 对齐。",
+              "items": {
+                "properties": {
+                  "background_color": {
+                    "type": "string"
+                  },
+                  "border_styles": {
+                    "type": "object",
+                    "description": "边框配置，结构同 +cells-set-style --border-styles。",
+                    "properties": {
+                      "bottom": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "left": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "right": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "top": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      }
+                    }
+                  },
+                  "font_color": {
+                    "type": "string"
+                  },
+                  "font_line": {
+                    "enum": [
+                      "none",
+                      "underline",
+                      "line-through"
+                    ],
+                    "type": "string"
+                  },
+                  "font_size": {
+                    "type": "number"
+                  },
+                  "font_style": {
+                    "enum": [
+                      "normal",
+                      "italic"
+                    ],
+                    "type": "string"
+                  },
+                  "font_weight": {
+                    "enum": [
+                      "normal",
+                      "bold"
+                    ],
+                    "type": "string"
+                  },
+                  "horizontal_alignment": {
+                    "enum": [
+                      "left",
+                      "center",
+                      "right"
+                    ],
+                    "type": "string"
+                  },
+                  "number_format": {
+                    "type": "string"
+                  },
+                  "range": {
+                    "description": "A1 单元格范围，必须落在该子表本次写入区域内；例如 A1:B1、B2。",
+                    "type": "string"
+                  },
+                  "vertical_alignment": {
+                    "enum": [
+                      "top",
+                      "middle",
+                      "bottom"
+                    ],
+                    "type": "string"
+                  },
+                  "word_wrap": {
+                    "enum": [
+                      "overflow",
+                      "auto-wrap",
+                      "word-clip"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "col_sizes": {
+              "description": "列宽操作数组；range 使用列范围如 A:C，type 为 pixel/standard，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "name": {
+              "description": "子表名。--sheets 模式下必须与同位置 --sheets.sheets[].name 一致；--values 模式下建议写 Sheet1（其 name 会被忽略）。",
+              "type": "string"
+            },
+            "row_sizes": {
+              "description": "行高操作数组；range 使用行范围如 1:3，type 为 pixel/standard/auto，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard",
+                      "auto"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            }
+          },
+          "required": [
+            "name"
+          ],
+          "type": "object"
+        },
+        "type": "array"
+      }
+    },
+    "+workbook-create": {
+      "sheets": {
+        "type": "array",
+        "minItems": 1,
+        "description": "一个或多个子表的 typed 数据，每个数组元素写入一张子表；支持多 DataFrame → 多子表一次写入。整体形状对齐 pandas `df.to_json(orient=\"split\")`：列名走 `columns`、二维取值走 `data`、每列的 pandas dtype 走 `dtypes`、可选的展示格式走 `formats`。一行式用法：`{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`。",
+        "items": {
+          "type": "object",
+          "required": [
+            "name",
+            "columns",
+            "data"
+          ],
+          "properties": {
+            "name": {
+              "type": "string",
+              "description": "目标子表名。按名匹配已有子表；不存在则新建该子表。同一次调用内子表名不可重复。"
+            },
+            "start_cell": {
+              "type": "string",
+              "default": "A1",
+              "description": "写入起点单元格（A1 记法，如 \"B2\"），默认 \"A1\"。mode=append 时忽略其行号、仅沿用其列。"
+            },
+            "mode": {
+              "type": "string",
+              "enum": [
+                "overwrite",
+                "append"
+              ],
+              "default": "overwrite",
+              "description": "overwrite（默认）：从 start_cell 起写「表头 + 数据」块；append：把数据追加到子表已有数据下方（默认不重复表头）。"
+            },
+            "header": {
+              "type": "boolean",
+              "description": "是否写一行列名表头。省略时按 mode 取默认:overwrite→true、append→false（避免在已有表头下重复）；显式给值可覆盖。"
+            },
+            "allow_overwrite": {
+              "type": "boolean",
+              "default": true,
+              "description": "为 false 时，若写入会落在非空单元格则拒写以保护原数据（返回 partial_success）。默认 true。"
+            },
+            "columns": {
+              "type": "array",
+              "minItems": 1,
+              "description": "列名字符串数组，顺序与 `data` 中每行取值一一对应。同一子表内列名不可重复。",
+              "items": {
+                "type": "string"
+              }
+            },
+            "data": {
+              "type": "array",
+              "description": "数据行；每行是一个数组，长度必须等于 `columns` 数。元素按 `dtypes` 推得的列类型取值（date 列写 ISO yyyy-mm-dd 字符串、number 列写数值、bool 列写布尔、其余写文本），null 表示空单元格。",
+              "items": {
+                "type": "array",
+                "items": {
+                  "type": [
+                    "string",
+                    "number",
+                    "boolean",
+                    "null"
+                  ],
+                  "description": "单元格值：date→ISO yyyy-mm-dd 字符串；number→数值（json.Number 精度保留）；bool→布尔；string→文本；null→空单元格。"
+                }
+              }
+            },
+            "dtypes": {
+              "type": "object",
+              "description": "可选。列名 → pandas dtype 字符串的映射；缺失项默认按 object（string + 文本格式 `@`）处理，所以省略整段时整张表按文本写入（导入 CSV-shaped 数据的最简形态）。dtype 解析规则：`int*` / `uint*` / `Int*` / `UInt*` / `float*` / `Float*` / `complex*` → number（精度保留），`bool` / `boolean` → bool，`datetime64[ns]` / 含时区的 `datetime64[ns, UTC]` 等 → date（默认 `yyyy-mm-dd` 格式），`object` / `string` / `category` / 未识别 → string + 文本格式 `@`（数字样字符串如「00123」不会塌缩成数字）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            },
+            "formats": {
+              "type": "object",
+              "description": "可选。列名 → Excel number_format 字符串的映射，覆盖 dtype 自带的默认格式（金额 `#,##0.00`、百分比 `0.0%`、自定义日期 `yyyy-mm` 等）。percent 列的数值尺度由调用方负责（0.0469 配 `0.00%` 显示 4.69%）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            }
+          }
+        }
+      },
+      "styles": {
+        "items": {
+          "properties": {
+            "cell_merges": {
+              "description": "单元格合并操作数组；range 使用 A1 单元格范围，merge_type 默认 all。",
+              "items": {
+                "properties": {
+                  "merge_type": {
+                    "enum": [
+                      "all",
+                      "rows",
+                      "columns"
+                    ],
+                    "type": "string"
+                  },
+                  "range": {
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "cell_styles": {
+              "description": "单元格样式操作数组；每项用 A1 单元格 range 指定范围，字段名与 +cells-set-style 对齐。",
+              "items": {
+                "properties": {
+                  "background_color": {
+                    "type": "string"
+                  },
+                  "border_styles": {
+                    "type": "object",
+                    "description": "边框配置，结构同 +cells-set-style --border-styles。",
+                    "properties": {
+                      "bottom": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "left": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "right": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "top": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      }
+                    }
+                  },
+                  "font_color": {
+                    "type": "string"
+                  },
+                  "font_line": {
+                    "enum": [
+                      "none",
+                      "underline",
+                      "line-through"
+                    ],
+                    "type": "string"
+                  },
+                  "font_size": {
+                    "type": "number"
+                  },
+                  "font_style": {
+                    "enum": [
+                      "normal",
+                      "italic"
+                    ],
+                    "type": "string"
+                  },
+                  "font_weight": {
+                    "enum": [
+                      "normal",
+                      "bold"
+                    ],
+                    "type": "string"
+                  },
+                  "horizontal_alignment": {
+                    "enum": [
+                      "left",
+                      "center",
+                      "right"
+                    ],
+                    "type": "string"
+                  },
+                  "number_format": {
+                    "type": "string"
+                  },
+                  "range": {
+                    "description": "A1 单元格范围，必须落在该子表本次写入区域内；例如 A1:B1、B2。",
+                    "type": "string"
+                  },
+                  "vertical_alignment": {
+                    "enum": [
+                      "top",
+                      "middle",
+                      "bottom"
+                    ],
+                    "type": "string"
+                  },
+                  "word_wrap": {
+                    "enum": [
+                      "overflow",
+                      "auto-wrap",
+                      "word-clip"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "col_sizes": {
+              "description": "列宽操作数组；range 使用列范围如 A:C，type 为 pixel/standard，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "name": {
+              "description": "子表名。--sheets 模式下必须与同位置 --sheets.sheets[].name 一致；--values 模式下建议写 Sheet1（其 name 会被忽略）。",
+              "type": "string"
+            },
+            "row_sizes": {
+              "description": "行高操作数组；range 使用行范围如 1:3，type 为 pixel/standard/auto，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard",
+                      "auto"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            }
+          },
+          "required": [
+            "name"
+          ],
+          "type": "object"
+        },
+        "type": "array"
+      }
     }
   }
 }

shortcuts/sheets/execute_paths_test.go

@@ -47,19 +47,132 @@ func TestExecute_WorkbookInfo_ToolError(t *testing.T) {
 			"data": map[string]interface{}{},
 		},
 	}
-	stdout, stderr, err := func() (string, string, error) {
+	_, _, err := func() (string, string, error) {
 		parent, stdout, stderr, reg := newTestRig(t, WorkbookInfo)
 		reg.Register(stub)
 		parent.SetArgs([]string{"+workbook-info", "--url", testURL})
 		err := parent.Execute()
 		return stdout.String(), stderr.String(), err
 	}()
-	if err == nil {
-		t.Fatalf("expected non-zero code to surface as error; stdout=%s stderr=%s", stdout, stderr)
+	p := requireProblem(t, err, errs.CategoryAPI, errs.SubtypeServerError, "")
+	if !strings.Contains(p.Message, "1310201") && !strings.Contains(p.Message, "not found") {
+		t.Errorf("expected error code or message in problem; got message=%q", p.Message)
 	}
-	combined := stdout + stderr + err.Error()
-	if !strings.Contains(combined, "1310201") && !strings.Contains(combined, "not found") {
-		t.Errorf("expected error code in envelope; got=%s|%s|%v", stdout, stderr, err)
+}
+
+// TestExecute_WikiURLResolvesToSheet covers the two-step wiki path: a /wiki/
+// URL is resolved via get_node to its spreadsheet obj_token, which then feeds
+// the tool invoke. The tool stub is keyed on the resolved obj_token, so the
+// test would fail if the node_token were used unresolved.
+func TestExecute_WikiURLResolvesToSheet(t *testing.T) {
+	t.Parallel()
+	getNode := &httpmock.Stub{
+		Method: "GET",
+		URL:    "/open-apis/wiki/v2/spaces/get_node",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"node": map[string]interface{}{
+					"obj_type":  "sheet",
+					"obj_token": testToken,
+				},
+			},
+		},
+	}
+	tool := toolOutputStub(testToken, "read", `{"sheets":[{"sheet_id":"sh1","title":"Sheet1","index":0}]}`)
+	out, err := runShortcutWithStubs(t, WorkbookInfo,
+		[]string{"--url", "https://example.feishu.cn/wiki/wikTestNODE"}, getNode, tool)
+	if err != nil {
+		t.Fatalf("execute failed: %v\nout=%s", err, out)
+	}
+	data := decodeEnvelopeData(t, out)
+	if sheets, _ := data["sheets"].([]interface{}); len(sheets) != 1 {
+		t.Fatalf("sheets len = %d, want 1; out=%s", len(sheets), out)
+	}
+}
+
+// TestExecute_WikiURLWrongObjType rejects a wiki node that resolves to a
+// non-spreadsheet obj_type before any tool invoke.
+func TestExecute_WikiURLWrongObjType(t *testing.T) {
+	t.Parallel()
+	getNode := &httpmock.Stub{
+		Method: "GET",
+		URL:    "/open-apis/wiki/v2/spaces/get_node",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"node": map[string]interface{}{
+					"obj_type":  "docx",
+					"obj_token": "docABC",
+				},
+			},
+		},
+	}
+	_, err := runShortcutWithStubs(t, WorkbookInfo,
+		[]string{"--url", "https://example.feishu.cn/wiki/wikTestNODE"}, getNode)
+	requireValidation(t, err, "obj_type")
+}
+
+// TestExecute_WikiURLIncompleteNode treats an incomplete get_node response
+// (missing obj_type/obj_token) as an internal/server error, not a user --url
+// validation error.
+func TestExecute_WikiURLIncompleteNode(t *testing.T) {
+	t.Parallel()
+	getNode := &httpmock.Stub{
+		Method: "GET",
+		URL:    "/open-apis/wiki/v2/spaces/get_node",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"node": map[string]interface{}{},
+			},
+		},
+	}
+	_, err := runShortcutWithStubs(t, WorkbookInfo,
+		[]string{"--url", "https://example.feishu.cn/wiki/wikTestNODE"}, getNode)
+	if err == nil {
+		t.Fatal("want error for incomplete get_node node data")
+	}
+	var ve *errs.ValidationError
+	if errors.As(err, &ve) {
+		t.Fatalf("incomplete-data error classified as validation (%v); want internal", err)
+	}
+}
+
+// TestExecute_RangeMove_WikiURL guards the transformExecuteFn path: +range-move
+// and +range-copy use a named Execute helper (not an inline func), so they must
+// still resolve a /wiki/ URL to the backing spreadsheet token before calling
+// transform_range. The tool stub is keyed on the resolved obj_token, so an
+// unresolved node_token would miss it and fail this test.
+func TestExecute_RangeMove_WikiURL(t *testing.T) {
+	t.Parallel()
+	getNode := &httpmock.Stub{
+		Method: "GET",
+		URL:    "/open-apis/wiki/v2/spaces/get_node",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"node": map[string]interface{}{
+					"obj_type":  "sheet",
+					"obj_token": testToken,
+				},
+			},
+		},
+	}
+	tool := toolOutputStub(testToken, "write", `{"updated_range":"A10:B11"}`)
+	out, err := runShortcutWithStubs(t, RangeMove,
+		[]string{
+			"--url", "https://example.feishu.cn/wiki/wikTestNODE",
+			"--sheet-id", testSheetID,
+			"--source-range", "A1:B2",
+			"--target-range", "A10",
+		}, getNode, tool)
+	if err != nil {
+		t.Fatalf("execute failed: %v\nout=%s", err, out)
 	}
 }
 
@@ -365,14 +478,17 @@ func TestExecute_WorkbookCreate(t *testing.T) {
 			},
 		},
 	}
-	// Initial fill first reads the workbook structure to resolve the default
-	// sheet's id (the create response doesn't echo it), then writes.
+	// The write reads the workbook structure to resolve the default sheet's id
+	// (the create response doesn't echo it). lookupFirstSheetID and
+	// writeTypedSheets' listSheetIDsByName both read it — one reusable stub serves
+	// both. The synthesized sheet is named "Sheet1", matching the default sheet,
+	// so it's adopted in place (no rename).
 	structure := toolOutputStub("shtcnBRAND", "read", `{"sheets":[{"sheet_id":"shtFirst","sheet_name":"Sheet1","index":0}]}`)
+	structure.Reusable = true
 	fill := toolOutputStub("shtcnBRAND", "write", `{"updated_cells":4}`)
 	out, err := runShortcutWithStubs(t, WorkbookCreate, []string{
 		"--title", "Sales",
-		"--headers", `["Name","Score"]`,
-		"--values", `[["alice",95]]`,
+		"--values", `[["Name","Score"],["alice",95]]`,
 	}, create, structure, fill)
 	if err != nil {
 		t.Fatalf("execute failed: %v\nout=%s", err, out)
@@ -382,8 +498,8 @@ func TestExecute_WorkbookCreate(t *testing.T) {
 	if ss["spreadsheet_token"] != "shtcnBRAND" {
 		t.Errorf("spreadsheet_token = %v", ss["spreadsheet_token"])
 	}
-	if data["initial_fill"] == nil {
-		t.Errorf("initial_fill missing in envelope")
+	if sheets, _ := data["sheets"].([]interface{}); len(sheets) != 1 {
+		t.Errorf("sheets summary missing in envelope; got %#v", data["sheets"])
 	}
 	// The fill must target the resolved first sheet, not an empty selector.
 	fillInput := decodeToolInput(t, decodeRawEnvelopeBody(t, fill.CapturedBody), "set_cell_range")
@@ -393,14 +509,13 @@ func TestExecute_WorkbookCreate(t *testing.T) {
 }
 
 // TestExecute_WorkbookCreate_EmptyArraysSkipFill locks the fix for the nil-map
-// panic / illegal-range bug: --values '[]' or --headers '[]' must short-circuit
-// the initial fill (no structure/fill calls fire) and finish with the
-// spreadsheet created but no initial_fill — never panic on a nil fill map.
+// panic / illegal-range bug: --values '[]' must short-circuit the initial fill
+// (no structure/fill calls fire) and finish with the spreadsheet created but no
+// sheets summary — never panic on a nil payload.
 func TestExecute_WorkbookCreate_EmptyArraysSkipFill(t *testing.T) {
 	t.Parallel()
 	for _, tc := range []struct{ name, flag, val string }{
 		{"empty values", "--values", "[]"},
-		{"empty headers", "--headers", "[]"},
 	} {
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
@@ -421,8 +536,8 @@ func TestExecute_WorkbookCreate_EmptyArraysSkipFill(t *testing.T) {
 				t.Fatalf("execute failed: %v\nout=%s", err, out)
 			}
 			data := decodeEnvelopeData(t, out)
-			if data["initial_fill"] != nil {
-				t.Errorf("initial_fill should be absent for %s %s; got %#v", tc.flag, tc.val, data["initial_fill"])
+			if data["sheets"] != nil {
+				t.Errorf("sheets should be absent for %s %s; got %#v", tc.flag, tc.val, data["sheets"])
 			}
 			if ss, _ := data["spreadsheet"].(map[string]interface{}); ss["spreadsheet_token"] != "shtNEW" {
 				t.Errorf("spreadsheet_token = %v, want shtNEW", ss["spreadsheet_token"])

shortcuts/sheets/flag_defs.go

@@ -80,3 +80,28 @@ func flagsFor(command string) []common.Flag {
 	}
 	return out
 }
+
+// flagAcceptsStdin reports whether the (command, flag) pair declares stdin as
+// an input source in flag-defs.json. Used to decide whether an "invalid JSON"
+// error should also steer the caller toward stdin. It runs on an error path,
+// so it returns false for an unknown command/flag rather than panicking the
+// way flagsFor does.
+func flagAcceptsStdin(command, name string) bool {
+	defs, _ := loadFlagDefs()
+	spec, ok := defs[command]
+	if !ok {
+		return false
+	}
+	for _, df := range spec.Flags {
+		if df.Name != name {
+			continue
+		}
+		for _, in := range df.Input {
+			if in == common.Stdin {
+				return true
+			}
+		}
+		return false
+	}
+	return false
+}

shortcuts/sheets/flag_defs_gen.go

@@ -75,7 +75,7 @@ var flagDefs = map[string]commandDef{
 			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
 			{Name: "range", Kind: "own", Type: "string", Required: "required", Desc: "A1 range, e.g. `A1:F10` (no sheet prefix — use `--sheet-id` / `--sheet-name` to select the sheet)"},
 			{Name: "include", Kind: "own", Type: "string_slice", Required: "optional", Desc: "Comma-separated info categories to include", Enum: []string{"value", "formula", "style", "comment", "data_validation"}},
-			{Name: "max-chars", Kind: "own", Type: "int", Required: "optional", Desc: "Safety cap; default 200000", Default: "200000", Hidden: true},
+			{Name: "max-chars", Kind: "own", Type: "int", Required: "optional", Desc: "Max output chars per call; default 500000 (safety cap). Large reads are usually better redirected to a file; only lower it (e.g. 25000) when you want results inline without triggering file offload, paging via has_more", Default: "500000"},
 			{Name: "skip-hidden", Kind: "own", Type: "bool", Required: "optional", Desc: "Skip hidden rows and columns; default `false`"},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
@@ -305,10 +305,9 @@ var flagDefs = map[string]commandDef{
 			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
 			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
 			{Name: "range", Kind: "own", Type: "string", Required: "required", Desc: "A1 range, e.g. `A1:F30` (no sheet prefix — use `--sheet-id` / `--sheet-name` to select the sheet)"},
-			{Name: "max-chars", Kind: "own", Type: "int", Required: "optional", Desc: "Safety cap; default 200000", Default: "200000", Hidden: true},
+			{Name: "max-chars", Kind: "own", Type: "int", Required: "optional", Desc: "Max output chars per call; default 500000 (safety cap). Large reads are usually better redirected to a file; only lower it (e.g. 25000) when you want results inline without triggering file offload, paging via has_more", Default: "500000"},
 			{Name: "include-row-prefix", Kind: "own", Type: "bool", Required: "optional", Desc: "Whether to prefix each row with `[row=N]`; default `true`", Default: "true"},
 			{Name: "skip-hidden", Kind: "own", Type: "bool", Required: "optional", Desc: "Skip hidden rows and columns; default `false`"},
-			{Name: "rows-json", Kind: "own", Type: "bool", Required: "optional", Desc: "Return structured rows ({row_number, values:{col→cell}}) instead of CSV text; default false", Default: "false"},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional", Desc: "Print the request path and parameters without executing"},
 		},
 	},
@@ -320,7 +319,7 @@ var flagDefs = map[string]commandDef{
 			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
 			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
 			{Name: "start-cell", Kind: "own", Type: "string", Required: "required", Desc: "Top-left A1 anchor (e.g. `A1`, `B5`; no sheet prefix — use `--sheet-id` / `--sheet-name` to select the sheet); must be a single cell, range notation not accepted; the bottom-right is inferred from CSV row/column counts", Default: "A1"},
-			{Name: "csv", Kind: "own", Type: "string", Required: "required", Desc: "RFC 4180 CSV text; plain values only (no formulas / styles / comments)", Input: []string{"file", "stdin"}},
+			{Name: "csv", Kind: "own", Type: "string", Required: "required", Desc: "RFC 4180 CSV text; values or formulas (a leading = is evaluated as a formula); no styles / comments / images (use +cells-set for those).", Input: []string{"file", "stdin"}},
 			{Name: "allow-overwrite", Kind: "own", Type: "bool", Required: "optional", Desc: "Allow overwriting (default true); set false to error if any target cell is non-empty", Default: "true"},
 			{Name: "range", Kind: "own", Type: "string", Required: "optional", Desc: "alias for --start-cell (parity with +csv-get / +cells-set, which locate with --range); a range like A1:H17 collapses to its top-left cell", Hidden: true},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
@@ -766,7 +765,7 @@ var flagDefs = map[string]commandDef{
 			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
 			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
 			{Name: "title", Kind: "own", Type: "string", Required: "required", Desc: "New sheet title"},
-			{Name: "index", Kind: "own", Type: "int", Required: "optional", Desc: "Insert position; appended to the end when omitted", Default: "-1"},
+			{Name: "index", Kind: "own", Type: "int", Required: "optional", Desc: "Insert position (0-based); appended to the end when omitted", Default: "-1"},
 			{Name: "row-count", Kind: "own", Type: "int", Required: "optional", Desc: "Initial row count (default 200, max 50000)", Default: "200"},
 			{Name: "col-count", Kind: "own", Type: "int", Required: "optional", Desc: "Initial column count (default 20, max 200)", Default: "20"},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
@@ -793,6 +792,16 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+sheet-hide-gridline": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
+			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+sheet-info": {
 		Risk: "read",
 		Flags: []flagDef{
@@ -839,6 +848,16 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+sheet-show-gridline": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
+			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+sheet-unhide": {
 		Risk: "write",
 		Flags: []flagDef{
@@ -895,13 +914,39 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+table-get": {
+		Risk: "read",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "sheet-id", Kind: "own", Type: "string", Required: "optional", Desc: "Read only this sheet (by id); omit to read all sheets"},
+			{Name: "sheet-name", Kind: "own", Type: "string", Required: "optional", Desc: "Read only this sheet (by name); omit to read all sheets"},
+			{Name: "range", Kind: "own", Type: "string", Required: "optional", Desc: "A1 range to read; omit to read each sheet's full used range (spans internal blank rows/columns, not just the A1 current region)"},
+			{Name: "no-header", Kind: "own", Type: "bool", Required: "optional", Desc: "Treat the first row as data instead of a header (columns get positional names col1, col2, ...)"},
+			{Name: "dataframe-out", Kind: "own", Type: "string", Required: "optional", Desc: "Write the typed table as one Arrow IPC file (Feather v2) instead of the default JSON. Pass `@<path>` for a file or `-` for binary stdout (same convention as other binary I/O flags). Mirror of the input-side `--dataframe` on `+table-put` / `+workbook-create` — pandas users round-trip via `df = pd.read_feather(\"x.arrow\")` or `pd.read_feather(io.BytesIO(stdout))`. Single-sheet only: requires `--sheet-id` or `--sheet-name`; whole-workbook reads keep the default JSON path. Column types come from the typed read-back (string/number/date/bool); per-column `number_format` is preserved as Arrow field metadata so the Arrow file can round-trip straight back through `+table-put --dataframe`."},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
+	"+table-put": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL to write into (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token to write into (XOR with `--url`)"},
+			{Name: "sheets", Kind: "own", Type: "string", Required: "xor", Desc: "Typed table payload (pandas-DataFrame-shaped) as JSON, XOR with `--dataframe`: a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it with `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. `dtypes` values are pandas dtype strings (`int64`, `float64`, `Int64`, `bool`, `boolean`, `datetime64[ns]`, `object`, ...); the writer maps them to internal string/number/date/bool — omit `dtypes` and a column writes as text (good for raw CSV-shaped data). `formats[col]` is an Excel number_format string (e.g. `#,##0.00`, `0.0%`, `yyyy-mm`); when absent, date columns default to `yyyy-mm-dd` and string columns to text format (`@`).", Input: []string{"file", "stdin"}},
+			{Name: "dataframe", Kind: "own", Type: "string", Required: "xor", Desc: "Single-sheet typed table from one Arrow IPC file (a.k.a. Feather v2 — what `pandas.DataFrame.to_feather()` writes), XOR with `--sheets`. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema (int*/uint*/float* → number, date32/date64/timestamp → date, utf8/large_utf8 → string, bool → bool); per-column `number_format` may be set via Arrow field metadata (`pa.field(\"price\", pa.float64(), metadata={b\"number_format\": b\"$#,##0.00\"})`). Writes the sheet at default placement: name `Sheet1` (created when absent), overwrite from A1 with header. For a different sheet name, anchor, mode, or to write multiple sheets, use `--sheets` instead."},
+			{Name: "styles", Kind: "own", Type: "string", Required: "optional", Desc: "Visual operations applied after the typed write, as JSON: top-level `{styles:[...]}`. Each item corresponds to one written sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. The styles array length/order/name must match the written sheets: with --sheets, match --sheets.sheets; with --dataframe (single sheet named Sheet1), pass exactly one styles item with name `Sheet1`. Run `+table-put --print-schema --flag-name styles` for the full cell_styles field schema.", Input: []string{"file", "stdin"}},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+workbook-create": {
 		Risk: "write",
 		Flags: []flagDef{
 			{Name: "title", Kind: "own", Type: "string", Required: "required", Desc: "Spreadsheet title"},
 			{Name: "folder-token", Kind: "own", Type: "string", Required: "optional", Desc: "Target folder token; placed at the drive root when omitted"},
-			{Name: "headers", Kind: "own", Type: "string", Required: "optional", Desc: "Header row as a JSON array: `[\"Col A\",\"Col B\"]`", Input: []string{"file", "stdin"}},
-			{Name: "values", Kind: "own", Type: "string", Required: "optional", Desc: "Initial data as a 2D JSON array: `[[\"alice\",95]]`", Input: []string{"file", "stdin"}},
+			{Name: "values", Kind: "own", Type: "string", Required: "optional", Desc: "Untyped initial data as one 2D JSON array (`[[\"alice\",95]]`); values are written as-is with their type auto-detected, through the same batched set_cell_range path as --sheets — pair with --styles for number formats, colors, merges, and row/col sizes", Input: []string{"file", "stdin"}},
+			{Name: "sheets", Kind: "own", Type: "string", Required: "optional", Desc: "Typed table payload as JSON (same shape as `+table-put`): a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it from a DataFrame via `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. Mutually exclusive with --values and --dataframe. Creates the workbook, then writes typed type-faithful data (dates land as real dates, numbers keep precision).", Input: []string{"file", "stdin"}},
+			{Name: "styles", Kind: "own", Type: "string", Required: "optional", Desc: "Initial visual operations as JSON: top-level `{styles:[...]}`. Each item corresponds to one target sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. With --sheets, styles array length/order/name must match --sheets.sheets. With --values, pass exactly one styles item for the initial sheet (its name is ignored).", Input: []string{"file", "stdin"}},
+			{Name: "dataframe", Kind: "own", Type: "string", Required: "optional", Desc: "Single-sheet typed table from one Arrow IPC file (Feather v2 — what `pandas.DataFrame.to_feather()` writes), mutually exclusive with --values and --sheets. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema; per-column `number_format` may be set via Arrow field metadata. Creates the workbook and fills its default sheet (`Sheet1` — adopted in place, no empty Sheet1 left behind). For multi-sheet or non-default placement, use `--sheets` instead."},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
@@ -912,10 +957,18 @@ var flagDefs = map[string]commandDef{
 			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
 			{Name: "file-extension", Kind: "own", Type: "string", Required: "optional", Desc: "Export file format; `csv` mode requires `--sheet-id`", Default: "xlsx", Enum: []string{"xlsx", "csv"}},
 			{Name: "sheet-id", Kind: "own", Type: "string", Required: "optional", Desc: "Required only in csv mode: which sheet to export as CSV. This is a `+workbook-export`-specific flag, unrelated to the common four-tuple sheet locator (this shortcut does not accept the common sheet locator)"},
-			{Name: "output-path", Kind: "own", Type: "string", Required: "optional", Desc: "Local save path; export is triggered but not downloaded when omitted"},
+			{Name: "output-path", Kind: "own", Type: "string", Required: "optional", Desc: "Local save path. When omitted, **only the export task is triggered + polled, the file is NOT downloaded** (returns file_token / status so a later step can resume the download). Pass a concrete path (e.g. `./out.xlsx`) or a directory (`.` keeps the server-provided filename) to download. Note: the equivalent `lark-cli drive +export --doc-type sheet` uses three separate flags (`--output-dir` / `--file-name` / `--overwrite`) and defaults to downloading into the current directory; this wrapper collapses them into a single `--output-path` for ergonomics but defaults to no-download — fall back to `drive +export` if the split flag set fits better."},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+workbook-import": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "file", Kind: "own", Type: "string", Required: "required", Desc: "Local file path (.xlsx / .xls / .csv)"},
+			{Name: "folder-token", Kind: "own", Type: "string", Required: "optional", Desc: "Target folder token; imported to the cloud drive root when omitted"},
+			{Name: "name", Kind: "own", Type: "string", Required: "optional", Desc: "Imported spreadsheet name; defaults to the local file name without its extension"},
+		},
+	},
 	"+workbook-info": {
 		Risk: "read",
 		Flags: []flagDef{

shortcuts/sheets/flag_defs_test.go

@@ -65,9 +65,9 @@ func TestFlagsFor_MapsAllFields(t *testing.T) {
 	if url == nil || url.Required {
 		t.Errorf("+sheet-create --url should not be cobra-required: %+v", url)
 	}
-	// hidden + int default
+	// visible + int default
 	cap := byName("+cells-get", "max-chars")
-	if cap == nil || !cap.Hidden || cap.Default != "200000" {
+	if cap == nil || cap.Hidden || cap.Default != "500000" {
 		t.Errorf("+cells-get --max-chars not mapped: %+v", cap)
 	}
 	// input sources
@@ -140,3 +140,24 @@ func TestFlagsFor_EveryRegisteredCommandHasDefs(t *testing.T) {
 		}
 	}
 }
+
+// TestFlagAcceptsStdin verifies the stdin-capability probe that decides whether
+// an "invalid JSON" error should also steer the caller toward stdin: a composite
+// flag (cells) accepts stdin, a plain locator (spreadsheet-token) does not, and
+// an unknown command/flag returns false without panicking (it runs on an error
+// path, unlike flagsFor).
+func TestFlagAcceptsStdin(t *testing.T) {
+	t.Parallel()
+	if !flagAcceptsStdin("+cells-set", "cells") {
+		t.Error("+cells-set --cells should accept stdin")
+	}
+	if flagAcceptsStdin("+cells-set", "spreadsheet-token") {
+		t.Error("--spreadsheet-token should not accept stdin")
+	}
+	if flagAcceptsStdin("+nope", "cells") {
+		t.Error("unknown command should be false (and must not panic)")
+	}
+	if flagAcceptsStdin("+cells-set", "nope") {
+		t.Error("unknown flag should be false")
+	}
+}

shortcuts/sheets/flag_schema.go

@@ -9,6 +9,8 @@ import (
 	"fmt"
 	"sort"
 	"sync"
+
+	"github.com/larksuite/cli/errs"
 )
 
 // ─── --print-schema runtime introspection ─────────────────────────────
@@ -91,7 +93,7 @@ func printFlagSchemaFor(command string) func(flagName string) ([]byte, error) {
 		}
 		entry, ok := idx.Flags[command]
 		if !ok || len(entry) == 0 {
-			return nil, fmt.Errorf("no JSON Schema registered for %s", command)
+			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "no JSON Schema registered for %s", command)
 		}
 		if flagName == "" {
 			flags := make([]string, 0, len(entry))
@@ -112,7 +114,9 @@ func printFlagSchemaFor(command string) func(flagName string) ([]byte, error) {
 				flags = append(flags, f)
 			}
 			sort.Strings(flags)
-			return nil, fmt.Errorf("no JSON Schema registered for %s --%s; available: %v", command, flagName, flags)
+			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"no JSON Schema registered for %s --%s; available: %v", command, flagName, flags).
+				WithParam("--flag-name")
 		}
 		// Reformat for readability — schema files store compact JSON.
 		var pretty interface{}

shortcuts/sheets/flag_schema_test.go

@@ -84,12 +84,12 @@ func TestPrintFlagSchema_NamedFlagReturnsSchemaSubtree(t *testing.T) {
 func TestPrintFlagSchema_UnknownFlagListsAvailable(t *testing.T) {
 	t.Parallel()
 	_, err := printFlagSchemaFor("+chart-create")("does-not-exist")
-	if err == nil {
-		t.Fatal("expected error for unknown flag, got nil")
+	ve := requireValidation(t, err, "+chart-create")
+	if !strings.Contains(ve.Message, "properties") {
+		t.Errorf("message should list available flags; got %q", ve.Message)
 	}
-	msg := err.Error()
-	if !strings.Contains(msg, "+chart-create") || !strings.Contains(msg, "properties") {
-		t.Errorf("error should mention shortcut + available flags; got %q", msg)
+	if ve.Param != "--flag-name" {
+		t.Errorf("param = %q, want --flag-name", ve.Param)
 	}
 }
 

shortcuts/sheets/flag_schema_validate.go

@@ -63,6 +63,7 @@ func validateParsedJSONFlag(fv flagView, name string, value interface{}) error {
 var parseJSONFlagSkip = map[string]struct{}{
 	"properties": {},
 	"operations": {},
+	"styles":     {},
 }
 
 // validateValueAgainstSchema is the (command, flag) → schema → check
@@ -93,7 +94,17 @@ func validateValueAgainstSchema(fv flagView, name string, value interface{}) err
 	var schema schemaProperty
 	json.Unmarshal(raw, &schema)
 	if vErr := validateAgainstSchema(value, &schema, ""); vErr != nil {
-		return sheetsValidationForFlag(name, "--%s: %s", name, vErr.Error())
+		// Composite-JSON shape errors (e.g. +cells-set --cells, chart
+		// --properties) are the highest-frequency usage-layer failure for
+		// sheets, and agents often burn several retries guessing the shape.
+		// Point them straight at --print-schema, which dumps the exact JSON
+		// Schema for this (command, flag) pair. The hint is always actionable:
+		// reaching this branch means entry[name] resolved a schema from the
+		// embedded index, and --print-schema reads that same index, so the
+		// suggested command is guaranteed to print it.
+		return sheetsValidationForFlag(name,
+			"--%s: %s; run `lark-cli sheets %s --print-schema --flag-name %s` to see the expected JSON Schema",
+			name, vErr.Error(), command, name)
 	}
 	return nil
 }

shortcuts/sheets/flag_schema_validate_test.go

@@ -478,11 +478,9 @@ func TestValidateInputAgainstSchema_RealSchema(t *testing.T) {
 		},
 	}
 	err := validateInputAgainstSchema(fv, bad)
-	if err == nil {
-		t.Fatal("expected enum violation, got nil")
-	}
-	if !strings.Contains(err.Error(), "summarize_by") || !strings.Contains(err.Error(), "not in enum") {
-		t.Errorf("error = %q, want summarize_by + enum hint", err.Error())
+	ve := requireValidation(t, err, "summarize_by")
+	if !strings.Contains(ve.Message, "not in enum") {
+		t.Errorf("error = %q, want enum hint", ve.Message)
 	}
 }
 
@@ -499,11 +497,9 @@ func TestValidateInputAgainstSchema_RealMinItems(t *testing.T) {
 		},
 	}
 	err := validateInputAgainstSchema(fv, bad)
-	if err == nil {
-		t.Fatal("expected minItems violation for empty values, got nil")
-	}
-	if !strings.Contains(err.Error(), "values") || !strings.Contains(err.Error(), "minimum is 1") {
-		t.Errorf("error = %q, want values + minimum-is-1 hint", err.Error())
+	ve := requireValidation(t, err, "values")
+	if !strings.Contains(ve.Message, "minimum is 1") {
+		t.Errorf("error = %q, want minimum-is-1 hint", ve.Message)
 	}
 }
 
@@ -520,11 +516,9 @@ func TestValidateInputAgainstSchema_RealMinimum(t *testing.T) {
 		},
 	}
 	err := validateInputAgainstSchema(fv, bad)
-	if err == nil {
-		t.Fatal("expected minimum violation for row:-1, got nil")
-	}
-	if !strings.Contains(err.Error(), "row") || !strings.Contains(err.Error(), "below minimum") {
-		t.Errorf("error = %q, want row + below-minimum hint", err.Error())
+	ve := requireValidation(t, err, "row")
+	if !strings.Contains(ve.Message, "below minimum") {
+		t.Errorf("error = %q, want below-minimum hint", ve.Message)
 	}
 }
 
@@ -554,11 +548,9 @@ func TestValidateInputAgainstSchema_RealAdditionalProperties(t *testing.T) {
 		},
 	}
 	err := validateInputAgainstSchema(fv, bad)
-	if err == nil {
-		t.Fatal("expected additionalProperties violation, got nil")
-	}
-	if !strings.Contains(err.Error(), "collapse") || !strings.Contains(err.Error(), `expected type "string"`) {
-		t.Errorf("error = %q, want collapse + string-type hint", err.Error())
+	ve := requireValidation(t, err, "collapse")
+	if !strings.Contains(ve.Message, `expected type "string"`) {
+		t.Errorf("error = %q, want string-type hint", ve.Message)
 	}
 }
 
@@ -587,3 +579,24 @@ func TestValidateInputAgainstSchema_SkipOperations(t *testing.T) {
 		t.Errorf("operations should be skipped; got %v", err)
 	}
 }
+
+// TestValidateValueAgainstSchema_PrintSchemaHint pins the highest-value
+// recovery affordance for composite-JSON flags: when the shape is wrong, the
+// error must point the agent straight at --print-schema (with the right
+// command + flag) instead of leaving it to guess across retries. +cells-set
+// --cells expects a 2-D array; a bare string trips the top-level type check.
+func TestValidateValueAgainstSchema_PrintSchemaHint(t *testing.T) {
+	t.Parallel()
+	fv := mapFlagView{command: "+cells-set"}
+	err := validateValueAgainstSchema(fv, "cells", "not-an-array")
+	// Underlying shape error is preserved (substring callers still match).
+	ve := requireValidation(t, err, `expected type "array"`)
+	// And the actionable --print-schema hint is appended with the exact
+	// command + flag, so a copy-paste fetches the schema for this pair.
+	if !strings.Contains(ve.Message, "lark-cli sheets +cells-set --print-schema --flag-name cells") {
+		t.Errorf("want --print-schema hint with command+flag; got %q", ve.Message)
+	}
+	if ve.Param != "--cells" {
+		t.Errorf("param = %q, want --cells", ve.Param)
+	}
+}

shortcuts/sheets/flag_schemas_gen.go

@@ -32,4 +32,6 @@ var commandsWithSchema = map[string]struct{}{
 	"+range-sort":            {},
 	"+sparkline-create":      {},
 	"+sparkline-update":      {},
+	"+table-put":             {},
+	"+workbook-create":       {},
 }

shortcuts/sheets/helpers.go

@@ -10,6 +10,8 @@ package sheets
 import (
 	"context"
 	"encoding/json"
+	"fmt"
+	neturl "net/url"
 	"strings"
 
 	"github.com/larksuite/cli/errs"
@@ -48,46 +50,151 @@ func sheetsInputStatError(flag string, err error) error {
 	return wrapped
 }
 
-// resolveSpreadsheetToken applies the public --url / --spreadsheet-token XOR
-// pair shared by every sheets canonical shortcut and returns the resolved
-// token. Network-free, safe to call from Validate and DryRun.
-func resolveSpreadsheetToken(runtime *common.RuntimeContext) (string, error) {
+// spreadsheetRef classification: a --url / --spreadsheet-token input names a
+// spreadsheet either directly (a /sheets/ URL or raw token) or indirectly via a
+// wiki node that must be resolved to its backing spreadsheet at Execute time.
+const (
+	spreadsheetRefSheet = "sheet"
+	spreadsheetRefWiki  = "wiki"
+)
+
+// spreadsheetRef is a parsed --url / --spreadsheet-token input. A wiki ref holds
+// the still-unresolved wiki node_token; resolveSpreadsheetTokenExec turns it
+// into the real spreadsheet token at Execute time.
+type spreadsheetRef struct {
+	Kind  string // spreadsheetRefSheet | spreadsheetRefWiki
+	Token string
+}
+
+// parseSpreadsheetRef applies the public --url / --spreadsheet-token XOR pair and
+// classifies the input. Network-free, safe to call from Validate and DryRun.
+//
+// Recognized --url shapes:
+//   - https://.../sheets/<token>        → {sheet, token}
+//   - https://.../spreadsheets/<token>  → {sheet, token}
+//   - https://.../wiki/<node_token>     → {wiki, node_token}  (resolved at Execute)
+//
+// A raw --spreadsheet-token is always treated as a spreadsheet token; wiki nodes
+// only ever arrive as a /wiki/ URL.
+func parseSpreadsheetRef(runtime *common.RuntimeContext) (spreadsheetRef, error) {
 	if err := common.ExactlyOneTyped(runtime, "url", "spreadsheet-token"); err != nil {
-		return "", err
+		return spreadsheetRef{}, err
 	}
 	if token := strings.TrimSpace(runtime.Str("spreadsheet-token")); token != "" {
 		if err := validate.RejectControlChars(token, "spreadsheet-token"); err != nil {
-			return "", sheetsValidationCauseForFlag("spreadsheet-token", err)
+			return spreadsheetRef{}, sheetsValidationCauseForFlag("spreadsheet-token", err)
 		}
-		return token, nil
+		return spreadsheetRef{Kind: spreadsheetRefSheet, Token: token}, nil
 	}
 
-	url := strings.TrimSpace(runtime.Str("url"))
-	token := extractSpreadsheetToken(url)
-	if token == "" || token == url {
-		return "", sheetsValidationForFlag("url", "--url must be a spreadsheet URL like https://.../sheets/<token>")
+	rawURL := strings.TrimSpace(runtime.Str("url"))
+	token, kind, ok := spreadsheetURLToken(rawURL)
+	if !ok {
+		return spreadsheetRef{}, sheetsValidationForFlag("url", "--url must be a spreadsheet URL like https://.../sheets/<token> or a wiki URL like https://.../wiki/<token>")
 	}
 	if err := validate.RejectControlChars(token, "url"); err != nil {
-		return "", sheetsValidationCauseForFlag("url", err)
+		return spreadsheetRef{}, sheetsValidationCauseForFlag("url", err)
 	}
-	return token, nil
+	return spreadsheetRef{Kind: kind, Token: token}, nil
 }
 
-// extractSpreadsheetToken pulls the token segment out of a /sheets/<token>
-// or /spreadsheets/<token> URL. Returns the input unchanged when no known
-// prefix is present (callers must check token != originalInput).
-func extractSpreadsheetToken(input string) string {
-	input = strings.TrimSpace(input)
-	for _, prefix := range []string{"/sheets/", "/spreadsheets/"} {
-		if idx := strings.Index(input, prefix); idx >= 0 {
-			token := input[idx+len(prefix):]
-			if idx2 := strings.IndexAny(token, "/?#"); idx2 >= 0 {
-				token = token[:idx2]
-			}
-			return token
+// spreadsheetURLToken extracts the token and its kind from a Lark URL, matching
+// only on the URL *path* segment (parsed via net/url). A /wiki/ or /sheets/ that
+// appears only in the query or fragment (e.g. a redirect or anchor param) never
+// hijacks classification. Returns ok=false when no known prefix heads the path.
+func spreadsheetURLToken(rawURL string) (token, kind string, ok bool) {
+	u, err := neturl.Parse(rawURL)
+	if err != nil || u.Path == "" {
+		return "", "", false
+	}
+	for _, m := range []struct {
+		prefix string
+		kind   string
+	}{
+		{"/sheets/", spreadsheetRefSheet},
+		{"/spreadsheets/", spreadsheetRefSheet},
+		{"/wiki/", spreadsheetRefWiki},
+	} {
+		if seg, found := pathSegmentAfter(u.Path, m.prefix); found {
+			return seg, m.kind, true
 		}
 	}
-	return input
+	return "", "", false
+}
+
+// pathSegmentAfter returns the first path segment after prefix when path begins
+// with prefix, else ("", false).
+func pathSegmentAfter(path, prefix string) (string, bool) {
+	if !strings.HasPrefix(path, prefix) {
+		return "", false
+	}
+	rest := path[len(prefix):]
+	if i := strings.IndexByte(rest, '/'); i >= 0 {
+		rest = rest[:i]
+	}
+	rest = strings.TrimSpace(rest)
+	if rest == "" {
+		return "", false
+	}
+	return rest, true
+}
+
+// resolveSpreadsheetToken applies the public --url / --spreadsheet-token XOR pair
+// and returns the resolved token. Network-free, safe to call from Validate and
+// DryRun.
+//
+// A /wiki/ URL yields the still-unresolved wiki node_token: turning it into the
+// backing spreadsheet token needs a get_node call, which only Execute may make.
+// Validate/DryRun only need a non-empty, control-char-clean token, so the
+// node_token passes through unchanged here; Execute paths call
+// resolveSpreadsheetTokenExec instead.
+func resolveSpreadsheetToken(runtime *common.RuntimeContext) (string, error) {
+	ref, err := parseSpreadsheetRef(runtime)
+	if err != nil {
+		return "", err
+	}
+	return ref.Token, nil
+}
+
+// resolveSpreadsheetTokenExec is the Execute-time counterpart of
+// resolveSpreadsheetToken: it additionally resolves a /wiki/ URL's node_token to
+// the backing spreadsheet token via wiki get_node, verifying obj_type=sheet.
+// Non-wiki inputs make no API call. Use this from every sheets Execute hook and
+// keep resolveSpreadsheetToken in Validate/DryRun so those stay network-free.
+func resolveSpreadsheetTokenExec(runtime *common.RuntimeContext) (string, error) {
+	ref, err := parseSpreadsheetRef(runtime)
+	if err != nil {
+		return "", err
+	}
+	if ref.Kind != spreadsheetRefWiki {
+		return ref.Token, nil
+	}
+	return resolveWikiNodeToSpreadsheetToken(runtime, ref.Token)
+}
+
+// resolveWikiNodeToSpreadsheetToken resolves a wiki node_token to the spreadsheet
+// obj_token it points at, erroring when the node is not a spreadsheet. The
+// wiki:node:read scope is only needed on this path, so it is enforced here rather
+// than declared unconditionally on every sheets shortcut.
+func resolveWikiNodeToSpreadsheetToken(runtime *common.RuntimeContext, nodeToken string) (string, error) {
+	if err := runtime.EnsureScopes([]string{"wiki:node:read"}); err != nil {
+		return "", err
+	}
+	data, err := runtime.CallAPITyped("GET", "/open-apis/wiki/v2/spaces/get_node",
+		map[string]interface{}{"token": nodeToken}, nil)
+	if err != nil {
+		return "", err
+	}
+	node := common.GetMap(data, "node")
+	objType := common.GetString(node, "obj_type")
+	objToken := common.GetString(node, "obj_token")
+	if objType == "" || objToken == "" {
+		return "", errs.NewInternalError(errs.SubtypeInvalidResponse, "wiki get_node returned incomplete node data for %q", nodeToken)
+	}
+	if objType != "sheet" {
+		return "", sheetsValidationForFlag("url", "wiki URL resolves to obj_type=%q, but a spreadsheet (obj_type=sheet) is required", objType)
+	}
+	return objToken, nil
 }
 
 // resolveSheetSelector validates the --sheet-id / --sheet-name XOR and
@@ -241,6 +348,16 @@ func parseJSONFlag(runtime flagView, name string) (interface{}, error) {
 	}
 	var out interface{}
 	if err := json.Unmarshal([]byte(raw), &out); err != nil {
+		// Composite payloads that embed formulas / quotes / commas are the
+		// classic source of this error: inlined into the shell, the JSON gets
+		// mangled (e.g. `\$` → "invalid character in string escape"). For any
+		// flag that accepts stdin, steer the caller there — passing the payload
+		// via `--<flag> - < file` sidesteps shell escaping entirely.
+		if flagAcceptsStdin(runtime.Command(), name) {
+			return nil, sheetsValidationForFlag(name,
+				"--%s: invalid JSON: %v; if the payload contains formulas / quotes / commas, pass it via stdin (`--%s - < file`) so the shell doesn't mangle the JSON",
+				name, err, name).WithCause(err)
+		}
 		return nil, sheetsValidationForFlag(name, "--%s: invalid JSON: %v", name, err).WithCause(err)
 	}
 	// Schema-driven flag validation at the user-input boundary. Skips
@@ -325,6 +442,72 @@ func buildCellStyleFromFlags(runtime flagView) map[string]interface{} {
 	return style
 }
 
+// cellStyleAliases maps shorthand cell_styles field names that models commonly
+// hallucinate (Excel / openpyxl / CSS conventions) onto the canonical field
+// names the backend expects. Only the unambiguous alignment shorthands are
+// aliased — they are the high-frequency miss; ambiguous guesses (e.g. "color",
+// "bg_color", "text_align") are intentionally left out so a wrong guess still
+// surfaces as an error rather than being silently reinterpreted.
+var cellStyleAliases = []struct{ alias, canonical string }{
+	{"horizontal_align", "horizontal_alignment"},
+	{"halign", "horizontal_alignment"},
+	{"vertical_align", "vertical_alignment"},
+	{"valign", "vertical_alignment"},
+}
+
+// normalizeCellStyleAliases renames known shorthand keys in a single
+// cell_styles map to their canonical equivalents, in place, so a model that
+// writes e.g. "horizontal_align" instead of "horizontal_alignment" still
+// applies the style instead of hitting an "unsupported field" error (--styles)
+// or having the field silently dropped by the backend (typed --cells). If both
+// the shorthand and its canonical key are present it returns a validation error
+// rather than picking one. path labels the map for the error message.
+func normalizeCellStyleAliases(style map[string]interface{}, path string) error {
+	if len(style) == 0 {
+		return nil
+	}
+	for _, a := range cellStyleAliases {
+		v, ok := style[a.alias]
+		if !ok {
+			continue
+		}
+		if _, exists := style[a.canonical]; exists {
+			return common.ValidationErrorf("%s.%s conflicts with %s; pass only %s", path, a.alias, a.canonical, a.canonical)
+		}
+		style[a.canonical] = v
+		delete(style, a.alias)
+	}
+	return nil
+}
+
+// normalizeTypedCellsStyleAliases walks a typed --cells 2D array and applies
+// normalizeCellStyleAliases to every cell's inline cell_styles object, so the
+// alignment shorthands are accepted on +cells-set the same as on --styles.
+// Structure is checked leniently to match the pass-through contract: any
+// element that isn't the expected shape is skipped, not rejected.
+func normalizeTypedCellsStyleAliases(cells []interface{}, path string) error {
+	for r, rowRaw := range cells {
+		row, ok := rowRaw.([]interface{})
+		if !ok {
+			continue
+		}
+		for c, cellRaw := range row {
+			cell, ok := cellRaw.(map[string]interface{})
+			if !ok {
+				continue
+			}
+			st, ok := cell["cell_styles"].(map[string]interface{})
+			if !ok {
+				continue
+			}
+			if err := normalizeCellStyleAliases(st, fmt.Sprintf("%s[%d][%d].cell_styles", path, r, c)); err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
 // borderStylesFromFlag parses --border-styles as a JSON object (top/bottom/
 // left/right with style sub-objects). Returns nil when the flag is empty.
 func borderStylesFromFlag(runtime flagView) (map[string]interface{}, error) {

shortcuts/sheets/helpers_test.go

@@ -81,6 +81,53 @@ func runShortcutWithStubs(t *testing.T, sc common.Shortcut, args []string, stubs
 	return stdout.String(), err
 }
 
+// requireProblem asserts err carries a typed errs.Problem with the given
+// category and (optional) subtype, and that its message contains msgContains
+// (skip the message check by passing ""). Returns the Problem so callers can
+// drill into the typed envelope's category-specific fields (e.g. cast to
+// *errs.ValidationError to read .Param / .Params / .Cause).
+//
+// Replaces the older "strings.Contains(stdout+stderr+err.Error(), ...)" pattern
+// across sheets tests: substring on a rendered envelope was brittle (any
+// message tweak silently broke it) and didn't verify that the typed contract —
+// category / subtype / cause preservation — held. Per coding guideline
+// "Error-path tests must assert typed metadata via errs.ProblemOf
+// (category / subtype / param) and cause preservation, not message substrings
+// alone."
+func requireProblem(t *testing.T, err error, wantCategory errs.Category, wantSubtype errs.Subtype, msgContains string) *errs.Problem {
+	t.Helper()
+	if err == nil {
+		t.Fatal("expected error, got nil")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed error carrying errs.Problem, got %T: %v", err, err)
+	}
+	if p.Category != wantCategory {
+		t.Errorf("category = %q, want %q (err=%v)", p.Category, wantCategory, err)
+	}
+	if wantSubtype != "" && p.Subtype != wantSubtype {
+		t.Errorf("subtype = %q, want %q (err=%v)", p.Subtype, wantSubtype, err)
+	}
+	if msgContains != "" && !strings.Contains(p.Message, msgContains) {
+		t.Errorf("message = %q, want containing %q", p.Message, msgContains)
+	}
+	return p
+}
+
+// requireValidation is shorthand for the most common case: a typed
+// CategoryValidation error with SubtypeInvalidArgument. Returns the
+// *errs.ValidationError so callers can also assert on .Param / .Params / .Cause.
+func requireValidation(t *testing.T, err error, msgContains string) *errs.ValidationError {
+	t.Helper()
+	requireProblem(t, err, errs.CategoryValidation, errs.SubtypeInvalidArgument, msgContains)
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) {
+		t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
+	}
+	return ve
+}
+
 func TestSheetHelpersValidationMetadata(t *testing.T) {
 	t.Parallel()
 
@@ -268,3 +315,52 @@ const (
 	testSheetID  = "shtSubA"
 	testSheetID2 = "shtSubB"
 )
+
+// TestParseSpreadsheetRef locks the network-free classification of
+// --url / --spreadsheet-token into a sheet token vs an (unresolved) wiki
+// node_token. The wiki node is resolved later, at Execute time only.
+func TestParseSpreadsheetRef(t *testing.T) {
+	t.Parallel()
+	mk := func(url, tok string) *common.RuntimeContext {
+		cmd := &cobra.Command{Use: "sheets"}
+		cmd.Flags().String("url", url, "")
+		cmd.Flags().String("spreadsheet-token", tok, "")
+		return common.TestNewRuntimeContext(cmd, testConfig(t))
+	}
+	cases := []struct {
+		name      string
+		url       string
+		tok       string
+		wantKind  string
+		wantToken string
+		wantErr   bool
+	}{
+		{name: "sheets url", url: "https://x.feishu.cn/sheets/shtABC", wantKind: spreadsheetRefSheet, wantToken: "shtABC"},
+		{name: "spreadsheets url", url: "https://x.feishu.cn/spreadsheets/shtABC", wantKind: spreadsheetRefSheet, wantToken: "shtABC"},
+		{name: "wiki url", url: "https://x.feishu.cn/wiki/wikDEF", wantKind: spreadsheetRefWiki, wantToken: "wikDEF"},
+		{name: "wiki url with query", url: "https://x.feishu.cn/wiki/wikDEF?sheet=xxxxxx", wantKind: spreadsheetRefWiki, wantToken: "wikDEF"},
+		{name: "raw token", tok: "shtRAW", wantKind: spreadsheetRefSheet, wantToken: "shtRAW"},
+		{name: "sheets url with /wiki/ in query stays sheet", url: "https://x.feishu.cn/sheets/shtABC?from=/wiki/wikX", wantKind: spreadsheetRefSheet, wantToken: "shtABC"},
+		{name: "sheets url with /wiki/ in fragment stays sheet", url: "https://x.feishu.cn/sheets/shtABC#/wiki/wikX", wantKind: spreadsheetRefSheet, wantToken: "shtABC"},
+		{name: "docx url unsupported", url: "https://x.feishu.cn/docx/docABC", wantErr: true},
+		{name: "neither provided", wantErr: true},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			ref, err := parseSpreadsheetRef(mk(tc.url, tc.tok))
+			if tc.wantErr {
+				if err == nil {
+					t.Fatalf("want error, got ref=%+v", ref)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("unexpected error: %v", err)
+			}
+			if ref.Kind != tc.wantKind || ref.Token != tc.wantToken {
+				t.Fatalf("ref = %+v, want {Kind:%s Token:%s}", ref, tc.wantKind, tc.wantToken)
+			}
+		})
+	}
+}

shortcuts/sheets/lark_sheet_batch_update.go

@@ -67,7 +67,7 @@ var BatchUpdate = common.Shortcut{
 		return invokeToolDryRun(token, ToolKindWrite, "batch_update", input)
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		token, err := resolveSpreadsheetToken(runtime)
+		token, err := resolveSpreadsheetTokenExec(runtime)
 		if err != nil {
 			return err
 		}
@@ -180,7 +180,7 @@ var CellsBatchSetStyle = common.Shortcut{
 		return invokeToolDryRun(token, ToolKindWrite, "batch_update", input)
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		token, err := resolveSpreadsheetToken(runtime)
+		token, err := resolveSpreadsheetTokenExec(runtime)
 		if err != nil {
 			return err
 		}
@@ -270,7 +270,7 @@ var CellsBatchClear = common.Shortcut{
 		return invokeToolDryRun(token, ToolKindWrite, "batch_update", input)
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		token, err := resolveSpreadsheetToken(runtime)
+		token, err := resolveSpreadsheetTokenExec(runtime)
 		if err != nil {
 			return err
 		}
@@ -350,7 +350,7 @@ var DropdownUpdate = common.Shortcut{
 		return invokeToolDryRun(token, ToolKindWrite, "batch_update", input)
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		token, err := resolveSpreadsheetToken(runtime)
+		token, err := resolveSpreadsheetTokenExec(runtime)
 		if err != nil {
 			return err
 		}
@@ -396,7 +396,7 @@ var DropdownDelete = common.Shortcut{
 		return invokeToolDryRun(token, ToolKindWrite, "batch_update", input)
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		token, err := resolveSpreadsheetToken(runtime)
+		token, err := resolveSpreadsheetTokenExec(runtime)
 		if err != nil {
 			return err
 		}

shortcuts/sheets/lark_sheet_batch_update_test.go

@@ -5,7 +5,6 @@ package sheets
 
 import (
 	"encoding/json"
-	"strings"
 	"testing"
 )
 
@@ -166,18 +165,16 @@ func TestCellsBatchClear_Guards(t *testing.T) {
 	t.Parallel()
 
 	// sheetless range → prefix guard (shared with the dropdown fan-outs).
-	stdout, stderr, err := runShortcutCapturingErr(t, CellsBatchClear, []string{
+	_, _, err := runShortcutCapturingErr(t, CellsBatchClear, []string{
 		"--url", testURL,
 		"--ranges", `["A1:A10"]`,
 		"--yes",
 		"--dry-run",
 	})
-	if err == nil || !strings.Contains(stdout+stderr+err.Error(), "must include a sheet prefix") {
-		t.Errorf("expected sheet-prefix guard; got=%s|%s|%v", stdout, stderr, err)
-	}
+	requireValidation(t, err, "must include a sheet prefix")
 
 	// missing --yes → confirmation_required (high-risk-write).
-	stdout, stderr, err = runShortcutCapturingErr(t, CellsBatchClear, []string{
+	stdout, stderr, err := runShortcutCapturingErr(t, CellsBatchClear, []string{
 		"--url", testURL,
 		"--ranges", `["sheet1!A1:A10"]`,
 	})
@@ -268,38 +265,32 @@ func TestBatchUpdate_ValidationGuards(t *testing.T) {
 	t.Parallel()
 
 	// dropdown-update with sheetless range
-	stdout, stderr, err := runShortcutCapturingErr(t, DropdownUpdate, []string{
+	_, _, err := runShortcutCapturingErr(t, DropdownUpdate, []string{
 		"--url", testURL,
 		"--ranges", `["A2:A5"]`,
 		"--options", `["a"]`,
 		"--dry-run",
 	})
-	if err == nil || !strings.Contains(stdout+stderr+err.Error(), "must include a sheet prefix") {
-		t.Errorf("expected sheet-prefix guard for +dropdown-update; got=%s|%s|%v", stdout, stderr, err)
-	}
+	requireValidation(t, err, "must include a sheet prefix")
 
 	// batch-update with empty operations
-	stdout, stderr, err = runShortcutCapturingErr(t, BatchUpdate, []string{
+	_, _, err = runShortcutCapturingErr(t, BatchUpdate, []string{
 		"--url", testURL,
 		"--operations", `[]`,
 		"--yes",
 		"--dry-run",
 	})
-	if err == nil || !strings.Contains(stdout+stderr+err.Error(), "non-empty JSON array") {
-		t.Errorf("expected empty-operations guard; got=%s|%s|%v", stdout, stderr, err)
-	}
+	requireValidation(t, err, "non-empty JSON array")
 
 	// dropdown-update with non-array --options (object instead) → array guard
 	// (now via schema validator at parseJSONFlag time)
-	stdout, stderr, err = runShortcutCapturingErr(t, DropdownUpdate, []string{
+	_, _, err = runShortcutCapturingErr(t, DropdownUpdate, []string{
 		"--url", testURL,
 		"--ranges", `["sheet1!A1:A2"]`,
 		"--options", `{"not":"array"}`,
 		"--dry-run",
 	})
-	if err == nil || !strings.Contains(stdout+stderr+err.Error(), `expected type "array"`) {
-		t.Errorf("expected JSON array guard; got=%s|%s|%v", stdout, stderr, err)
-	}
+	requireValidation(t, err, `expected type "array"`)
 }
 
 // TestValidateDropdownRanges_RejectsMalformedRange locks the up-front sheet!range
@@ -322,15 +313,13 @@ func TestValidateDropdownRanges_RejectsMalformedRange(t *testing.T) {
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
-			stdout, stderr, err := runShortcutCapturingErr(t, DropdownUpdate, []string{
+			_, _, err := runShortcutCapturingErr(t, DropdownUpdate, []string{
 				"--url", testURL,
 				"--ranges", tc.ranges,
 				"--options", `["a"]`,
 				"--dry-run",
 			})
-			if err == nil || !strings.Contains(stdout+stderr+err.Error(), tc.want) {
-				t.Errorf("ranges=%s: expected error containing %q; got=%s|%s|%v", tc.ranges, tc.want, stdout, stderr, err)
-			}
+			requireValidation(t, err, tc.want)
 		})
 	}
 }
@@ -419,18 +408,13 @@ func TestBatchUpdate_TranslatorRejects(t *testing.T) {
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
-			stdout, stderr, err := runShortcutCapturingErr(t, BatchUpdate, []string{
+			_, _, err := runShortcutCapturingErr(t, BatchUpdate, []string{
 				"--url", testURL,
 				"--operations", tc.opsJSON,
 				"--yes",
 				"--dry-run",
 			})
-			if err == nil {
-				t.Fatalf("expected error containing %q; got stdout=%s stderr=%s", tc.wantMatch, stdout, stderr)
-			}
-			if !strings.Contains(stdout+stderr+err.Error(), tc.wantMatch) {
-				t.Errorf("expected error containing %q; got: %s | %s | %v", tc.wantMatch, stdout, stderr, err)
-			}
+			requireValidation(t, err, tc.wantMatch)
 		})
 	}
 }

shortcuts/sheets/lark_sheet_dataframe.go

@@ -0,0 +1,621 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"io"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/apache/arrow/go/v17/arrow"
+	"github.com/apache/arrow/go/v17/arrow/array"
+	"github.com/apache/arrow/go/v17/arrow/ipc"
+	"github.com/apache/arrow/go/v17/arrow/memory"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/extension/fileio"
+	"github.com/larksuite/cli/shortcuts/common"
+)
+
+// ─── --dataframe (Arrow IPC / Feather v2 binary input) ────────────────
+//
+// --dataframe is the binary-typed twin of --sheets. The wire payload is one
+// Arrow IPC file (a.k.a. Feather v2 — what `pandas.DataFrame.to_feather()`
+// writes), single schema, optionally multi-batch. Type / format are read off
+// the Arrow schema (no separate dtypes/formats maps), and per-column number
+// format can be set via the field's `number_format` metadata key:
+//
+//	pa.field("price", pa.float64(), metadata={b"number_format": b"$#,##0.00"})
+//
+// One DataFrame writes into one sub-sheet at fixed defaults: name `Sheet1`
+// (adopted in place by +workbook-create; created when absent by +table-put),
+// overwrite from A1 with header on, allow_overwrite=true. The shortcut
+// surface is deliberately the one flag — anything that needs a different
+// sheet name / anchor / mode / multi-sheet falls back to --sheets, whose
+// JSON payload already carries every knob.
+//
+// Binary IO note: --dataframe bypasses the text-oriented Input resolver
+// (`runtime.Str("dataframe")` carries a *path*, not file contents). Reading
+// the Arrow bytes through that resolver would TrimSpace the trailing IPC
+// magic / corrupt non-UTF8 bytes. Path → FileIO.Open → io.ReadAll keeps the
+// stream byte-exact. "-" reads from stdin directly.
+
+// dataframeDefaultSheetName is the sub-sheet name --dataframe writes into.
+// Matches valuesSheetName so +workbook-create adopts the brand-new
+// workbook's default sheet in place (no stray empty Sheet1 left behind);
+// +table-put creates Sheet1 if it doesn't already exist.
+const dataframeDefaultSheetName = valuesSheetName
+
+// parseDataframePayload reads the --dataframe path (Arrow IPC file) and
+// composes a single-sheet tablePayload at the fixed default placement.
+// Network-free: safe from Validate and DryRun. The resulting tableSheetSpec
+// rides the same buildSheetMatrix / buildTypedCell path as a --sheets entry,
+// so downstream is unaware of where the rows came from.
+func parseDataframePayload(rctx *common.RuntimeContext) (*tablePayload, error) {
+	raw := strings.TrimSpace(rctx.Str("dataframe"))
+	if raw == "" {
+		return nil, common.ValidationErrorf("--dataframe is required")
+	}
+	data, err := readDataframeBytes(rctx, raw)
+	if err != nil {
+		return nil, err
+	}
+	spec, err := decodeArrowToSheet(data, dataframeDefaultSheetName)
+	if err != nil {
+		return nil, common.ValidationErrorf("--dataframe: %v", err).WithCause(err)
+	}
+	payload := &tablePayload{Sheets: []tableSheetSpec{spec}}
+	if err := payload.validate(); err != nil {
+		return nil, err
+	}
+	return payload, nil
+}
+
+// dataframeStdinCache holds the bytes read from stdin on the first call so a
+// later call (Validate → Execute / DryRun) gets the same bytes instead of an
+// empty stream — stdin is single-shot, but parseDataframePayload runs
+// multiple times per command invocation. Process-wide is fine: lark-cli is
+// one-shot (one command per process). Tests reset by setting it back to nil.
+var dataframeStdinCache []byte
+
+// Memory caps for --dataframe. The Arrow IPC reader allocates large buffers up
+// front, and arrowRecordToRows materializes every cell into [][]interface{}, so
+// an unbounded input could OOM the CLI before the backend's per-write limits
+// kicked in. The caps mirror the backend's per-sheet hard ceilings (200 cols,
+// 50000 rows) plus a generous overall byte cap that still fits the worst-case
+// dense numeric payload (200 × 50000 cells × ~25 bytes Arrow overhead ≈ 250 MB).
+const (
+	dataframeMaxBytes = 256 * 1024 * 1024 // 256 MiB raw IPC payload
+	dataframeMaxCols  = 200               // backend hard ceiling
+	dataframeMaxRows  = 50000             // backend hard ceiling
+)
+
+// readDataframeBytes resolves --dataframe to raw binary. A literal `@` prefix
+// is tolerated for symmetry with --sheets (`@/tmp/x.arrow` and `/tmp/x.arrow`
+// both work). `-` reads stdin verbatim — cached on first call so Validate /
+// Execute / DryRun all see the same bytes. Bytes are returned untouched: no
+// TrimSpace, no BOM strip — both would corrupt an Arrow IPC stream.
+func readDataframeBytes(rctx *common.RuntimeContext, raw string) ([]byte, error) {
+	// readCapped pulls up to dataframeMaxBytes+1 bytes from r so we can detect
+	// "exceeded cap" without allocating the entire oversized payload up front.
+	readCapped := func(r io.Reader) ([]byte, error) {
+		data, err := io.ReadAll(io.LimitReader(r, dataframeMaxBytes+1))
+		if err != nil {
+			return nil, err
+		}
+		if len(data) > dataframeMaxBytes {
+			return nil, common.ValidationErrorf(
+				"--dataframe: payload exceeds %d MiB cap (limits CLI memory; the backend per-sheet ceilings are 200 cols × 50000 rows)",
+				dataframeMaxBytes/(1024*1024))
+		}
+		return data, nil
+	}
+
+	if raw == "-" {
+		if dataframeStdinCache != nil {
+			return dataframeStdinCache, nil
+		}
+		// A process has a single stdin: --dataframe is binary and bypasses the
+		// common Input resolver, so we have to share the stdin-consumed flag with
+		// it explicitly. Without this, e.g. `+table-put --dataframe - --styles -`
+		// would be accepted and one of them would silently see an empty stream.
+		if rctx.StdinConsumed() {
+			return nil, common.ValidationErrorf("--dataframe: stdin (-) can only be used by one flag").
+				WithHint("a process has a single stdin, so only one flag per call may use '-'; pass the others as @file (e.g. --styles @/path/to/styles.json)")
+		}
+		ios := rctx.IO()
+		if ios == nil || ios.In == nil {
+			return nil, common.ValidationErrorf("--dataframe: stdin is not available")
+		}
+		data, err := readCapped(ios.In)
+		if err != nil {
+			if errs.IsTyped(err) {
+				return nil, err
+			}
+			return nil, common.ValidationErrorf("--dataframe: read stdin: %v", err).WithCause(err)
+		}
+		if len(data) == 0 {
+			return nil, common.ValidationErrorf("--dataframe: stdin is empty")
+		}
+		dataframeStdinCache = data
+		rctx.MarkStdinConsumed()
+		return data, nil
+	}
+	path := strings.TrimPrefix(raw, "@")
+	fio := rctx.FileIO()
+	if fio == nil {
+		return nil, common.ValidationErrorf("--dataframe: file input is not available in this context")
+	}
+	// Pre-check size via Stat so a multi-GB file is rejected immediately
+	// instead of being streamed all the way to the cap.
+	if info, statErr := fio.Stat(path); statErr == nil && info.Size() > dataframeMaxBytes {
+		return nil, common.ValidationErrorf(
+			"--dataframe: file %q is %d MiB, exceeds %d MiB cap",
+			path, info.Size()/(1024*1024), dataframeMaxBytes/(1024*1024))
+	}
+	f, err := fio.Open(path)
+	if err != nil {
+		return nil, common.ValidationErrorf("--dataframe: %v", err).WithCause(err)
+	}
+	defer f.Close()
+	data, err := readCapped(f)
+	if err != nil {
+		if errs.IsTyped(err) {
+			return nil, err
+		}
+		return nil, common.ValidationErrorf("--dataframe: %v", err).WithCause(err)
+	}
+	if len(data) == 0 {
+		return nil, common.ValidationErrorf("--dataframe: file %q is empty", path)
+	}
+	return data, nil
+}
+
+// decodeArrowToSheet reads `data` as an Arrow IPC file (single schema,
+// possibly multi-batch) and produces a tableSheetSpec with name + columns +
+// rows filled in. Sheet placement (start_cell / mode / header / overwrite) is
+// not touched here — parseDataframePayload layers those on from CLI flags.
+func decodeArrowToSheet(data []byte, sheetName string) (tableSheetSpec, error) {
+	reader, err := ipc.NewFileReader(bytes.NewReader(data))
+	if err != nil {
+		return tableSheetSpec{}, fmt.Errorf("invalid Arrow IPC file (expected pandas df.to_feather output): %w", err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	defer reader.Close()
+
+	schema := reader.Schema()
+	if schema == nil || schema.NumFields() == 0 {
+		return tableSheetSpec{}, fmt.Errorf("Arrow schema has no fields") //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+
+	ncols := schema.NumFields()
+	if ncols > dataframeMaxCols {
+		// Fail fast at the schema layer before allocating per-column slices.
+		// 200 cols matches the backend's per-sheet hard ceiling — anything past
+		// that would error on the first set_cell_range anyway.
+		return tableSheetSpec{}, fmt.Errorf("%d columns exceeds the per-sheet ceiling of %d", ncols, dataframeMaxCols) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	cols := make([]tableColumnSpec, ncols)
+	seen := make(map[string]bool, ncols)
+	for i := 0; i < ncols; i++ {
+		f := schema.Field(i)
+		name := f.Name
+		if strings.TrimSpace(name) == "" {
+			return tableSheetSpec{}, fmt.Errorf("column %d has empty name", i) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		if seen[name] {
+			return tableSheetSpec{}, fmt.Errorf("duplicate column name %q", name) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		seen[name] = true
+		typ, format, err := arrowFieldToTypeFormat(f)
+		if err != nil {
+			return tableSheetSpec{}, fmt.Errorf("column %q: %w", name, err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		cols[i] = tableColumnSpec{Name: name, Type: typ, Format: format}
+	}
+
+	var rows [][]interface{}
+	for b := 0; b < reader.NumRecords(); b++ {
+		rec, err := reader.RecordAt(b)
+		if err != nil {
+			return tableSheetSpec{}, fmt.Errorf("read record batch %d: %w", b, err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		// Reject early during iteration before materializing more rows into the
+		// [][]interface{} buffer — without this, a 1M-row Arrow file would be
+		// fully decoded into memory before the writer's per-batch size check
+		// kicks in.
+		if int64(len(rows))+rec.NumRows() > int64(dataframeMaxRows) {
+			rec.Release()
+			return tableSheetSpec{}, fmt.Errorf("%d rows exceeds the per-sheet ceiling of %d", int64(len(rows))+rec.NumRows(), dataframeMaxRows) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		batchRows, err := arrowRecordToRows(rec, cols)
+		rec.Release()
+		if err != nil {
+			return tableSheetSpec{}, err
+		}
+		rows = append(rows, batchRows...)
+	}
+
+	return tableSheetSpec{Name: sheetName, Columns: cols, Rows: rows}, nil
+}
+
+// arrowFieldToTypeFormat maps an Arrow field to the internal (type, format)
+// pair. The field's `number_format` metadata key — when present — sets the
+// Excel number_format string verbatim; otherwise sensible defaults are
+// applied per type (`@` text for strings, `yyyy-mm-dd` for dates).
+func arrowFieldToTypeFormat(f arrow.Field) (typ, format string, err error) {
+	if v, ok := f.Metadata.GetValue("number_format"); ok {
+		format = strings.TrimSpace(v)
+	}
+	switch f.Type.(type) {
+	case *arrow.StringType, *arrow.LargeStringType:
+		if format == "" {
+			format = "@"
+		}
+		return "string", format, nil
+	case *arrow.BooleanType:
+		return "bool", format, nil
+	case *arrow.Date32Type, *arrow.Date64Type, *arrow.TimestampType:
+		if format == "" {
+			format = "yyyy-mm-dd"
+		}
+		return "date", format, nil
+	}
+	if isArrowNumericType(f.Type) {
+		return "number", format, nil
+	}
+	return "", "", fmt.Errorf("unsupported Arrow type %s (want string/number/date/bool)", f.Type.Name()) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+}
+
+func isArrowNumericType(t arrow.DataType) bool {
+	switch t.ID() {
+	case arrow.INT8, arrow.INT16, arrow.INT32, arrow.INT64,
+		arrow.UINT8, arrow.UINT16, arrow.UINT32, arrow.UINT64,
+		arrow.FLOAT16, arrow.FLOAT32, arrow.FLOAT64:
+		return true
+	}
+	return false
+}
+
+// arrowRecordToRows transposes one column-batch into row-major
+// [][]interface{} matched to `cols`. Cells are stamped with the same value
+// shapes buildTypedCell expects from the JSON path: nil for nulls,
+// json.Number for numerics (precision-preserving), `yyyy-mm-dd` strings for
+// dates/timestamps, bool for booleans, string for strings.
+func arrowRecordToRows(rec arrow.Record, cols []tableColumnSpec) ([][]interface{}, error) {
+	if int(rec.NumCols()) != len(cols) {
+		return nil, fmt.Errorf("record has %d cols, schema declared %d", rec.NumCols(), len(cols)) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	nrows := int(rec.NumRows())
+	rows := make([][]interface{}, nrows)
+	for r := range rows {
+		rows[r] = make([]interface{}, len(cols))
+	}
+	for c := range cols {
+		arr := rec.Column(c)
+		for r := 0; r < nrows; r++ {
+			if arr.IsNull(r) {
+				continue
+			}
+			v, err := arrowCellValue(arr, r)
+			if err != nil {
+				return nil, fmt.Errorf("row %d column %q: %w", r, cols[c].Name, err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+			}
+			rows[r][c] = v
+		}
+	}
+	return rows, nil
+}
+
+func arrowCellValue(arr arrow.Array, i int) (interface{}, error) {
+	switch a := arr.(type) {
+	case *array.String:
+		return a.Value(i), nil
+	case *array.LargeString:
+		return a.Value(i), nil
+	case *array.Boolean:
+		return a.Value(i), nil
+	case *array.Int8:
+		return json.Number(strconv.FormatInt(int64(a.Value(i)), 10)), nil
+	case *array.Int16:
+		return json.Number(strconv.FormatInt(int64(a.Value(i)), 10)), nil
+	case *array.Int32:
+		return json.Number(strconv.FormatInt(int64(a.Value(i)), 10)), nil
+	case *array.Int64:
+		return json.Number(strconv.FormatInt(a.Value(i), 10)), nil
+	case *array.Uint8:
+		return json.Number(strconv.FormatUint(uint64(a.Value(i)), 10)), nil
+	case *array.Uint16:
+		return json.Number(strconv.FormatUint(uint64(a.Value(i)), 10)), nil
+	case *array.Uint32:
+		return json.Number(strconv.FormatUint(uint64(a.Value(i)), 10)), nil
+	case *array.Uint64:
+		return json.Number(strconv.FormatUint(a.Value(i), 10)), nil
+	case *array.Float16:
+		return json.Number(strconv.FormatFloat(float64(a.Value(i).Float32()), 'f', -1, 32)), nil
+	case *array.Float32:
+		return json.Number(strconv.FormatFloat(float64(a.Value(i)), 'f', -1, 32)), nil
+	case *array.Float64:
+		return json.Number(strconv.FormatFloat(a.Value(i), 'f', -1, 64)), nil
+	case *array.Date32:
+		// Date32: days since 1970-01-01 (epoch). Multiply to seconds, format
+		// in UTC so timezone offset can't flip the calendar date.
+		t := time.Unix(int64(a.Value(i))*86400, 0).UTC()
+		return t.Format("2006-01-02"), nil
+	case *array.Date64:
+		t := time.UnixMilli(int64(a.Value(i))).UTC()
+		return t.Format("2006-01-02"), nil
+	case *array.Timestamp:
+		ts := int64(a.Value(i))
+		unit := a.DataType().(*arrow.TimestampType).Unit
+		var t time.Time
+		switch unit {
+		case arrow.Second:
+			t = time.Unix(ts, 0).UTC()
+		case arrow.Millisecond:
+			t = time.UnixMilli(ts).UTC()
+		case arrow.Microsecond:
+			t = time.UnixMicro(ts).UTC()
+		case arrow.Nanosecond:
+			t = time.Unix(0, ts).UTC()
+		default:
+			return nil, fmt.Errorf("unsupported timestamp unit %v", unit) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		return t.Format("2006-01-02"), nil
+	}
+	return nil, fmt.Errorf("unsupported Arrow array %T", arr) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+}
+
+// ─── --dataframe-out (Arrow IPC binary output, mirror of --dataframe) ──
+//
+// +table-get's binary read-back: encode one sheet's typed read-back as an
+// Arrow IPC file (Feather v2), so pandas can `pd.read_feather(path)` /
+// `pd.read_feather(BytesIO(stdout))` symmetrically with the put side.
+// Single-sheet only — Arrow IPC carries one schema per file. The JSON path
+// is unchanged; --dataframe-out swaps the encoder for callers that already
+// have pandas / pyarrow in their pipeline.
+
+// encodeSheetMapToArrowIPC turns one readSheetAsSpec output into an Arrow IPC
+// file blob. Internal column types are recovered from `dtypes` (the wire
+// proxy for the typed protocol), and per-column `number_format` rides through
+// as Arrow field metadata so the file feeds straight back into
+// `+table-put --dataframe`.
+func encodeSheetMapToArrowIPC(sheet map[string]interface{}) ([]byte, error) {
+	columns, _ := sheet["columns"].([]interface{})
+	if len(columns) == 0 {
+		return nil, fmt.Errorf("sheet has no columns") //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	dtypes, _ := sheet["dtypes"].(map[string]interface{})
+	formats, _ := sheet["formats"].(map[string]interface{})
+	// `data` arrives as either []interface{} (when the sheet came through a
+	// JSON round-trip / unit-test fixture) or [][]interface{} (the shape
+	// readSheetAsSpec directly emits in production). Accept both — anything
+	// else falls through to a zero-row table.
+	var rawData [][]interface{}
+	switch d := sheet["data"].(type) {
+	case [][]interface{}:
+		rawData = d
+	case []interface{}:
+		rawData = make([][]interface{}, len(d))
+		for i, r := range d {
+			rawData[i], _ = r.([]interface{})
+		}
+	}
+
+	ncols := len(columns)
+	colNames := make([]string, ncols)
+	colTypes := make([]string, ncols)
+	fields := make([]arrow.Field, ncols)
+	for i, c := range columns {
+		name, _ := c.(string)
+		if name == "" {
+			return nil, fmt.Errorf("column %d has empty name", i) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		colNames[i] = name
+		dt, _ := dtypes[name].(string)
+		colTypes[i] = dtypeToInternalType(dt)
+		var meta arrow.Metadata
+		if formats != nil {
+			if nf, ok := formats[name].(string); ok && strings.TrimSpace(nf) != "" {
+				meta = arrow.NewMetadata([]string{"number_format"}, []string{nf})
+			}
+		}
+		fields[i] = arrow.Field{
+			Name:     name,
+			Type:     internalTypeToArrowType(colTypes[i]),
+			Nullable: true,
+			Metadata: meta,
+		}
+	}
+	schema := arrow.NewSchema(fields, nil)
+
+	mem := memory.NewGoAllocator()
+	rb := array.NewRecordBuilder(mem, schema)
+	defer rb.Release()
+	for r, row := range rawData {
+		if len(row) != ncols {
+			return nil, fmt.Errorf("row %d has %d cells, want %d", r, len(row), ncols) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		for c := 0; c < ncols; c++ {
+			if err := appendArrowCell(rb.Field(c), colTypes[c], row[c]); err != nil {
+				return nil, fmt.Errorf("row %d column %q: %w", r, colNames[c], err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+			}
+		}
+	}
+	rec := rb.NewRecord()
+	defer rec.Release()
+
+	var buf bytesWriterSeeker
+	w, err := ipc.NewFileWriter(&buf, ipc.WithSchema(schema), ipc.WithAllocator(mem))
+	if err != nil {
+		return nil, fmt.Errorf("ipc.NewFileWriter: %w", err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	if err := w.Write(rec); err != nil {
+		return nil, fmt.Errorf("write record: %w", err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	if err := w.Close(); err != nil {
+		return nil, fmt.Errorf("close writer: %w", err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	return buf.buf, nil
+}
+
+// dtypeToInternalType inverts typeToDtype so the Arrow encoder can pick an
+// internal column type from the wire-level dtype string. Unknown / object
+// falls back to string (lossless: every cell is already typed as such).
+func dtypeToInternalType(dtype string) string {
+	switch strings.ToLower(strings.TrimSpace(dtype)) {
+	case "float64", "float32", "int64", "int32", "int16", "int8",
+		"uint64", "uint32", "uint16", "uint8":
+		return "number"
+	case "bool", "boolean":
+		return "bool"
+	}
+	if strings.HasPrefix(strings.ToLower(dtype), "datetime") {
+		return "date"
+	}
+	return "string"
+}
+
+// internalTypeToArrowType is the put-side dtypeToTypeFormat dual: maps the
+// internal column type to the Arrow data type the encoder builds a column
+// with. Numbers go to float64 because +table-get can't tell int from float
+// from a number_format alone — float64 covers both losslessly for the cell
+// ranges Lark Sheets accepts.
+func internalTypeToArrowType(typ string) arrow.DataType {
+	switch typ {
+	case "number":
+		return arrow.PrimitiveTypes.Float64
+	case "date":
+		return arrow.FixedWidthTypes.Date32
+	case "bool":
+		return arrow.FixedWidthTypes.Boolean
+	}
+	return arrow.BinaryTypes.String
+}
+
+// appendArrowCell stamps one cell into its column builder. Cell shape matches
+// what cellToTyped emits on the JSON path: json.Number for numbers, ISO
+// `yyyy-mm-dd` string for dates, plain string for strings, bool for bools,
+// nil for empty. Anything off-shape errors so the caller doesn't silently
+// emit nulls for malformed data.
+func appendArrowCell(b array.Builder, typ string, v interface{}) error {
+	if v == nil {
+		b.AppendNull()
+		return nil
+	}
+	switch typ {
+	case "string":
+		s, ok := v.(string)
+		if !ok {
+			return fmt.Errorf("string expects string value, got %T", v) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		b.(*array.StringBuilder).Append(s)
+	case "number":
+		f, err := arrowNumber(v)
+		if err != nil {
+			return err
+		}
+		b.(*array.Float64Builder).Append(f)
+	case "date":
+		s, ok := v.(string)
+		if !ok {
+			return fmt.Errorf("date expects ISO yyyy-mm-dd string, got %T", v) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		t, err := time.Parse("2006-01-02", strings.TrimSpace(s))
+		if err != nil {
+			return fmt.Errorf("date parse %q: %w", s, err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		b.(*array.Date32Builder).Append(arrow.Date32FromTime(t))
+	case "bool":
+		bb, ok := v.(bool)
+		if !ok {
+			return fmt.Errorf("bool expects bool, got %T", v) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		b.(*array.BooleanBuilder).Append(bb)
+	default:
+		return fmt.Errorf("unsupported internal type %q", typ) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	return nil
+}
+
+// arrowNumber converts the number cell shape readSheetAsSpec emits
+// (json.Number) plus the float fallback to float64 for the Arrow builder.
+func arrowNumber(v interface{}) (float64, error) {
+	switch n := v.(type) {
+	case json.Number:
+		f, err := n.Float64()
+		if err != nil {
+			return 0, fmt.Errorf("number parse %q: %w", n.String(), err) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+		}
+		return f, nil
+	case float64:
+		return n, nil
+	}
+	return 0, fmt.Errorf("number expects numeric value, got %T", v) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+}
+
+// bytesWriterSeeker is a 10-line in-memory io.WriteSeeker for
+// ipc.NewFileWriter, which seeks back to patch a footer offset. Using a
+// buffer (instead of a temp file or os.Stdout, which isn't seekable) keeps
+// --dataframe-out's stdout path zero-IO and stays straightforward.
+type bytesWriterSeeker struct {
+	buf []byte
+	pos int64
+}
+
+func (w *bytesWriterSeeker) Write(p []byte) (int, error) {
+	end := w.pos + int64(len(p))
+	if end > int64(len(w.buf)) {
+		w.buf = append(w.buf, make([]byte, end-int64(len(w.buf)))...)
+	}
+	n := copy(w.buf[w.pos:], p)
+	w.pos = end
+	return n, nil
+}
+
+func (w *bytesWriterSeeker) Seek(offset int64, whence int) (int64, error) {
+	switch whence {
+	case io.SeekStart:
+		w.pos = offset
+	case io.SeekCurrent:
+		w.pos += offset
+	case io.SeekEnd:
+		w.pos = int64(len(w.buf)) + offset
+	default:
+		return 0, fmt.Errorf("unknown whence %d", whence) //nolint:forbidigo // intermediate error; the command layer wraps it into a typed --dataframe/--dataframe-out validation error
+	}
+	return w.pos, nil
+}
+
+// writeDataframeOut dispatches the encoded Arrow bytes to wherever --dataframe-out
+// points: `-` → process stdout, `@<path>` or plain path → local file. Symmetric
+// with readDataframeBytes on the input side: same `@` tolerance, same TrimPrefix
+// semantics, and an absolute path will still get rejected by FileIO's SafePath.
+func writeDataframeOut(rctx *common.RuntimeContext, raw string, data []byte) error {
+	if raw == "-" {
+		out := rctx.IO()
+		if out == nil || out.Out == nil {
+			return common.ValidationErrorf("--dataframe-out: stdout is not available")
+		}
+		if _, err := out.Out.Write(data); err != nil {
+			return errs.NewInternalError(errs.SubtypeFileIO, "--dataframe-out: write stdout").WithCause(err)
+		}
+		return nil
+	}
+	path := strings.TrimPrefix(raw, "@")
+	fio := rctx.FileIO()
+	if fio == nil {
+		return common.ValidationErrorf("--dataframe-out: file output is not available in this context")
+	}
+	// FileIO.Save validates the path via SafeOutputPath (the same sandbox
+	// readDataframeBytes hits on the input side) and writes atomically, so we
+	// don't need an extra ValidatePath call here.
+	if _, err := fio.Save(path, fileio.SaveOptions{ContentLength: int64(len(data))}, bytes.NewReader(data)); err != nil {
+		return errs.NewInternalError(errs.SubtypeFileIO, "--dataframe-out: write %q", path).WithCause(err)
+	}
+	return nil
+}

shortcuts/sheets/lark_sheet_dataframe_test.go

@@ -0,0 +1,421 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/apache/arrow/go/v17/arrow"
+	"github.com/apache/arrow/go/v17/arrow/array"
+	"github.com/apache/arrow/go/v17/arrow/ipc"
+	"github.com/apache/arrow/go/v17/arrow/memory"
+	"github.com/larksuite/cli/shortcuts/common"
+)
+
+// buildArrowIPC writes one record into a Feather v2 (Arrow IPC file) blob.
+// Used by the round-trip tests below to stand in for what
+// `pandas.DataFrame.to_feather(path)` would produce; saves the tests from
+// depending on a pandas-shaped fixture file.
+//
+// ipc.NewFileWriter wants an io.WriteSeeker (it back-patches a footer
+// offset), so we write to a temp file and read the bytes back — simpler than
+// re-implementing a seekable in-memory buffer.
+func buildArrowIPC(t *testing.T, schema *arrow.Schema, build func(b *array.RecordBuilder)) []byte {
+	t.Helper()
+	mem := memory.NewGoAllocator()
+	rb := array.NewRecordBuilder(mem, schema)
+	defer rb.Release()
+	build(rb)
+	rec := rb.NewRecord()
+	defer rec.Release()
+
+	path := filepath.Join(t.TempDir(), "df.arrow")
+	f, err := os.Create(path)
+	if err != nil {
+		t.Fatalf("create temp arrow file: %v", err)
+	}
+	w, err := ipc.NewFileWriter(f, ipc.WithSchema(schema), ipc.WithAllocator(mem))
+	if err != nil {
+		f.Close()
+		t.Fatalf("ipc.NewFileWriter: %v", err)
+	}
+	if err := w.Write(rec); err != nil {
+		t.Fatalf("write record: %v", err)
+	}
+	if err := w.Close(); err != nil {
+		t.Fatalf("close writer: %v", err)
+	}
+	if err := f.Close(); err != nil {
+		t.Fatalf("close file: %v", err)
+	}
+	data, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("read temp arrow file: %v", err)
+	}
+	return data
+}
+
+// TestDataframe_RoundTripCoreTypes pins down the Arrow-schema → internal
+// (type, format) mapping and the per-cell value shape that buildTypedCell
+// expects: number cells are json.Number (precision-preserving), date cells
+// are `yyyy-mm-dd` strings, bool/string come through verbatim. Numbers, dates,
+// strings, bools, and nulls all in one record so a future Arrow-Go bump can't
+// quietly regress any one family.
+func TestDataframe_RoundTripCoreTypes(t *testing.T) {
+	t.Parallel()
+
+	schema := arrow.NewSchema([]arrow.Field{
+		{Name: "name", Type: arrow.BinaryTypes.String},
+		{Name: "qty", Type: arrow.PrimitiveTypes.Int64},
+		{Name: "price", Type: arrow.PrimitiveTypes.Float64, Metadata: arrow.NewMetadata(
+			[]string{"number_format"}, []string{"$#,##0.00"},
+		)},
+		{Name: "active", Type: arrow.FixedWidthTypes.Boolean},
+		{Name: "shipped_on", Type: arrow.FixedWidthTypes.Date32},
+	}, nil)
+
+	jan15 := arrow.Date32FromTime(time.Date(2024, 1, 15, 0, 0, 0, 0, time.UTC))
+	feb02 := arrow.Date32FromTime(time.Date(2024, 2, 2, 0, 0, 0, 0, time.UTC))
+
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		b.Field(0).(*array.StringBuilder).AppendValues([]string{"alice", ""}, []bool{true, false})
+		b.Field(1).(*array.Int64Builder).AppendValues([]int64{42, 0}, []bool{true, false})
+		b.Field(2).(*array.Float64Builder).AppendValues([]float64{19.95, 0}, []bool{true, false})
+		b.Field(3).(*array.BooleanBuilder).AppendValues([]bool{true, false}, []bool{true, true})
+		b.Field(4).(*array.Date32Builder).AppendValues([]arrow.Date32{jan15, feb02}, []bool{true, true})
+	})
+
+	spec, err := decodeArrowToSheet(buf, "S1")
+	if err != nil {
+		t.Fatalf("decodeArrowToSheet: %v", err)
+	}
+	if spec.Name != "S1" {
+		t.Errorf("sheet name = %q, want S1", spec.Name)
+	}
+	if len(spec.Columns) != 5 {
+		t.Fatalf("got %d columns, want 5", len(spec.Columns))
+	}
+	want := []struct{ typ, format string }{
+		{"string", "@"},
+		{"number", ""},
+		{"number", "$#,##0.00"},
+		{"bool", ""},
+		{"date", "yyyy-mm-dd"},
+	}
+	for i, w := range want {
+		if spec.Columns[i].Type != w.typ {
+			t.Errorf("columns[%d].Type = %q, want %q", i, spec.Columns[i].Type, w.typ)
+		}
+		if spec.Columns[i].Format != w.format {
+			t.Errorf("columns[%d].Format = %q, want %q", i, spec.Columns[i].Format, w.format)
+		}
+	}
+
+	if len(spec.Rows) != 2 {
+		t.Fatalf("got %d rows, want 2", len(spec.Rows))
+	}
+	// Row 0: every field present, types match what buildTypedCell will accept.
+	row0 := spec.Rows[0]
+	if row0[0] != "alice" {
+		t.Errorf("row0[name] = %#v, want \"alice\"", row0[0])
+	}
+	if n, ok := row0[1].(json.Number); !ok || n.String() != "42" {
+		t.Errorf("row0[qty] = %#v, want json.Number(\"42\")", row0[1])
+	}
+	if n, ok := row0[2].(json.Number); !ok || n.String() != "19.95" {
+		t.Errorf("row0[price] = %#v, want json.Number(\"19.95\")", row0[2])
+	}
+	if row0[3] != true {
+		t.Errorf("row0[active] = %#v, want true", row0[3])
+	}
+	if row0[4] != "2024-01-15" {
+		t.Errorf("row0[shipped_on] = %#v, want \"2024-01-15\"", row0[4])
+	}
+
+	// Row 1: nulls on name/qty/price (despite the buffer values) must become nil
+	// so buildTypedCell paints an empty cell that still carries number_format.
+	row1 := spec.Rows[1]
+	for _, c := range []int{0, 1, 2} {
+		if row1[c] != nil {
+			t.Errorf("row1[%d] = %#v, want nil (null in arrow)", c, row1[c])
+		}
+	}
+	if row1[3] != false {
+		t.Errorf("row1[active] = %#v, want false", row1[3])
+	}
+	if row1[4] != "2024-02-02" {
+		t.Errorf("row1[shipped_on] = %#v, want \"2024-02-02\"", row1[4])
+	}
+}
+
+// TestDataframe_Timestamp pins the timestamp → date conversion for the
+// timestamp[us] case (pandas default for `pd.Timestamp` columns once written
+// via `to_feather`). Only the calendar date matters for our `yyyy-mm-dd`
+// landing — guard against TZ drift from the wrong unit pick.
+func TestDataframe_Timestamp(t *testing.T) {
+	t.Parallel()
+	schema := arrow.NewSchema([]arrow.Field{
+		{Name: "ts", Type: &arrow.TimestampType{Unit: arrow.Microsecond}},
+	}, nil)
+	ts := arrow.Timestamp(time.Date(2024, 6, 12, 14, 30, 0, 0, time.UTC).UnixMicro())
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		b.Field(0).(*array.TimestampBuilder).AppendValues([]arrow.Timestamp{ts}, []bool{true})
+	})
+	spec, err := decodeArrowToSheet(buf, "S")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if spec.Columns[0].Type != "date" {
+		t.Errorf("type = %q, want date", spec.Columns[0].Type)
+	}
+	if got := spec.Rows[0][0]; got != "2024-06-12" {
+		t.Errorf("ts = %#v, want \"2024-06-12\"", got)
+	}
+}
+
+// TestDataframe_EmptySchema rejects an Arrow file whose schema has no fields:
+// a 0-column "DataFrame" would write a header-less, data-less block that
+// validates as "writer ran successfully" but produces nothing — the test ties
+// that off as an explicit error rather than letting it slip through.
+func TestDataframe_EmptySchema(t *testing.T) {
+	t.Parallel()
+	schema := arrow.NewSchema(nil, nil)
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {})
+	_, err := decodeArrowToSheet(buf, "S")
+	if err == nil || !strings.Contains(err.Error(), "no fields") {
+		t.Errorf("err = %v, want 'no fields' error", err)
+	}
+}
+
+// TestDataframe_DuplicateColumn catches duplicate-name columns at decode
+// time. Validate already rejects duplicate column names for the JSON path;
+// the Arrow path mirrors that so the error surfaces with the same shape.
+func TestDataframe_DuplicateColumn(t *testing.T) {
+	t.Parallel()
+	schema := arrow.NewSchema([]arrow.Field{
+		{Name: "x", Type: arrow.BinaryTypes.String},
+		{Name: "x", Type: arrow.PrimitiveTypes.Int64},
+	}, nil)
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		b.Field(0).(*array.StringBuilder).Append("")
+		b.Field(1).(*array.Int64Builder).Append(0)
+	})
+	_, err := decodeArrowToSheet(buf, "S")
+	if err == nil || !strings.Contains(err.Error(), "duplicate") {
+		t.Errorf("err = %v, want duplicate-column error", err)
+	}
+}
+
+// TestDataframe_BadBytes rejects a non-Arrow blob with a hint pointing at
+// pandas df.to_feather so users see what producer is expected without having
+// to grep the docs.
+func TestDataframe_BadBytes(t *testing.T) {
+	t.Parallel()
+	_, err := decodeArrowToSheet([]byte("not arrow"), "S")
+	if err == nil || !strings.Contains(err.Error(), "Arrow") {
+		t.Errorf("err = %v, want Arrow-decode error", err)
+	}
+}
+
+// TestDecodeArrowToSheet_RejectsTooManyColumns guards the per-sheet column cap.
+// Without it a wide Arrow file (e.g. 201+ columns) would allocate a long
+// tableColumnSpec slice + decode every batch before the backend's 200-column
+// per-sheet ceiling rejected the first set_cell_range — wasting both CPU and
+// memory before the failure surfaces. Cap matches the backend hard ceiling.
+func TestDecodeArrowToSheet_RejectsTooManyColumns(t *testing.T) {
+	t.Parallel()
+	fields := make([]arrow.Field, dataframeMaxCols+1)
+	for i := range fields {
+		fields[i] = arrow.Field{Name: "c" + strings.TrimSpace(string(rune('0'+i%10))) + "_" + strings.Repeat("x", i/10+1), Type: arrow.BinaryTypes.String}
+	}
+	schema := arrow.NewSchema(fields, nil)
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		for i := range fields {
+			b.Field(i).(*array.StringBuilder).Append("")
+		}
+	})
+	_, err := decodeArrowToSheet(buf, "S")
+	if err == nil || !strings.Contains(err.Error(), "exceeds the per-sheet ceiling") {
+		t.Errorf("err = %v, want column-cap error", err)
+	}
+}
+
+// TestReadDataframeBytes_RejectsSecondStdinConsumer covers the case where another
+// flag (e.g. --styles) has already consumed stdin via the common Input resolver:
+// since --dataframe bypasses that resolver, the only thing keeping the two from
+// racing for an empty stream is the explicit StdinConsumed() check in
+// readDataframeBytes. Without that check, fangshuyu's report holds — both flags
+// silently accept '-' and one of them sees empty bytes downstream.
+func TestReadDataframeBytes_RejectsSecondStdinConsumer(t *testing.T) {
+	// process-wide cache must be reset so the test isn't served from a prior run.
+	saved := dataframeStdinCache
+	dataframeStdinCache = nil
+	t.Cleanup(func() { dataframeStdinCache = saved })
+
+	rctx := &common.RuntimeContext{}
+	rctx.MarkStdinConsumed()
+
+	_, err := readDataframeBytes(rctx, "-")
+	requireValidation(t, err, "stdin (-) can only be used by one flag")
+}
+
+// TestDataframe_EncodeRoundTrip checks --dataframe-out's encoder against its
+// own decoder: build a +table-get-shaped sheet map (the same one
+// readSheetAsSpec emits), encode to Arrow IPC, decode back via the put-side
+// decoder, and require the column types / formats / row values to match. If
+// any encoder choice drifts from what the decoder expects, the round-trip
+// breaks here long before a real put → get round-trip in production would.
+func TestDataframe_EncodeRoundTrip(t *testing.T) {
+	t.Parallel()
+	sheet := map[string]interface{}{
+		"name":    "S1",
+		"columns": []interface{}{"name", "qty", "price", "active", "ts"},
+		"dtypes": map[string]interface{}{
+			"name":   "object",
+			"qty":    "float64",
+			"price":  "float64",
+			"active": "bool",
+			"ts":     "datetime64[ns]",
+		},
+		"formats": map[string]interface{}{
+			// `@` is the writer convention for string columns; readSheetAsSpec
+			// strips it via isTextNumberFormat, so an Arrow file built from a
+			// real read won't carry @ either. Keep it absent here to mirror
+			// the production wire shape.
+			"price": "$#,##0.00",
+		},
+		"data": []interface{}{
+			[]interface{}{"alice", json.Number("42"), json.Number("19.95"), true, "2024-01-15"},
+			[]interface{}{"bob", nil, json.Number("8.5"), false, "2024-02-02"},
+		},
+	}
+	blob, err := encodeSheetMapToArrowIPC(sheet)
+	if err != nil {
+		t.Fatalf("encodeSheetMapToArrowIPC: %v", err)
+	}
+	spec, err := decodeArrowToSheet(blob, "S1")
+	if err != nil {
+		t.Fatalf("decodeArrowToSheet: %v", err)
+	}
+	wantTypes := []string{"string", "number", "number", "bool", "date"}
+	wantFormats := []string{"@", "", "$#,##0.00", "", "yyyy-mm-dd"}
+	if len(spec.Columns) != len(wantTypes) {
+		t.Fatalf("got %d columns, want %d", len(spec.Columns), len(wantTypes))
+	}
+	for i, w := range wantTypes {
+		if spec.Columns[i].Type != w {
+			t.Errorf("columns[%d].Type = %q, want %q", i, spec.Columns[i].Type, w)
+		}
+		if spec.Columns[i].Format != wantFormats[i] {
+			t.Errorf("columns[%d].Format = %q, want %q", i, spec.Columns[i].Format, wantFormats[i])
+		}
+	}
+	if len(spec.Rows) != 2 {
+		t.Fatalf("got %d rows, want 2", len(spec.Rows))
+	}
+	if spec.Rows[0][0] != "alice" {
+		t.Errorf("row0[name] = %#v, want alice", spec.Rows[0][0])
+	}
+	if n, ok := spec.Rows[0][1].(json.Number); !ok || n.String() != "42" {
+		t.Errorf("row0[qty] = %#v, want json.Number(\"42\")", spec.Rows[0][1])
+	}
+	if spec.Rows[0][3] != true {
+		t.Errorf("row0[active] = %#v, want true", spec.Rows[0][3])
+	}
+	if spec.Rows[0][4] != "2024-01-15" {
+		t.Errorf("row0[ts] = %#v, want 2024-01-15", spec.Rows[0][4])
+	}
+	// qty is null on row1, must come back as nil (not a zero-valued
+	// json.Number that would later round-trip as 0).
+	if spec.Rows[1][1] != nil {
+		t.Errorf("row1[qty] = %#v, want nil (null arrow cell)", spec.Rows[1][1])
+	}
+}
+
+// TestDataframe_EncodeAcceptsBothRowShapes pins the encoder against the two
+// shapes `sheet["data"]` actually arrives in: `[][]interface{}` from a live
+// readSheetAsSpec call (production), and `[]interface{}` from a JSON
+// unmarshal (round-trip / fixtures). Either must produce non-empty Arrow
+// output — early on the production shape silently fell through the
+// `[]interface{}` type assertion and we shipped a 0-row Arrow blob.
+func TestDataframe_EncodeAcceptsBothRowShapes(t *testing.T) {
+	t.Parallel()
+	base := func(data interface{}) map[string]interface{} {
+		return map[string]interface{}{
+			"name":    "S",
+			"columns": []interface{}{"city"},
+			"dtypes":  map[string]interface{}{"city": "object"},
+			"data":    data,
+		}
+	}
+	for label, data := range map[string]interface{}{
+		"production [][]interface{}": [][]interface{}{{"BJ"}, {"SH"}},
+		"unmarshal []interface{}":    []interface{}{[]interface{}{"BJ"}, []interface{}{"SH"}},
+	} {
+		blob, err := encodeSheetMapToArrowIPC(base(data))
+		if err != nil {
+			t.Errorf("%s: encode: %v", label, err)
+			continue
+		}
+		spec, err := decodeArrowToSheet(blob, "S")
+		if err != nil {
+			t.Errorf("%s: decode: %v", label, err)
+			continue
+		}
+		if len(spec.Rows) != 2 {
+			t.Errorf("%s: got %d rows, want 2", label, len(spec.Rows))
+		}
+	}
+}
+
+// TestDataframe_DtypeToInternalType pins the inverse of typeToDtype so
+// readSheetAsSpec's dtype labels recover the right internal type. Covers the
+// dtype families +table-get emits today plus the safe fallback for unknown
+// labels (string, lossless).
+func TestDataframe_DtypeToInternalType(t *testing.T) {
+	t.Parallel()
+	cases := map[string]string{
+		"float64":         "number",
+		"int64":           "number",
+		"Int64":           "number",
+		"bool":            "bool",
+		"boolean":         "bool",
+		"datetime64[ns]":  "date",
+		"datetime64[ms]":  "date",
+		"object":          "string",
+		"":                "string",
+		"weird-new-dtype": "string",
+	}
+	for in, want := range cases {
+		if got := dtypeToInternalType(in); got != want {
+			t.Errorf("dtypeToInternalType(%q) = %q, want %q", in, got, want)
+		}
+	}
+}
+
+// TestDataframe_BytesWriterSeeker confirms the in-memory WriteSeeker handles
+// the Seek-and-overwrite pattern ipc.NewFileWriter uses to patch the footer
+// offset: write some bytes, seek back to the middle, overwrite, end up with
+// the buffer reflecting the overwritten bytes (not a tail-extended duplicate).
+func TestDataframe_BytesWriterSeeker(t *testing.T) {
+	t.Parallel()
+	var w bytesWriterSeeker
+	if _, err := w.Write([]byte("hello world")); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := w.Seek(6, 0); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := w.Write([]byte("WORLD")); err != nil {
+		t.Fatal(err)
+	}
+	if got := string(w.buf); got != "hello WORLD" {
+		t.Errorf("buf = %q, want \"hello WORLD\"", got)
+	}
+}