📝 Add docstrings to feat/whiteboard-skill-cli-v0.2.12

Docstrings generation was requested by @cl900811.

* https://github.com/larksuite/cli/pull/1559#issuecomment-4787542760

The following files were modified:

* `shortcuts/whiteboard/whiteboard_query.go`
* `shortcuts/whiteboard/whiteboard_update.go`

.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");

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

cmd/event/list_test.go

@@ -26,6 +26,7 @@ func TestRunList_TextOutput(t *testing.T) {
 		"KEY", "AUTH", "PARAMS", "DESCRIPTION",
 		"im.message.receive_v1",
 		"im.message.message_read_v1",
+		"task.task.update_user_access_v2",
 	} {
 		if !strings.Contains(out, want) {
 			t.Errorf("list output missing %q; full output:\n%s", want, out)
@@ -55,4 +56,17 @@ func TestRunList_JSONOutput(t *testing.T) {
 			}
 		}
 	}
+
+	var foundTask bool
+	for _, row := range rows {
+		if row["key"] == "task.task.update_user_access_v2" {
+			foundTask = true
+			if row["single_consumer"] != true {
+				t.Errorf("task row single_consumer = %v, want true", row["single_consumer"])
+			}
+		}
+	}
+	if !foundTask {
+		t.Fatal("event list JSON missing task.task.update_user_access_v2")
+	}
 }

cmd/event/schema_test.go

@@ -96,6 +96,34 @@ func TestRunSchema_JSONOutput(t *testing.T) {
 	}
 }
 
+func TestRunSchema_TaskUpdateUserAccessJSON(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, &core.CliConfig{AppID: "test"})
+
+	if err := runSchema(f, "task.task.update_user_access_v2", true); err != nil {
+		t.Fatalf("runSchema json: %v", err)
+	}
+
+	var payload map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &payload); err != nil {
+		t.Fatalf("output is not valid JSON: %v\n%s", err, stdout.String())
+	}
+	if payload["jq_root_path"] != ".event" {
+		t.Errorf("jq_root_path = %v, want .event", payload["jq_root_path"])
+	}
+	if payload["single_consumer"] != true {
+		t.Errorf("single_consumer = %v, want true", payload["single_consumer"])
+	}
+	resolved := payload["resolved_output_schema"].(map[string]interface{})
+	props := resolved["properties"].(map[string]interface{})
+	eventProps := props["event"].(map[string]interface{})["properties"].(map[string]interface{})
+	if got := eventProps["task_guid"].(map[string]interface{})["format"]; got != "task_guid" {
+		t.Errorf("task_guid format = %v, want task_guid", got)
+	}
+	if _, ok := eventProps["event_types"].(map[string]interface{})["items"].(map[string]interface{})["enum"]; !ok {
+		t.Fatalf("event_types enum missing in schema: %#v", eventProps["event_types"])
+	}
+}
+
 func TestSchema_RendersSubscriptionKeyMarker(t *testing.T) {
 	const syntheticKey = "test.evt_sub"
 	t.Cleanup(func() { eventlib.UnregisterKeyForTest(syntheticKey) })

events/register.go

@@ -7,6 +7,7 @@ package events
 import (
 	"github.com/larksuite/cli/events/im"
 	"github.com/larksuite/cli/events/minutes"
+	"github.com/larksuite/cli/events/task"
 	"github.com/larksuite/cli/events/vc"
 	"github.com/larksuite/cli/events/whiteboard"
 	"github.com/larksuite/cli/internal/event"
@@ -17,6 +18,7 @@ func init() {
 	all := [][]event.KeyDefinition{
 		im.Keys(),
 		minutes.Keys(),
+		task.Keys(),
 		vc.Keys(),
 		whiteboard.Keys(),
 	}

events/task/native.go

@@ -0,0 +1,23 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package task
+
+// TaskUpdateUserAccessV2Data is the Task v2 update event payload under the
+// standard Lark V2 event envelope.
+type TaskUpdateUserAccessV2Data struct {
+	EventTypes []string `json:"event_types,omitempty" desc:"Task commit types included in this event" enum:"task_create,task_deleted,task_summary_update,task_desc_update,task_assignees_update,task_followers_update,task_reminders_update,task_start_due_update,task_completed_update"`
+	TaskGUID   string   `json:"task_guid,omitempty" desc:"Task GUID that changed" kind:"task_guid"`
+}
+
+var taskUpdateUserAccessCommitTypes = []string{
+	"task_create",
+	"task_deleted",
+	"task_summary_update",
+	"task_desc_update",
+	"task_assignees_update",
+	"task_followers_update",
+	"task_reminders_update",
+	"task_start_due_update",
+	"task_completed_update",
+}

events/task/preconsume.go

@@ -0,0 +1,32 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package task
+
+import (
+	"context"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/event"
+)
+
+const taskSubscriptionPath = "/open-apis/task/v2/task_v2/task_subscription?user_id_type=open_id"
+
+func taskSubscriptionPreConsume(ctx context.Context, rt event.APIClient, _ map[string]string) (func() error, error) {
+	if rt == nil {
+		return nil, errs.NewInternalError(errs.SubtypeUnknown,
+			"runtime API client is required for pre-consume subscription")
+	}
+
+	if _, err := rt.CallAPI(ctx, "POST", taskSubscriptionPath, nil); err != nil {
+		if _, ok := errs.ProblemOf(err); ok {
+			return nil, err
+		}
+		return nil, errs.NewNetworkError(
+			errs.SubtypeNetworkTransport,
+			"failed to subscribe task event",
+		).WithCause(err)
+	}
+
+	return nil, nil
+}

events/task/preconsume_test.go

@@ -0,0 +1,119 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package task
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+)
+
+type stubAPIClient struct {
+	err error
+
+	method string
+	path   string
+	body   interface{}
+	calls  int
+}
+
+func (s *stubAPIClient) CallAPI(_ context.Context, method, path string, body interface{}) (json.RawMessage, error) {
+	s.method = method
+	s.path = path
+	s.body = body
+	s.calls++
+	if s.err != nil {
+		return nil, s.err
+	}
+	return json.RawMessage(`{"code":0,"msg":"success","data":{}}`), nil
+}
+
+func TestTaskSubscriptionPreConsumeCallsSubscribeAPI(t *testing.T) {
+	rt := &stubAPIClient{}
+	cleanup, err := taskSubscriptionPreConsume(context.Background(), rt, nil)
+	if err != nil {
+		t.Fatalf("taskSubscriptionPreConsume error = %v", err)
+	}
+	if cleanup != nil {
+		t.Fatal("cleanup = non-nil, want nil because task subscription has no unsubscribe API")
+	}
+	if rt.calls != 1 {
+		t.Fatalf("calls = %d, want 1", rt.calls)
+	}
+	if rt.method != "POST" {
+		t.Errorf("method = %q, want POST", rt.method)
+	}
+	if rt.path != taskSubscriptionPath {
+		t.Errorf("path = %q, want %q", rt.path, taskSubscriptionPath)
+	}
+	if rt.body != nil {
+		t.Errorf("body = %#v, want nil", rt.body)
+	}
+}
+
+func TestTaskSubscriptionPreConsumeRequiresRuntime(t *testing.T) {
+	_, err := taskSubscriptionPreConsume(context.Background(), nil, nil)
+	if err == nil {
+		t.Fatal("expected error")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed error, got %T: %v", err, err)
+	}
+	if p.Category != errs.CategoryInternal {
+		t.Errorf("category = %s, want %s", p.Category, errs.CategoryInternal)
+	}
+	if p.Subtype != errs.SubtypeUnknown {
+		t.Errorf("subtype = %s, want %s", p.Subtype, errs.SubtypeUnknown)
+	}
+}
+
+func TestTaskSubscriptionPreConsumePassesThroughAPIError(t *testing.T) {
+	wantErr := errs.NewValidationError(errs.SubtypeFailedPrecondition, "subscription already exists")
+	rt := &stubAPIClient{err: wantErr}
+
+	_, err := taskSubscriptionPreConsume(context.Background(), rt, nil)
+	if err != wantErr {
+		t.Fatalf("err identity changed: got %T %v, want original %T %v", err, err, wantErr, wantErr)
+	}
+	if !errors.Is(err, wantErr) {
+		t.Fatalf("err = %v, want %v", err, wantErr)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed error, got %T: %v", err, err)
+	}
+	if p.Category != errs.CategoryValidation {
+		t.Errorf("category = %s, want %s", p.Category, errs.CategoryValidation)
+	}
+	if p.Subtype != errs.SubtypeFailedPrecondition {
+		t.Errorf("subtype = %s, want %s", p.Subtype, errs.SubtypeFailedPrecondition)
+	}
+}
+
+func TestTaskSubscriptionPreConsumeWrapsUntypedAPIError(t *testing.T) {
+	cause := errors.New("connection reset")
+	rt := &stubAPIClient{err: cause}
+
+	_, err := taskSubscriptionPreConsume(context.Background(), rt, nil)
+	if err == nil {
+		t.Fatal("expected error")
+	}
+	if !errors.Is(err, cause) {
+		t.Fatalf("err = %v, want cause %v", err, cause)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("expected typed error, got %T: %v", err, err)
+	}
+	if p.Category != errs.CategoryNetwork {
+		t.Errorf("category = %s, want %s", p.Category, errs.CategoryNetwork)
+	}
+	if p.Subtype != errs.SubtypeNetworkTransport {
+		t.Errorf("subtype = %s, want %s", p.Subtype, errs.SubtypeNetworkTransport)
+	}
+}

events/task/register.go

@@ -0,0 +1,33 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package task registers Task-domain EventKeys.
+package task
+
+import (
+	"reflect"
+
+	"github.com/larksuite/cli/internal/event"
+)
+
+const eventTypeTaskUpdateUserAccessV2 = "task.task.update_user_access_v2"
+
+// Keys returns all Task-domain EventKey definitions.
+func Keys() []event.KeyDefinition {
+	return []event.KeyDefinition{
+		{
+			Key:         eventTypeTaskUpdateUserAccessV2,
+			DisplayName: "Task updated",
+			Description: "Triggered when tasks visible to the current user or app are created, deleted, or updated",
+			EventType:   eventTypeTaskUpdateUserAccessV2,
+			Schema: event.SchemaDef{
+				Native: &event.SchemaSpec{Type: reflect.TypeOf(TaskUpdateUserAccessV2Data{})},
+			},
+			PreConsume:            taskSubscriptionPreConsume,
+			Scopes:                []string{"task:task:read"},
+			AuthTypes:             []string{"user", "bot"},
+			RequiredConsoleEvents: []string{eventTypeTaskUpdateUserAccessV2},
+			SingleConsumer:        true,
+		},
+	}
+}

events/task/register_test.go

@@ -0,0 +1,95 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package task
+
+import (
+	"encoding/json"
+	"reflect"
+	"testing"
+
+	"github.com/larksuite/cli/internal/event"
+	"github.com/larksuite/cli/internal/event/schemas"
+)
+
+func TestKeysTaskUpdateUserAccessMetadata(t *testing.T) {
+	keys := Keys()
+	if len(keys) != 1 {
+		t.Fatalf("len(Keys()) = %d, want 1", len(keys))
+	}
+
+	def := keys[0]
+	if def.Key != eventTypeTaskUpdateUserAccessV2 {
+		t.Errorf("Key = %q, want %q", def.Key, eventTypeTaskUpdateUserAccessV2)
+	}
+	if def.EventType != eventTypeTaskUpdateUserAccessV2 {
+		t.Errorf("EventType = %q, want %q", def.EventType, eventTypeTaskUpdateUserAccessV2)
+	}
+	if def.Schema.Native == nil {
+		t.Fatal("Schema.Native is nil")
+	}
+	if def.Schema.Native.Type != reflect.TypeOf(TaskUpdateUserAccessV2Data{}) {
+		t.Errorf("native type = %v, want TaskUpdateUserAccessV2Data", def.Schema.Native.Type)
+	}
+	if def.Process != nil {
+		t.Fatal("Native Task EventKey must not set Process")
+	}
+	if def.PreConsume == nil {
+		t.Fatal("PreConsume is nil")
+	}
+	if !def.SingleConsumer {
+		t.Fatal("SingleConsumer = false, want true")
+	}
+	if !reflect.DeepEqual(def.Scopes, []string{"task:task:read"}) {
+		t.Errorf("Scopes = %#v", def.Scopes)
+	}
+	if !reflect.DeepEqual(def.AuthTypes, []string{"user", "bot"}) {
+		t.Errorf("AuthTypes = %#v", def.AuthTypes)
+	}
+	if !reflect.DeepEqual(def.RequiredConsoleEvents, []string{eventTypeTaskUpdateUserAccessV2}) {
+		t.Errorf("RequiredConsoleEvents = %#v", def.RequiredConsoleEvents)
+	}
+}
+
+func TestTaskUpdateUserAccessSchemaAnnotations(t *testing.T) {
+	raw := schemas.WrapV2Envelope(schemas.FromType(reflect.TypeOf(TaskUpdateUserAccessV2Data{})))
+	var schema map[string]interface{}
+	if err := json.Unmarshal(raw, &schema); err != nil {
+		t.Fatalf("unmarshal schema: %v", err)
+	}
+
+	eventProps := schema["properties"].(map[string]interface{})["event"].(map[string]interface{})["properties"].(map[string]interface{})
+	taskGUID := eventProps["task_guid"].(map[string]interface{})
+	if got := taskGUID["format"]; got != "task_guid" {
+		t.Errorf("task_guid format = %v, want task_guid", got)
+	}
+
+	eventTypes := eventProps["event_types"].(map[string]interface{})
+	items := eventTypes["items"].(map[string]interface{})
+	rawEnum, ok := items["enum"].([]interface{})
+	if !ok {
+		t.Fatalf("event_types item enum missing: %#v", items["enum"])
+	}
+	got := make(map[string]bool, len(rawEnum))
+	for _, v := range rawEnum {
+		got[v.(string)] = true
+	}
+	for _, want := range taskUpdateUserAccessCommitTypes {
+		if !got[want] {
+			t.Errorf("event_types enum missing %q; enum=%v", want, rawEnum)
+		}
+	}
+}
+
+func TestTaskUpdateUserAccessRegistersCleanly(t *testing.T) {
+	const key = eventTypeTaskUpdateUserAccessV2
+	event.UnregisterKeyForTest(key)
+	t.Cleanup(func() { event.UnregisterKeyForTest(key) })
+
+	for _, def := range Keys() {
+		event.RegisterKey(def)
+	}
+	if _, ok := event.Lookup(key); !ok {
+		t.Fatalf("event.Lookup(%q) not registered", key)
+	}
+}

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/im/builders_test.go

@@ -469,6 +469,29 @@ func TestShortcutValidateBranches(t *testing.T) {
 		}
 	})
 
+	t.Run("ImMessagesSend audio rejects non-opus local file", func(t *testing.T) {
+		runtime := newTestRuntimeContext(t, map[string]string{
+			"chat-id": "oc_123",
+			"audio":   "./voice.mp3",
+		}, nil)
+		err := ImMessagesSend.Validate(context.Background(), runtime)
+		if err == nil || !strings.Contains(err.Error(), "--audio supports only Opus audio files") {
+			t.Fatalf("ImMessagesSend.Validate() error = %v", err)
+		}
+	})
+
+	t.Run("ImMessagesSend audio accepts opus and ogg local files", func(t *testing.T) {
+		for _, audio := range []string{"./voice.opus", "./voice.ogg"} {
+			runtime := newTestRuntimeContext(t, map[string]string{
+				"chat-id": "oc_123",
+				"audio":   audio,
+			}, nil)
+			if err := ImMessagesSend.Validate(context.Background(), runtime); err != nil {
+				t.Fatalf("ImMessagesSend.Validate(%q) unexpected error = %v", audio, err)
+			}
+		}
+	})
+
 	t.Run("ImMessagesSend conflicting explicit msg-type", func(t *testing.T) {
 		runtime := newTestRuntimeContext(t, map[string]string{
 			"chat-id":  "oc_123",
@@ -492,6 +515,17 @@ func TestShortcutValidateBranches(t *testing.T) {
 		}
 	})
 
+	t.Run("ImMessagesReply audio rejects non-opus local file", func(t *testing.T) {
+		runtime := newTestRuntimeContext(t, map[string]string{
+			"message-id": "om_123",
+			"audio":      "./voice.mp3",
+		}, nil)
+		err := ImMessagesReply.Validate(context.Background(), runtime)
+		if err == nil || !strings.Contains(err.Error(), "--audio supports only Opus audio files") {
+			t.Fatalf("ImMessagesReply.Validate() error = %v", err)
+		}
+	})
+
 	t.Run("ImThreadsMessagesList invalid thread", func(t *testing.T) {
 		runtime := newTestRuntimeContext(t, map[string]string{
 			"thread": "bad_thread",

shortcuts/im/helpers.go

@@ -1048,6 +1048,42 @@ func detectIMFileType(filePath string) string {
 	}
 }
 
+const (
+	audioMessageInputDesc = "audio file key (file_xxx), URL, or cwd-relative local path for a voice message (absolute paths and .. are rejected); local paths and URLs must be Opus (.opus or Ogg Opus .ogg). For mp3/wav, convert to .opus first, or use --file to send as an attachment"
+	audioMessageHint      = "Convert non-Opus audio to .opus and use --audio for a voice message, for example: ffmpeg -i input.mp3 -acodec libopus -ac 1 -ar 16000 output.opus. To send the original mp3/wav as an attachment, use --file instead."
+)
+
+func validateAudioMessageInput(flagName, value string) error {
+	value = strings.TrimSpace(value)
+	if value == "" || isMediaKey(value) {
+		return nil
+	}
+
+	ext := audioInputExt(value)
+	if ext == "" {
+		return nil
+	}
+	if ext == ".opus" || ext == ".ogg" {
+		return nil
+	}
+	return errs.NewValidationError(
+		errs.SubtypeInvalidArgument,
+		"%s supports only Opus audio files for audio messages, such as .opus files or Ogg Opus (.ogg) files",
+		flagName,
+	).WithParam(flagName).WithHint("%s", audioMessageHint)
+}
+
+func audioInputExt(value string) string {
+	if isURL(value) {
+		parsed, err := url.Parse(value)
+		if err != nil {
+			return ""
+		}
+		return strings.ToLower(path.Ext(parsed.Path))
+	}
+	return strings.ToLower(filepath.Ext(value))
+}
+
 const maxImageUploadSize = 5 * 1024 * 1024  // 5MB — Lark API limit for images
 const maxFileUploadSize = 100 * 1024 * 1024 // 100MB — Lark API limit for files
 

shortcuts/im/helpers_test.go

@@ -13,6 +13,7 @@ import (
 	"strings"
 	"testing"
 
+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/shortcuts/common"
 )
 
@@ -50,6 +51,56 @@ func TestDetectIMFileType(t *testing.T) {
 	}
 }
 
+func TestValidateAudioMessageInput(t *testing.T) {
+	tests := []struct {
+		name    string
+		value   string
+		wantErr bool
+	}{
+		{name: "empty", value: ""},
+		{name: "existing file key", value: "file_abc"},
+		{name: "opus file", value: "./voice.opus"},
+		{name: "ogg opus file", value: "./voice.ogg"},
+		{name: "uppercase opus", value: "./VOICE.OPUS"},
+		{name: "mp3 local file", value: "./voice.mp3", wantErr: true},
+		{name: "wav local file", value: "./voice.wav", wantErr: true},
+		{name: "extensionless local path", value: "./voice"},
+		{name: "opus url", value: "https://example.com/voice.opus?download=1"},
+		{name: "ogg url", value: "https://example.com/voice.ogg?download=1"},
+		{name: "mp3 url", value: "https://example.com/voice.mp3?download=1", wantErr: true},
+		{name: "extensionless url", value: "https://example.com/download?id=1"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := validateAudioMessageInput("--audio", tt.value)
+			if tt.wantErr {
+				if err == nil || !strings.Contains(err.Error(), "--audio supports only Opus audio files") {
+					t.Fatalf("validateAudioMessageInput(%q) error = %v", tt.value, err)
+				}
+				p, ok := errs.ProblemOf(err)
+				if !ok {
+					t.Fatalf("validateAudioMessageInput(%q) error is not typed: %v", tt.value, err)
+				}
+				if p.Category != errs.CategoryValidation || p.Subtype != errs.SubtypeInvalidArgument {
+					t.Fatalf("ProblemOf(%q) = category %q subtype %q", tt.value, p.Category, p.Subtype)
+				}
+				validationErr, ok := err.(*errs.ValidationError)
+				if !ok || validationErr.Param != "--audio" {
+					t.Fatalf("validateAudioMessageInput(%q) param = %q, want --audio", tt.value, validationErr.Param)
+				}
+				if !strings.Contains(p.Hint, "use --file") || !strings.Contains(p.Hint, "ffmpeg") {
+					t.Fatalf("validateAudioMessageInput(%q) hint = %q, want --file and ffmpeg guidance", tt.value, p.Hint)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("validateAudioMessageInput(%q) unexpected error = %v", tt.value, err)
+			}
+		})
+	}
+}
+
 // TestSplitCSV covers the shared helper that replaced the three identical functions
 func TestSplitCSV(t *testing.T) {
 	tests := []struct {

shortcuts/im/im_messages_reply.go

@@ -33,7 +33,7 @@ var ImMessagesReply = common.Shortcut{
 		{Name: "file", Desc: "file key (file_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected)"},
 		{Name: "video", Desc: "video file key (file_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected); must be used together with --video-cover"},
 		{Name: "video-cover", Desc: "video cover image key (img_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected); required when using --video"},
-		{Name: "audio", Desc: "audio file key (file_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected)"},
+		{Name: "audio", Desc: audioMessageInputDesc},
 		{Name: "reply-in-thread", Type: "bool", Desc: "reply in thread (message appears in thread stream instead of main chat)"},
 		{Name: "idempotency-key", Desc: "idempotency key (prevents duplicate sends)"},
 	},
@@ -100,6 +100,9 @@ var ImMessagesReply = common.Shortcut{
 				return err
 			}
 		}
+		if err := validateAudioMessageInput("--audio", audioKey); err != nil {
+			return err
+		}
 
 		if messageId == "" {
 			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--message-id is required (om_xxx)").WithParam("--message-id")

shortcuts/im/im_messages_send.go

@@ -37,7 +37,7 @@ var ImMessagesSend = common.Shortcut{
 		{Name: "file", Desc: "file key (file_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected)"},
 		{Name: "video", Desc: "video file key (file_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected); must be used together with --video-cover"},
 		{Name: "video-cover", Desc: "video cover image key (img_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected); required when using --video"},
-		{Name: "audio", Desc: "audio file key (file_xxx), URL, or cwd-relative local path (absolute paths and .. are rejected)"},
+		{Name: "audio", Desc: audioMessageInputDesc},
 	},
 	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
 		chatFlag := runtime.Str("chat-id")
@@ -112,6 +112,9 @@ var ImMessagesSend = common.Shortcut{
 				return err
 			}
 		}
+		if err := validateAudioMessageInput("--audio", audioKey); err != nil {
+			return err
+		}
 
 		if err := common.ExactlyOneTyped(runtime, "chat-id", "user-id"); err != nil {
 			return err

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/task/shortcuts.go

@@ -242,7 +242,6 @@ func Shortcuts() []common.Shortcut {
 		GetMyTasks,
 		GetRelatedTasks,
 		SearchTask,
-		SubscribeTaskEvent,
 		UploadAttachmentTask,
 		CreateTasklist,
 		SearchTasklist,

shortcuts/task/task_subscribe_event.go

@@ -1,40 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package task
-
-import (
-	"context"
-	"fmt"
-	"io"
-	"net/http"
-
-	"github.com/larksuite/cli/shortcuts/common"
-)
-
-var SubscribeTaskEvent = common.Shortcut{
-	Service:     "task",
-	Command:     "+subscribe-event",
-	Description: "subscribe to task events",
-	Risk:        "write",
-	Scopes:      []string{"task:task:read"},
-	AuthTypes:   []string{"user", "bot"},
-	HasFormat:   true,
-	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		return common.NewDryRunAPI().
-			POST("/open-apis/task/v2/task_v2/task_subscription").
-			Params(map[string]interface{}{"user_id_type": "open_id"})
-	},
-	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		params := map[string]interface{}{"user_id_type": "open_id"}
-		if _, err := callTaskAPITyped(runtime, http.MethodPost, "/open-apis/task/v2/task_v2/task_subscription", params, nil); err != nil {
-			return err
-		}
-
-		outData := map[string]interface{}{"ok": true}
-		runtime.OutFormat(outData, nil, func(w io.Writer) {
-			fmt.Fprintln(w, "✅ Task event subscription created successfully!")
-		})
-		return nil
-	},
-}

shortcuts/task/task_subscribe_event_test.go

@@ -1,163 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package task
-
-import (
-	"errors"
-	"strings"
-	"testing"
-
-	"github.com/spf13/cobra"
-
-	"github.com/larksuite/cli/errs"
-	"github.com/larksuite/cli/internal/httpmock"
-	"github.com/larksuite/cli/internal/output"
-	"github.com/larksuite/cli/shortcuts/common"
-)
-
-func TestSubscribeTaskEvent(t *testing.T) {
-	tests := []struct {
-		name      string
-		mode      string
-		args      []string
-		register  func(*httpmock.Registry)
-		wantErr   bool
-		wantParts []string
-	}{
-		{
-			name: "execute json (user identity)",
-			mode: "execute",
-			args: []string{"+subscribe-event", "--as", "user", "--format", "json"},
-			register: func(reg *httpmock.Registry) {
-				reg.Register(&httpmock.Stub{
-					Method: "POST",
-					URL:    "/open-apis/task/v2/task_v2/task_subscription",
-					Body: map[string]interface{}{
-						"code": 0,
-						"msg":  "success",
-						"data": map[string]interface{}{},
-					},
-				})
-			},
-			wantParts: []string{`"ok": true`},
-		},
-		{
-			name: "execute json (bot identity)",
-			mode: "execute",
-			args: []string{"+subscribe-event", "--as", "bot", "--format", "json"},
-			register: func(reg *httpmock.Registry) {
-				reg.Register(&httpmock.Stub{
-					Method: "POST",
-					URL:    "/open-apis/task/v2/task_v2/task_subscription",
-					Body: map[string]interface{}{
-						"code": 0,
-						"msg":  "success",
-						"data": map[string]interface{}{},
-					},
-				})
-			},
-			wantParts: []string{`"ok": true`},
-		},
-		{
-			name: "execute api error",
-			mode: "execute",
-			args: []string{"+subscribe-event", "--as", "bot", "--format", "json"},
-			register: func(reg *httpmock.Registry) {
-				reg.Register(&httpmock.Stub{
-					Method: "POST",
-					URL:    "/open-apis/task/v2/task_v2/task_subscription",
-					Body: map[string]interface{}{
-						"code": 401,
-						"msg":  "Unauthorized",
-						"error": map[string]interface{}{
-							"log_id": "test-log-id",
-						},
-					},
-				})
-			},
-			wantErr:   true,
-			wantParts: []string{"Unauthorized"},
-		},
-		{
-			name:      "dry run",
-			mode:      "dryrun",
-			wantParts: []string{"POST /open-apis/task/v2/task_v2/task_subscription"},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			switch tt.mode {
-			case "execute":
-				f, stdout, _, reg := taskShortcutTestFactory(t)
-				warmTenantToken(t, f, reg)
-				if tt.register != nil {
-					tt.register(reg)
-				}
-
-				err := runMountedTaskShortcut(t, SubscribeTaskEvent, tt.args, f, stdout)
-				if tt.wantErr {
-					if err == nil {
-						t.Fatal("expected error, got nil")
-					}
-					out := err.Error()
-					for _, want := range tt.wantParts {
-						if !strings.Contains(out, want) {
-							t.Fatalf("error missing %q: %s", want, out)
-						}
-					}
-					return
-				}
-				if err != nil {
-					t.Fatalf("runMountedTaskShortcut() error = %v", err)
-				}
-
-				out := stdout.String()
-				outNorm := strings.ReplaceAll(out, `":"`, `": "`)
-				for _, want := range tt.wantParts {
-					if !strings.Contains(out, want) && !strings.Contains(outNorm, want) {
-						t.Fatalf("output missing %q: %s", want, out)
-					}
-				}
-			case "dryrun":
-				runtime := common.TestNewRuntimeContextWithIdentity(&cobra.Command{Use: "test"}, taskTestConfig(t), "user")
-				out := SubscribeTaskEvent.DryRun(nil, runtime).Format()
-				for _, want := range tt.wantParts {
-					if !strings.Contains(out, want) {
-						t.Fatalf("dry run output missing %q: %s", want, out)
-					}
-				}
-			}
-		})
-	}
-}
-
-// TestSubscribeTaskEvent_MalformedResponse covers the parse-response arm: a 200
-// with an unparseable body surfaces a typed internal invalid_response error
-// (exit 5).
-func TestSubscribeTaskEvent_MalformedResponse(t *testing.T) {
-	f, stdout, _, reg := taskShortcutTestFactory(t)
-	warmTenantToken(t, f, reg)
-
-	reg.Register(&httpmock.Stub{
-		Method:  "POST",
-		URL:     "/open-apis/task/v2/task_v2/task_subscription",
-		Status:  200,
-		RawBody: []byte("{not-json"),
-	})
-
-	args := []string{"+subscribe-event", "--as", "bot", "--format", "json"}
-	err := runMountedTaskShortcut(t, SubscribeTaskEvent, args, f, stdout)
-
-	var ie *errs.InternalError
-	if !errors.As(err, &ie) {
-		t.Fatalf("err = %T, want *errs.InternalError; err = %v", err, err)
-	}
-	if ie.Subtype != errs.SubtypeInvalidResponse {
-		t.Errorf("subtype = %q, want %q", ie.Subtype, errs.SubtypeInvalidResponse)
-	}
-	if got := output.ExitCodeOf(err); got != output.ExitInternal {
-		t.Errorf("exit code = %d, want %d", got, output.ExitInternal)
-	}
-}

shortcuts/whiteboard/whiteboard_query.go

@@ -5,6 +5,7 @@ package whiteboard
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -21,40 +22,54 @@ import (
 )
 
 const (
+	// WhiteboardQueryAsImage exports a whiteboard preview image.
 	WhiteboardQueryAsImage = "image"
-	WhiteboardQueryAsCode  = "code"
-	WhiteboardQueryAsRaw   = "raw"
+	// WhiteboardQueryAsSvg exports a whiteboard as SVG.
+	WhiteboardQueryAsSvg = "svg"
+	// WhiteboardQueryAsCode exports Mermaid or PlantUML source extracted from the whiteboard.
+	WhiteboardQueryAsCode = "code"
+	// WhiteboardQueryAsRaw exports the raw whiteboard node payload.
+	WhiteboardQueryAsRaw = "raw"
 )
 
+// SyntaxType identifies the diagram syntax extracted from whiteboard code blocks.
 type SyntaxType int
 
 const (
+	// SyntaxTypePlantUML marks PlantUML code blocks.
 	SyntaxTypePlantUML SyntaxType = 1
-	SyntaxTypeMermaid  SyntaxType = 2
+	// SyntaxTypeMermaid marks Mermaid code blocks.
+	SyntaxTypeMermaid SyntaxType = 2
 )
 
+// SyntaxTypeNameMap maps whiteboard syntax types to their CLI output names.
 var SyntaxTypeNameMap = map[SyntaxType]string{
 	SyntaxTypePlantUML: "plantuml",
 	SyntaxTypeMermaid:  "mermaid",
 }
 
+// SyntaxTypeExtensionMap maps whiteboard syntax types to their default file extensions.
 var SyntaxTypeExtensionMap = map[SyntaxType]string{
 	SyntaxTypePlantUML: ".puml",
 	SyntaxTypeMermaid:  ".mmd",
 }
 
+// String returns the CLI-facing name for the syntax type.
 func (s SyntaxType) String() string {
 	return SyntaxTypeNameMap[s]
 }
 
+// ExtensionName returns the default file extension for the syntax type.
 func (s SyntaxType) ExtensionName() string {
 	return SyntaxTypeExtensionMap[s]
 }
 
+// IsValid reports whether the syntax type is one of the supported whiteboard code syntaxes.
 func (s SyntaxType) IsValid() bool {
 	return s == SyntaxTypePlantUML || s == SyntaxTypeMermaid
 }
 
+// WhiteboardQuery registers the `whiteboard +query` shortcut.
 var WhiteboardQuery = common.Shortcut{
 	Service:     "whiteboard",
 	Command:     "+query",
@@ -64,8 +79,8 @@ var WhiteboardQuery = common.Shortcut{
 	AuthTypes:   []string{"user", "bot"},
 	Flags: []common.Flag{
 		{Name: "whiteboard-token", Desc: "whiteboard token of the whiteboard. You will need read permission to download preview image.", Required: true},
-		{Name: "output_as", Desc: "output whiteboard as: image | code | raw.", Required: true},
-		{Name: "output", Desc: "output directory. It is required when output as image. If not specified when --output_as code/raw, it will output directly.", Required: false},
+		{Name: "output_as", Desc: "output whiteboard as: image | svg | code | raw.", Required: true},
+		{Name: "output", Desc: "output directory. It is required when output as image. If not specified when --output_as svg/code/raw, it will output directly.", Required: false},
 		{Name: "overwrite", Desc: "overwrite existing file if it exists", Required: false, Type: "bool"},
 	},
 	HasFormat: true,
@@ -86,8 +101,8 @@ var WhiteboardQuery = common.Shortcut{
 		}
 
 		as := runtime.Str("output_as")
-		if as != WhiteboardQueryAsImage && as != WhiteboardQueryAsCode && as != WhiteboardQueryAsRaw {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--output_as flag must be one of: image | code | raw").WithParam("--output_as")
+		if as != WhiteboardQueryAsImage && as != WhiteboardQueryAsSvg && as != WhiteboardQueryAsCode && as != WhiteboardQueryAsRaw {
+			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--output_as flag must be one of: image | svg | code | raw").WithParam("--output_as")
 		}
 		return nil
 	},
@@ -107,8 +122,13 @@ var WhiteboardQuery = common.Shortcut{
 			return common.NewDryRunAPI().
 				GET(fmt.Sprintf("/open-apis/board/v1/whiteboards/%s/nodes", common.MaskToken(url.PathEscape(token)))).
 				Desc("Extract raw nodes structure from given whiteboard")
+		case WhiteboardQueryAsSvg:
+			return common.NewDryRunAPI().
+				POST(fmt.Sprintf("/open-apis/board/v1/whiteboards/%s/export", common.MaskToken(url.PathEscape(token)))).
+				Body(map[string]string{"export_type": "svg"}).
+				Desc("Export SVG of given whiteboard")
 		default:
-			return common.NewDryRunAPI().Desc("invalid --output_as flag, must be one of: image | code | raw")
+			return common.NewDryRunAPI().Desc("invalid --output_as flag, must be one of: image | svg | code | raw")
 		}
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
@@ -119,17 +139,110 @@ var WhiteboardQuery = common.Shortcut{
 		switch as {
 		case WhiteboardQueryAsImage:
 			return exportWhiteboardPreview(ctx, runtime, token, outDir)
+		case WhiteboardQueryAsSvg:
+			return exportWhiteboardSvg(runtime, token, outDir)
 		case WhiteboardQueryAsCode:
 			return exportWhiteboardCode(runtime, token, outDir)
 		case WhiteboardQueryAsRaw:
 			return exportWhiteboardRaw(runtime, token, outDir)
 		default:
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--output_as flag must be one of: image | code | raw").WithParam("--output_as")
+			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--output_as flag must be one of: image | svg | code | raw").WithParam("--output_as")
 		}
 
 	},
 }
 
+// exportReq defines the request body for whiteboard export APIs.
+type exportReq struct {
+	ExportType string `json:"export_type"`
+}
+
+// exportResp models the whiteboard export response envelope.
+type exportResp struct {
+	Code int    `json:"code"`
+	Msg  string `json:"msg"`
+	Data struct {
+		Content  string `json:"content"`
+		MimeType string `json:"mime_type"`
+	} `json:"data"`
+}
+
+// exportWhiteboardSvg exports a whiteboard as SVG and writes the result to stdout or a file.
+// It requests the SVG export for the given whiteboard token and saves the decoded content when an output path is provided.
+func exportWhiteboardSvg(runtime *common.RuntimeContext, wbToken, outDir string) error {
+	reqBody := exportReq{ExportType: "svg"}
+	req := &larkcore.ApiReq{
+		HttpMethod: http.MethodPost,
+		ApiPath:    fmt.Sprintf("/open-apis/board/v1/whiteboards/%s/export", url.PathEscape(wbToken)),
+		Body:       reqBody,
+	}
+
+	resp, err := runtime.DoAPI(req)
+	if err != nil {
+		return wrapWbNetworkErr(err, "export whiteboard svg failed: %v", err)
+	}
+
+	var exportData exportResp
+	if err := json.Unmarshal(resp.RawBody, &exportData); err == nil {
+		if exportData.Code != 0 {
+			subtype := errs.SubtypeUnknown
+			if resp.StatusCode == http.StatusNotFound {
+				subtype = errs.SubtypeNotFound
+			}
+			return errs.NewAPIError(subtype, "export whiteboard svg failed: %s", exportData.Msg).WithCode(exportData.Code)
+		}
+	} else if resp.StatusCode == http.StatusOK {
+		return errs.NewInternalError(errs.SubtypeInvalidResponse, "parse export response failed: %v", err).WithCause(err)
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		body := common.TruncateStr(strings.TrimSpace(string(resp.RawBody)), 500)
+		if resp.StatusCode >= 500 {
+			return errs.NewNetworkError(errs.SubtypeNetworkServer, "export whiteboard svg failed: HTTP %d: %s", resp.StatusCode, body).
+				WithCode(resp.StatusCode).
+				WithRetryable()
+		}
+		subtype := errs.SubtypeUnknown
+		if resp.StatusCode == http.StatusNotFound {
+			subtype = errs.SubtypeNotFound
+		}
+		return errs.NewAPIError(subtype, "export whiteboard svg failed: HTTP %d: %s", resp.StatusCode, body).
+			WithCode(resp.StatusCode)
+	}
+
+	svgBytes, err := base64.StdEncoding.DecodeString(exportData.Data.Content)
+	if err != nil {
+		return errs.NewInternalError(errs.SubtypeInvalidResponse, "decode svg base64 failed: %v", err).WithCause(err)
+	}
+
+	if outDir == "" {
+		runtime.OutFormat(map[string]interface{}{
+			"svg_content": string(svgBytes),
+		}, nil, func(w io.Writer) {
+			fmt.Fprintf(w, "%s\n", string(svgBytes))
+		})
+		return nil
+	}
+
+	finalPath, size, err := saveOutputFile(outDir, ".svg", wbToken, runtime, bytes.NewReader(svgBytes))
+	if err != nil {
+		return err
+	}
+
+	runtime.OutFormat(map[string]interface{}{
+		"svg_path":   finalPath,
+		"size_bytes": size,
+	}, nil, func(w io.Writer) {
+		fmt.Fprintf(w, "SVG saved to %s\n", finalPath)
+		fmt.Fprintf(w, "File size: %d bytes", size)
+	})
+	return nil
+}
+
+// exportWhiteboardPreview downloads a whiteboard preview image and saves it as a PNG file.
+//
+// It reports the saved file path and image size on success.
+// Returns an error if the API request fails, the response is rejected, or the file cannot be saved.
 func exportWhiteboardPreview(ctx context.Context, runtime *common.RuntimeContext, wbToken, outDir string) error {
 	req := &larkcore.ApiReq{
 		HttpMethod: http.MethodGet,
@@ -331,6 +444,9 @@ func exportWhiteboardRaw(runtime *common.RuntimeContext, wbToken, outDir string)
 	return nil
 }
 
+// saveOutputFile writes exported content to a file or directory and returns the final path and written size.
+// If outPath is a directory, it creates a file named whiteboard_<token><ext>. If outPath is a file path,
+// it adjusts the file extension to ext, validates the path, and respects the overwrite flag.
 func saveOutputFile(outPath, ext, token string, runtime *common.RuntimeContext, data io.Reader) (string, int64, error) {
 	// Step 1: Get final output path
 	info, err := runtime.FileIO().Stat(outPath)
@@ -367,6 +483,8 @@ func saveOutputFile(outPath, ext, token string, runtime *common.RuntimeContext,
 	switch ext {
 	case ".png":
 		contentType = "image/png"
+	case ".svg":
+		contentType = "image/svg+xml"
 	case ".json":
 		contentType = "application/json"
 	case ".mmd", ".puml":

shortcuts/whiteboard/whiteboard_query_test.go

@@ -6,6 +6,8 @@ package whiteboard
 import (
 	"bytes"
 	"context"
+	"encoding/base64"
+	"encoding/json"
 	"errors"
 	"os"
 	"path/filepath"
@@ -13,6 +15,7 @@ import (
 	"testing"
 
 	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/extension/fileio"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/httpmock"
@@ -20,6 +23,7 @@ import (
 	"github.com/spf13/cobra"
 )
 
+// TestSyntaxType verifies syntax names, extensions, and validity checks.
 func TestSyntaxType(t *testing.T) {
 	t.Parallel()
 
@@ -75,6 +79,7 @@ func TestSyntaxType(t *testing.T) {
 	}
 }
 
+// TestWhiteboardQuery_Validate verifies query flag validation for supported output modes.
 func TestWhiteboardQuery_Validate(t *testing.T) {
 	ctx := context.Background()
 	chdirTemp(t)
@@ -199,6 +204,9 @@ func TestWhiteboardQuery_Validate_TypedErrors(t *testing.T) {
 			if p.Category != errs.CategoryValidation {
 				t.Errorf("Category = %q, want %q", p.Category, errs.CategoryValidation)
 			}
+			if p.Subtype != errs.SubtypeInvalidArgument {
+				t.Errorf("Problem subtype = %q, want %q", p.Subtype, errs.SubtypeInvalidArgument)
+			}
 		})
 	}
 }
@@ -232,6 +240,7 @@ func TestExportWhiteboardPreview_HTTPError(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardPreview_HTTPNotFoundIsAPIError verifies 404 preview downloads surface as typed API errors.
 func TestExportWhiteboardPreview_HTTPNotFoundIsAPIError(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 	chdirTemp(t)
@@ -255,6 +264,7 @@ func TestExportWhiteboardPreview_HTTPNotFoundIsAPIError(t *testing.T) {
 	}
 }
 
+// TestWhiteboardQuery_DryRun verifies dry-run output for the supported query modes.
 func TestWhiteboardQuery_DryRun(t *testing.T) {
 	t.Parallel()
 
@@ -307,6 +317,64 @@ func TestWhiteboardQuery_DryRun(t *testing.T) {
 	}
 }
 
+// TestWhiteboardQuery_DryRun_InvalidOutputAs verifies dry-run guidance for unsupported output modes.
+func TestWhiteboardQuery_DryRun_InvalidOutputAs(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+	rt := newTestRuntime(map[string]string{
+		"whiteboard-token": "test-token-123",
+		"output_as":        "invalid",
+	}, nil)
+
+	dryRun := WhiteboardQuery.DryRun(ctx, rt)
+	if dryRun == nil {
+		t.Fatal("WhiteboardQuery.DryRun() returned nil")
+	}
+
+	data, err := json.Marshal(dryRun)
+	if err != nil {
+		t.Fatalf("Marshal() error = %v", err)
+	}
+	if !strings.Contains(string(data), "image | svg | code | raw") {
+		t.Fatalf("dry run desc = %s, want invalid output_as guidance", string(data))
+	}
+}
+
+// TestWhiteboardQuery_Execute_InvalidOutputAs_TypedError verifies invalid output modes return typed validation errors.
+func TestWhiteboardQuery_Execute_InvalidOutputAs_TypedError(t *testing.T) {
+	rt := newTestRuntime(map[string]string{
+		"whiteboard-token": "test-token-123",
+		"output_as":        "invalid",
+	}, nil)
+
+	err := WhiteboardQuery.Execute(context.Background(), rt)
+	if err == nil {
+		t.Fatal("expected error, got nil")
+	}
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) {
+		t.Fatalf("error is not *errs.ValidationError: %T (%v)", err, err)
+	}
+	if ve.Subtype != errs.SubtypeInvalidArgument {
+		t.Errorf("Subtype = %q, want %q", ve.Subtype, errs.SubtypeInvalidArgument)
+	}
+	if ve.Param != "--output_as" {
+		t.Errorf("Param = %q, want %q", ve.Param, "--output_as")
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("errs.ProblemOf returned false")
+	}
+	if p.Category != errs.CategoryValidation {
+		t.Errorf("Category = %q, want %q", p.Category, errs.CategoryValidation)
+	}
+	if p.Subtype != errs.SubtypeInvalidArgument {
+		t.Errorf("Problem subtype = %q, want %q", p.Subtype, errs.SubtypeInvalidArgument)
+	}
+}
+
+// TestWhiteboardQuery_ShortcutRegistration verifies the whiteboard query shortcut metadata.
 func TestWhiteboardQuery_ShortcutRegistration(t *testing.T) {
 	t.Parallel()
 
@@ -325,6 +393,7 @@ func TestWhiteboardQuery_ShortcutRegistration(t *testing.T) {
 	}
 }
 
+// TestSaveOutputFile verifies output saving, overwrite handling, and extension-specific paths.
 func TestSaveOutputFile(t *testing.T) {
 	t.Parallel()
 
@@ -476,6 +545,7 @@ func TestSaveOutputFile(t *testing.T) {
 	}
 }
 
+// TestSaveOutputFile_InvalidFinalPathTypedError verifies invalid save paths return typed validation errors.
 func TestSaveOutputFile_InvalidFinalPathTypedError(t *testing.T) {
 	chdirTemp(t)
 
@@ -491,6 +561,19 @@ func TestSaveOutputFile_InvalidFinalPathTypedError(t *testing.T) {
 	if ve.Subtype != errs.SubtypeInvalidArgument || ve.Param != "--output" {
 		t.Fatalf("validation details = subtype %q param %q, want %q --output", ve.Subtype, ve.Param, errs.SubtypeInvalidArgument)
 	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("errs.ProblemOf returned false")
+	}
+	if p.Category != errs.CategoryValidation {
+		t.Errorf("Category = %q, want %q", p.Category, errs.CategoryValidation)
+	}
+	if p.Subtype != errs.SubtypeInvalidArgument {
+		t.Errorf("Problem subtype = %q, want %q", p.Subtype, errs.SubtypeInvalidArgument)
+	}
+	if !errors.Is(err, fileio.ErrPathValidation) {
+		t.Fatalf("expected path-validation cause to be preserved, err=%v", err)
+	}
 }
 
 func newExecuteFactory(t *testing.T) (*cmdutil.Factory, *bytes.Buffer, *httpmock.Registry) {
@@ -525,6 +608,7 @@ func runShortcut(t *testing.T, shortcut common.Shortcut, args []string, factory
 	return err
 }
 
+// TestWhiteboardQueryExecute_AsRaw verifies raw query execution emits the raw node payload.
 func TestWhiteboardQueryExecute_AsRaw(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -553,6 +637,7 @@ func TestWhiteboardQueryExecute_AsRaw(t *testing.T) {
 	}
 }
 
+// TestWhiteboardQueryExecute_AsCode verifies code query execution emits extracted diagram source.
 func TestWhiteboardQueryExecute_AsCode(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 	chdirTemp(t)
@@ -583,6 +668,7 @@ func TestWhiteboardQueryExecute_AsCode(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardCode_EmptyNodes verifies code export handles empty whiteboards.
 func TestExportWhiteboardCode_EmptyNodes(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -605,6 +691,7 @@ func TestExportWhiteboardCode_EmptyNodes(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardCode_NoCodeBlocks verifies code export reports whiteboards without code blocks.
 func TestExportWhiteboardCode_NoCodeBlocks(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -629,6 +716,7 @@ func TestExportWhiteboardCode_NoCodeBlocks(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardCode_InvalidSyntaxType verifies unknown syntax types are rejected.
 func TestExportWhiteboardCode_InvalidSyntaxType(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -658,6 +746,7 @@ func TestExportWhiteboardCode_InvalidSyntaxType(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardCode_MultipleCodeBlocks verifies multiple code blocks are exported together.
 func TestExportWhiteboardCode_MultipleCodeBlocks(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -697,6 +786,7 @@ func TestExportWhiteboardCode_MultipleCodeBlocks(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardCode_SingleBlock_PlantUML_DirectOutput verifies direct PlantUML output for a single code block.
 func TestExportWhiteboardCode_SingleBlock_PlantUML_DirectOutput(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -730,6 +820,7 @@ func TestExportWhiteboardCode_SingleBlock_PlantUML_DirectOutput(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardCode_SingleBlock_Mermaid_DirectOutput verifies direct Mermaid output for a single code block.
 func TestExportWhiteboardCode_SingleBlock_Mermaid_DirectOutput(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -763,6 +854,7 @@ func TestExportWhiteboardCode_SingleBlock_Mermaid_DirectOutput(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardPreview verifies preview downloads can be written to disk.
 func TestExportWhiteboardPreview(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -791,6 +883,7 @@ func TestExportWhiteboardPreview(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardRaw_EmptyNodes verifies raw export reports empty whiteboards.
 func TestExportWhiteboardRaw_EmptyNodes(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -813,6 +906,7 @@ func TestExportWhiteboardRaw_EmptyNodes(t *testing.T) {
 	}
 }
 
+// TestFetchWhiteboardNodes_APIError verifies node fetch failures preserve typed API errors.
 func TestFetchWhiteboardNodes_APIError(t *testing.T) {
 	factory, stdout, reg := newExecuteFactory(t)
 
@@ -842,6 +936,7 @@ func TestFetchWhiteboardNodes_APIError(t *testing.T) {
 	}
 }
 
+// TestFetchWhiteboardNodes_InvalidResponseTypedError verifies malformed node responses become typed invalid-response errors.
 func TestFetchWhiteboardNodes_InvalidResponseTypedError(t *testing.T) {
 	tests := []struct {
 		name  string
@@ -901,6 +996,474 @@ func TestFetchWhiteboardNodes_MissingNodesIsEmpty(t *testing.T) {
 	}
 }
 
+// TestExportWhiteboardSvg_DirectOutput verifies SVG export is printed when no output path is provided.
+func TestExportWhiteboardSvg_DirectOutput(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	svgContent := `<svg xmlns="http://www.w3.org/2000/svg"><rect width="100" height="100"/></svg>`
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg/export",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"content":   base64.StdEncoding.EncodeToString([]byte(svgContent)),
+				"mime_type": "image/svg+xml",
+			},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg", "--output_as", "svg"}
+	if err := runShortcut(t, WhiteboardQuery, args, factory, stdout); err != nil {
+		t.Fatalf("err=%v", err)
+	}
+
+	if !strings.Contains(stdout.String(), "svg_content") {
+		t.Fatalf("stdout missing svg_content key: %s", stdout.String())
+	}
+}
+
+// TestExportWhiteboardSvg_SaveToFile verifies SVG export is written to the requested file.
+func TestExportWhiteboardSvg_SaveToFile(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+	chdirTemp(t)
+
+	svgContent := `<svg xmlns="http://www.w3.org/2000/svg"><circle cx="50" cy="50" r="40"/></svg>`
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg-file/export",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"content":   base64.StdEncoding.EncodeToString([]byte(svgContent)),
+				"mime_type": "image/svg+xml",
+			},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-file", "--output_as", "svg", "--output", "output", "--overwrite"}
+	if err := runShortcut(t, WhiteboardQuery, args, factory, stdout); err != nil {
+		t.Fatalf("err=%v", err)
+	}
+
+	data, err := os.ReadFile("output.svg")
+	if err != nil {
+		t.Fatalf("ReadFile() error: %v", err)
+	}
+	if string(data) != svgContent {
+		t.Fatalf("svg content = %q, want %q", string(data), svgContent)
+	}
+}
+
+// TestExportWhiteboardSvg_PrettyOutput verifies pretty output includes inline SVG content.
+func TestExportWhiteboardSvg_PrettyOutput(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	svgContent := `<svg xmlns="http://www.w3.org/2000/svg"><path d="M0 0L10 10"/></svg>`
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg-pretty/export",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"content":   base64.StdEncoding.EncodeToString([]byte(svgContent)),
+				"mime_type": "image/svg+xml",
+			},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-pretty", "--output_as", "svg", "--format", "pretty"}
+	if err := runShortcut(t, WhiteboardQuery, args, factory, stdout); err != nil {
+		t.Fatalf("err=%v", err)
+	}
+
+	if got := stdout.String(); !strings.Contains(got, svgContent) {
+		t.Fatalf("stdout = %q, want svg content", got)
+	}
+}
+
+// TestExportWhiteboardSvg_SaveToFile_PrettyOutput verifies pretty output reports the saved SVG path and size.
+func TestExportWhiteboardSvg_SaveToFile_PrettyOutput(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+	chdirTemp(t)
+
+	svgContent := `<svg xmlns="http://www.w3.org/2000/svg"><ellipse cx="60" cy="40" rx="50" ry="30"/></svg>`
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg-file-pretty/export",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"content":   base64.StdEncoding.EncodeToString([]byte(svgContent)),
+				"mime_type": "image/svg+xml",
+			},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-file-pretty", "--output_as", "svg", "--output", "output", "--overwrite", "--format", "pretty"}
+	if err := runShortcut(t, WhiteboardQuery, args, factory, stdout); err != nil {
+		t.Fatalf("err=%v", err)
+	}
+
+	if got := stdout.String(); !strings.Contains(got, "SVG saved to output.svg") || !strings.Contains(got, "File size:") {
+		t.Fatalf("stdout = %q, want save summary", got)
+	}
+}
+
+// TestExportWhiteboardSvg_SaveToFile_ExistingFileWithoutOverwrite verifies existing SVG outputs require --overwrite.
+func TestExportWhiteboardSvg_SaveToFile_ExistingFileWithoutOverwrite(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+	chdirTemp(t)
+
+	if err := os.WriteFile("output.svg", []byte("existing content"), 0644); err != nil {
+		t.Fatalf("WriteFile() error = %v", err)
+	}
+
+	svgContent := `<svg xmlns="http://www.w3.org/2000/svg"><line x1="0" y1="0" x2="1" y2="1"/></svg>`
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg-existing/export",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"content":   base64.StdEncoding.EncodeToString([]byte(svgContent)),
+				"mime_type": "image/svg+xml",
+			},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-existing", "--output_as", "svg", "--output", "output"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for existing output without overwrite")
+	}
+
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) {
+		t.Fatalf("error is not *errs.ValidationError: %T (%v)", err, err)
+	}
+	if ve.Subtype != errs.SubtypeInvalidArgument {
+		t.Errorf("Subtype = %q, want %q", ve.Subtype, errs.SubtypeInvalidArgument)
+	}
+	if ve.Param != "--overwrite" {
+		t.Errorf("Param = %q, want %q", ve.Param, "--overwrite")
+	}
+}
+
+// TestExportWhiteboardSvg_HTTP5xx verifies plain HTTP 5xx failures are classified as retryable network errors.
+func TestExportWhiteboardSvg_HTTP5xx(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/test-token-svg-5xx/export",
+		Status:      502,
+		RawBody:     []byte("bad gateway"),
+		ContentType: "text/plain",
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-5xx", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for HTTP 502")
+	}
+	var ne *errs.NetworkError
+	if !errors.As(err, &ne) {
+		t.Fatalf("error is not *errs.NetworkError: %T (%v)", err, err)
+	}
+	if ne.Subtype != errs.SubtypeNetworkServer {
+		t.Errorf("Subtype = %q, want %q", ne.Subtype, errs.SubtypeNetworkServer)
+	}
+	if ne.Code != 502 {
+		t.Errorf("Code = %d, want 502", ne.Code)
+	}
+	if !ne.Retryable {
+		t.Error("expected Retryable = true")
+	}
+}
+
+// TestExportWhiteboardSvg_HTTP5xxJSONEnvelopeReturnsAPIError verifies API envelopes take precedence over generic 5xx handling.
+func TestExportWhiteboardSvg_HTTP5xxJSONEnvelopeReturnsAPIError(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/test-token-svg-5xx-json/export",
+		Status:      502,
+		ContentType: "application/json",
+		RawBody:     []byte(`{"code":99002,"msg":"export task failed"}`),
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-5xx-json", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for HTTP 502 JSON envelope")
+	}
+	var apiErr *errs.APIError
+	if !errors.As(err, &apiErr) {
+		t.Fatalf("error is not *errs.APIError: %T (%v)", err, err)
+	}
+	var ne *errs.NetworkError
+	if errors.As(err, &ne) {
+		t.Fatalf("expected JSON envelope to win over HTTP 5xx fallback, got *errs.NetworkError: %v", err)
+	}
+	if apiErr.Subtype != errs.SubtypeUnknown {
+		t.Errorf("Subtype = %q, want %q", apiErr.Subtype, errs.SubtypeUnknown)
+	}
+	if apiErr.Code != 99002 {
+		t.Errorf("Code = %d, want 99002", apiErr.Code)
+	}
+}
+
+// TestExportWhiteboardSvg_HTTP4xx verifies plain HTTP 4xx failures are surfaced as API errors.
+func TestExportWhiteboardSvg_HTTP4xx(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/test-token-svg-403/export",
+		Status:      403,
+		RawBody:     []byte("forbidden"),
+		ContentType: "text/plain",
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-403", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for HTTP 403")
+	}
+	var apiErr *errs.APIError
+	if !errors.As(err, &apiErr) {
+		t.Fatalf("error is not *errs.APIError: %T (%v)", err, err)
+	}
+	if apiErr.Subtype != errs.SubtypeUnknown {
+		t.Errorf("Subtype = %q, want %q", apiErr.Subtype, errs.SubtypeUnknown)
+	}
+	if apiErr.Code != 403 {
+		t.Errorf("Code = %d, want 403", apiErr.Code)
+	}
+}
+
+// TestExportWhiteboardSvg_HTTPNotFoundJSONEnvelopeIsAPIError verifies not-found envelopes preserve the typed API error classification.
+func TestExportWhiteboardSvg_HTTPNotFoundJSONEnvelopeIsAPIError(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/missing-token-svg/export",
+		Status:      404,
+		ContentType: "application/json",
+		RawBody:     []byte(`{"code":99001,"msg":"whiteboard not found"}`),
+	})
+
+	args := []string{"+query", "--whiteboard-token", "missing-token-svg", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for HTTP 404 JSON envelope")
+	}
+	var apiErr *errs.APIError
+	if !errors.As(err, &apiErr) {
+		t.Fatalf("error is not *errs.APIError: %T (%v)", err, err)
+	}
+	if apiErr.Subtype != errs.SubtypeNotFound {
+		t.Errorf("Subtype = %q, want %q", apiErr.Subtype, errs.SubtypeNotFound)
+	}
+	if apiErr.Code != 99001 {
+		t.Errorf("Code = %d, want 99001", apiErr.Code)
+	}
+}
+
+// TestExportWhiteboardSvg_HTTPNotFoundPlainText verifies plain-text 404 responses surface as not-found API errors.
+func TestExportWhiteboardSvg_HTTPNotFoundPlainText(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/missing-token-svg-plain/export",
+		Status:      404,
+		ContentType: "text/plain",
+		RawBody:     []byte("whiteboard not found"),
+	})
+
+	args := []string{"+query", "--whiteboard-token", "missing-token-svg-plain", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for HTTP 404 plain text response")
+	}
+
+	var apiErr *errs.APIError
+	if !errors.As(err, &apiErr) {
+		t.Fatalf("error is not *errs.APIError: %T (%v)", err, err)
+	}
+	if apiErr.Subtype != errs.SubtypeNotFound {
+		t.Errorf("Subtype = %q, want %q", apiErr.Subtype, errs.SubtypeNotFound)
+	}
+	if apiErr.Code != 404 {
+		t.Errorf("Code = %d, want 404", apiErr.Code)
+	}
+}
+
+// TestExportWhiteboardSvg_InvalidJSON verifies malformed success responses are rejected as invalid responses.
+func TestExportWhiteboardSvg_InvalidJSON(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/test-token-svg-badjson/export",
+		Status:      200,
+		RawBody:     []byte("not json at all"),
+		ContentType: "application/json",
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-badjson", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for invalid JSON")
+	}
+	assertInvalidResponse(t, err)
+}
+
+// TestExportWhiteboardSvg_InvalidBody200PlainText verifies plain-text 200 responses are rejected as invalid export responses.
+func TestExportWhiteboardSvg_InvalidBody200PlainText(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method:      "POST",
+		URL:         "/open-apis/board/v1/whiteboards/test-token-svg-plain-200/export",
+		Status:      200,
+		RawBody:     []byte("not json at all"),
+		ContentType: "text/plain",
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-plain-200", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for plain text success response")
+	}
+	assertInvalidResponse(t, err)
+}
+
+// TestExportWhiteboardSvg_NonZeroCode verifies non-zero API codes are returned as typed API errors.
+func TestExportWhiteboardSvg_NonZeroCode(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg-apierr/export",
+		Body: map[string]interface{}{
+			"code": 99001,
+			"msg":  "whiteboard not found",
+			"data": map[string]interface{}{},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-apierr", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for non-zero code")
+	}
+	var apiErr *errs.APIError
+	if !errors.As(err, &apiErr) {
+		t.Fatalf("error is not *errs.APIError: %T (%v)", err, err)
+	}
+	if apiErr.Code != 99001 {
+		t.Errorf("Code = %d, want 99001", apiErr.Code)
+	}
+}
+
+// TestExportWhiteboardSvg_InvalidBase64 verifies invalid SVG payload encoding is rejected.
+func TestExportWhiteboardSvg_InvalidBase64(t *testing.T) {
+	factory, stdout, reg := newExecuteFactory(t)
+
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg-badbase64/export",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"content":   "!!!not-valid-base64!!!",
+				"mime_type": "image/svg+xml",
+			},
+		},
+	})
+
+	args := []string{"+query", "--whiteboard-token", "test-token-svg-badbase64", "--output_as", "svg"}
+	err := runShortcut(t, WhiteboardQuery, args, factory, stdout)
+	if err == nil {
+		t.Fatal("expected error for invalid base64")
+	}
+	assertInvalidResponse(t, err)
+}
+
+// TestWhiteboardQuery_Validate_SvgValid verifies svg is accepted as a valid query output format.
+func TestWhiteboardQuery_Validate_SvgValid(t *testing.T) {
+	ctx := context.Background()
+	chdirTemp(t)
+
+	rt := newTestRuntime(map[string]string{
+		"whiteboard-token": "test-token-123",
+		"output_as":        "svg",
+	}, nil)
+	if err := WhiteboardQuery.Validate(ctx, rt); err != nil {
+		t.Fatalf("expected svg to be valid, got err=%v", err)
+	}
+}
+
+// TestWhiteboardQuery_DryRun_Svg verifies the svg dry-run request uses the export endpoint and body.
+func TestWhiteboardQuery_DryRun_Svg(t *testing.T) {
+	t.Parallel()
+	ctx := context.Background()
+
+	rt := newTestRuntime(map[string]string{
+		"whiteboard-token": "test-token-123",
+		"output_as":        "svg",
+	}, nil)
+	dryRun := WhiteboardQuery.DryRun(ctx, rt)
+	if dryRun == nil {
+		t.Fatal("DryRun() returned nil for svg")
+	}
+
+	data, err := json.Marshal(dryRun)
+	if err != nil {
+		t.Fatalf("Marshal() error = %v", err)
+	}
+
+	var got struct {
+		API []struct {
+			Method string                 `json:"method"`
+			URL    string                 `json:"url"`
+			Params map[string]interface{} `json:"params"`
+			Body   map[string]interface{} `json:"body"`
+		} `json:"api"`
+	}
+	if err := json.Unmarshal(data, &got); err != nil {
+		t.Fatalf("Unmarshal() error = %v", err)
+	}
+
+	if len(got.API) != 1 {
+		t.Fatalf("len(api) = %d, want 1", len(got.API))
+	}
+	if got.API[0].Method != "POST" {
+		t.Fatalf("method = %q, want POST", got.API[0].Method)
+	}
+	if got.API[0].URL != "/open-apis/board/v1/whiteboards/test...-123/export" {
+		t.Fatalf("url = %q", got.API[0].URL)
+	}
+	if got.API[0].Body["export_type"] != "svg" {
+		t.Fatalf("body = %#v, want export_type=svg", got.API[0].Body)
+	}
+	if _, ok := got.API[0].Params["export_type"]; ok {
+		t.Fatalf("params should not include export_type, got %#v", got.API[0].Params)
+	}
+}
+
+// assertInvalidResponse verifies an error is classified as a typed invalid-response failure.
 func assertInvalidResponse(t *testing.T, err error) {
 	t.Helper()
 	if err == nil {

shortcuts/whiteboard/whiteboard_update.go

@@ -17,15 +17,21 @@ import (
 )
 
 const (
-	FormatRaw      = "raw"
+	// FormatRaw sends raw whiteboard node JSON to the create-nodes API.
+	FormatRaw = "raw"
+	// FormatPlantUML sends PlantUML source through the diagram import API.
 	FormatPlantUML = "plantuml"
-	FormatMermaid  = "mermaid"
+	// FormatMermaid sends Mermaid source through the diagram import API.
+	FormatMermaid = "mermaid"
+	// FormatSVG sends SVG source through the diagram import API.
+	FormatSVG = "svg"
 )
 
 var formatCodeMap = map[string]int{
 	FormatRaw:      0,
 	FormatPlantUML: 1,
 	FormatMermaid:  2,
+	FormatSVG:      3,
 }
 
 var wbUpdateScopes = []string{"board:whiteboard:node:create"}
@@ -35,9 +41,14 @@ var wbUpdateFlags = []common.Flag{
 	{Name: "whiteboard-token", Desc: "whiteboard token of the whiteboard to update. You will need edit permission to update the whiteboard.", Required: true},
 	{Name: "overwrite", Desc: "overwrite the whiteboard content, delete all existing content before update. Default is false.", Required: false, Type: "bool"},
 	{Name: "source", Desc: "Input whiteboard data.", Required: true, Input: []string{common.Stdin, common.File}},
-	{Name: "input_format", Desc: "format of input data: raw | plantuml | mermaid. Default is raw.", Required: false},
+	{Name: "input_format", Desc: "format of input data: raw | plantuml | mermaid | svg. Default is raw.", Required: false},
 }
 
+// wbUpdateValidate validates the whiteboard update command arguments.
+//
+// It checks the whiteboard token and idempotent token for dangerous control
+// characters, enforces a minimum length for a non-empty idempotent token, and
+// ensures the input format is one of raw, plantuml, mermaid, or svg.
 func wbUpdateValidate(ctx context.Context, runtime *common.RuntimeContext) error {
 	// 检查 token 是否包含控制字符（空字符串下自动跳过了）
 	if err := common.RejectDangerousCharsTyped("--whiteboard-token", runtime.Str("whiteboard-token")); err != nil {
@@ -53,8 +64,8 @@ func wbUpdateValidate(ctx context.Context, runtime *common.RuntimeContext) error
 
 	// 检查 --input_format 标志
 	format := getFormat(runtime)
-	if format != FormatRaw && format != FormatPlantUML && format != FormatMermaid {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "--input_format must be one of: raw | plantuml | mermaid").WithParam("--input_format")
+	if format != FormatRaw && format != FormatPlantUML && format != FormatMermaid && format != FormatSVG {
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "--input_format must be one of: raw | plantuml | mermaid | svg").WithParam("--input_format")
 	}
 	return nil
 }
@@ -68,6 +79,8 @@ func getFormat(runtime *common.RuntimeContext) string {
 	return format
 }
 
+// wbUpdateDryRun describes the HTTP request used to update a whiteboard.
+// It returns a failure description when source is missing or cannot be parsed.
 func wbUpdateDryRun(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
 	// 读取输入内容
 	input := runtime.Str("source")
@@ -91,7 +104,7 @@ func wbUpdateDryRun(ctx context.Context, runtime *common.RuntimeContext) *common
 			Overwrite: overwrite,
 		}
 		desc.POST(fmt.Sprintf("/open-apis/board/v1/whiteboards/%s/nodes", common.MaskToken(url.PathEscape(token)))).Body(reqBody).Desc("create all nodes of the whiteboard.")
-	case FormatPlantUML, FormatMermaid:
+	case FormatPlantUML, FormatMermaid, FormatSVG:
 		syntaxType := formatCodeMap[format]
 		reqBody := plantumlCreateReq{
 			PlantUmlCode: input,
@@ -106,6 +119,10 @@ func wbUpdateDryRun(ctx context.Context, runtime *common.RuntimeContext) *common
 	return desc
 }
 
+// wbUpdateExecute updates a whiteboard from the supplied source input.
+// It requires --source and dispatches to the raw node update path for raw input
+// or the diagram import path for PlantUML, Mermaid, and SVG input.
+// It returns an error if the source is missing or the input format is unsupported.
 func wbUpdateExecute(ctx context.Context, runtime *common.RuntimeContext) error {
 	token := runtime.Str("whiteboard-token")
 	overwrite := runtime.Bool("overwrite")
@@ -120,15 +137,17 @@ func wbUpdateExecute(ctx context.Context, runtime *common.RuntimeContext) error
 	switch format {
 	case FormatRaw:
 		return updateWhiteboardByRawNodes(ctx, runtime, token, []byte(input), overwrite, idempotentToken)
-	case FormatPlantUML, FormatMermaid:
+	case FormatPlantUML, FormatMermaid, FormatSVG:
 		return updateWhiteboardByCode(ctx, runtime, token, []byte(input), format, overwrite, idempotentToken)
 	default:
 		return errs.NewValidationError(errs.SubtypeInvalidArgument, "unsupported format: %s", format).WithParam("--input_format")
 	}
 }
 
+// WhiteboardUpdateDescription describes the whiteboard update shortcut.
 const WhiteboardUpdateDescription = "Update an existing whiteboard in lark document with mermaid, plantuml or whiteboard dsl. refer to lark-whiteboard skill for more details."
 
+// WhiteboardUpdate registers the `whiteboard +update` shortcut.
 var WhiteboardUpdate = common.Shortcut{
 	Service:     "whiteboard",
 	Command:     "+update",

shortcuts/whiteboard/whiteboard_update_test.go

@@ -6,6 +6,7 @@ package whiteboard
 import (
 	"bytes"
 	"context"
+	"encoding/json"
 	"errors"
 	"strings"
 	"testing"
@@ -18,6 +19,7 @@ import (
 	"github.com/spf13/cobra"
 )
 
+// TestWhiteboardUpdate_Validate verifies update flag validation for supported input formats.
 func TestWhiteboardUpdate_Validate(t *testing.T) {
 	ctx := context.Background()
 
@@ -53,6 +55,15 @@ func TestWhiteboardUpdate_Validate(t *testing.T) {
 			},
 			wantErr: false,
 		},
+		{
+			name: "valid: svg format",
+			flags: map[string]string{
+				"whiteboard-token": "test-token-123",
+				"input_format":     "svg",
+				"source":           "<svg/>",
+			},
+			wantErr: false,
+		},
 		{
 			name: "valid: with idempotent-token",
 			flags: map[string]string{
@@ -117,25 +128,26 @@ func TestWhiteboardUpdate_Validate_TypedErrors(t *testing.T) {
 			"idempotent-token": "short",
 			"source":           "{}",
 		}, nil)
-		assertValidationParam(t, wbUpdateValidate(ctx, rt), "--idempotent-token")
+		assertValidationParam(t, wbUpdateValidate(ctx, rt), "--idempotent-token", false)
 	})
 
 	t.Run("bad input_format", func(t *testing.T) {
 		rt := newTestRuntime(map[string]string{
 			"whiteboard-token": "t",
-			"input_format":     "svg",
+			"input_format":     "png",
 			"source":           "{}",
 		}, nil)
-		assertValidationParam(t, wbUpdateValidate(ctx, rt), "--input_format")
+		assertValidationParam(t, wbUpdateValidate(ctx, rt), "--input_format", false)
 	})
 
 	t.Run("malformed source json", func(t *testing.T) {
 		_, err, _ := parseWBcliNodes([]byte("not-json"))
-		assertValidationParam(t, err, "--source")
+		assertValidationParam(t, err, "--source", true)
 	})
 }
 
-func assertValidationParam(t *testing.T, err error, wantParam string) {
+// assertValidationParam verifies a validation error carries the expected flag param.
+func assertValidationParam(t *testing.T, err error, wantParam string, wantJSONCause bool) {
 	t.Helper()
 	if err == nil {
 		t.Fatalf("expected error, got nil")
@@ -150,8 +162,25 @@ func assertValidationParam(t *testing.T, err error, wantParam string) {
 	if ve.Param != wantParam {
 		t.Errorf("Param = %q, want %q", ve.Param, wantParam)
 	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("errs.ProblemOf returned false")
+	}
+	if p.Category != errs.CategoryValidation {
+		t.Errorf("Category = %q, want %q", p.Category, errs.CategoryValidation)
+	}
+	if p.Subtype != errs.SubtypeInvalidArgument {
+		t.Errorf("Problem subtype = %q, want %q", p.Subtype, errs.SubtypeInvalidArgument)
+	}
+	if wantJSONCause {
+		var syntaxErr *json.SyntaxError
+		if !errors.As(err, &syntaxErr) {
+			t.Fatalf("expected json syntax cause to be preserved, err=%v", err)
+		}
+	}
 }
 
+// TestGetFormat verifies input format defaults and explicit format selection.
 func TestGetFormat(t *testing.T) {
 	t.Parallel()
 
@@ -180,6 +209,11 @@ func TestGetFormat(t *testing.T) {
 			flagVal:  FormatMermaid,
 			expected: FormatMermaid,
 		},
+		{
+			name:     "svg returns svg",
+			flagVal:  FormatSVG,
+			expected: FormatSVG,
+		},
 	}
 
 	for _, tt := range tests {
@@ -193,6 +227,7 @@ func TestGetFormat(t *testing.T) {
 	}
 }
 
+// TestWhiteboardUpdate_ShortcutRegistration verifies the shortcut metadata for update commands.
 func TestWhiteboardUpdate_ShortcutRegistration(t *testing.T) {
 	t.Parallel()
 
@@ -213,6 +248,7 @@ func TestWhiteboardUpdate_ShortcutRegistration(t *testing.T) {
 	}
 }
 
+// TestShortcutsIncludesExpectedCommands verifies the whiteboard shortcut registry includes query and update.
 func TestShortcutsIncludesExpectedCommands(t *testing.T) {
 	t.Parallel()
 
@@ -237,6 +273,7 @@ func TestShortcutsIncludesExpectedCommands(t *testing.T) {
 	}
 }
 
+// TestParseWBcliNodes verifies whiteboard CLI output parsing for raw and wrapped node payloads.
 func TestParseWBcliNodes(t *testing.T) {
 	t.Parallel()
 
@@ -285,6 +322,7 @@ func TestParseWBcliNodes(t *testing.T) {
 	}
 }
 
+// TestWBUpdateDryRun verifies dry-run requests for the supported whiteboard update formats.
 func TestWBUpdateDryRun(t *testing.T) {
 	ctx := context.Background()
 
@@ -317,6 +355,14 @@ func TestWBUpdateDryRun(t *testing.T) {
 				"source":           "graph TD\nA-->B",
 			},
 		},
+		{
+			name: "dry run svg format",
+			flags: map[string]string{
+				"whiteboard-token": "test-token-123",
+				"input_format":     "svg",
+				"source":           "<svg/>",
+			},
+		},
 	}
 
 	for _, tt := range tests {
@@ -362,6 +408,7 @@ func runUpdateShortcut(t *testing.T, shortcut common.Shortcut, args []string, fa
 	return err
 }
 
+// TestWhiteboardUpdateExecute_RawFormat verifies raw node updates call the raw nodes endpoint.
 func TestWhiteboardUpdateExecute_RawFormat(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -385,6 +432,7 @@ func TestWhiteboardUpdateExecute_RawFormat(t *testing.T) {
 	}
 }
 
+// TestWhiteboardUpdateExecute_PlantUMLFormat verifies PlantUML updates use the diagram import endpoint.
 func TestWhiteboardUpdateExecute_PlantUMLFormat(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -410,6 +458,7 @@ Bob -> Alice : hello
 	}
 }
 
+// TestWhiteboardUpdateExecute_PlantUMLInvalidResponse verifies missing node IDs are treated as invalid responses.
 func TestWhiteboardUpdateExecute_PlantUMLInvalidResponse(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -431,6 +480,7 @@ Bob -> Alice : hello
 	assertInvalidResponse(t, err)
 }
 
+// TestWhiteboardUpdateExecute_MermaidFormat verifies Mermaid updates use the diagram import endpoint.
 func TestWhiteboardUpdateExecute_MermaidFormat(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -455,6 +505,44 @@ A-->B`
 	}
 }
 
+// TestWhiteboardUpdateExecute_SVGFormat verifies svg update requests use syntax_type=3 and send the source payload.
+func TestWhiteboardUpdateExecute_SVGFormat(t *testing.T) {
+	factory, stdout, reg := newUpdateExecuteFactory(t)
+
+	// SVG shares the /nodes/plantuml endpoint with plantuml/mermaid via syntax_type=3.
+	stub := &httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/board/v1/whiteboards/test-token-svg/nodes/plantuml",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "success",
+			"data": map[string]interface{}{
+				"node_id": "node1",
+			},
+		},
+	}
+	reg.Register(stub)
+
+	source := `<svg xmlns="http://www.w3.org/2000/svg"/>`
+	args := []string{"+update", "--whiteboard-token", "test-token-svg", "--input_format", "svg", "--source", source}
+	if err := runUpdateShortcut(t, WhiteboardUpdate, args, factory, stdout); err != nil {
+		t.Fatalf("err=%v", err)
+	}
+
+	var body map[string]interface{}
+	if err := json.Unmarshal(stub.CapturedBody, &body); err != nil {
+		t.Fatalf("unmarshal captured body: %v\nraw=%s", err, string(stub.CapturedBody))
+	}
+
+	if got := body["syntax_type"]; got != float64(3) {
+		t.Fatalf("syntax_type = %#v, want 3; body=%s", got, string(stub.CapturedBody))
+	}
+	if got := body["plant_uml_code"]; got != source {
+		t.Fatalf("plant_uml_code = %#v, want %q; body=%s", got, source, string(stub.CapturedBody))
+	}
+}
+
+// TestWhiteboardUpdateExecute_RawInvalidResponse verifies malformed raw update responses are rejected.
 func TestWhiteboardUpdateExecute_RawInvalidResponse(t *testing.T) {
 	tests := []struct {
 		name  string
@@ -494,6 +582,7 @@ func TestWhiteboardUpdateExecute_RawInvalidResponse(t *testing.T) {
 	}
 }
 
+// TestWhiteboardUpdateExecute_RawWithIdempotent verifies raw updates pass through the idempotency token.
 func TestWhiteboardUpdateExecute_RawWithIdempotent(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -518,6 +607,7 @@ func TestWhiteboardUpdateExecute_RawWithIdempotent(t *testing.T) {
 	}
 }
 
+// TestWhiteboardUpdateExecute_RawFormatWithRawNodes verifies raw-node payloads are forwarded without DSL wrapping.
 func TestWhiteboardUpdateExecute_RawFormatWithRawNodes(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -541,6 +631,7 @@ func TestWhiteboardUpdateExecute_RawFormatWithRawNodes(t *testing.T) {
 	}
 }
 
+// TestWhiteboardUpdateExecute_RawAPIError verifies raw update API failures preserve typed error metadata and hints.
 func TestWhiteboardUpdateExecute_RawAPIError(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -577,6 +668,7 @@ func TestWhiteboardUpdateExecute_RawAPIError(t *testing.T) {
 	}
 }
 
+// TestWhiteboardUpdateExecute_PlantUMLAPIError verifies diagram update API failures preserve typed error metadata.
 func TestWhiteboardUpdateExecute_PlantUMLAPIError(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -607,6 +699,7 @@ invalid
 	}
 }
 
+// TestWhiteboardUpdateExecute_WithOverwrite verifies diagram updates send overwrite=true when requested.
 func TestWhiteboardUpdateExecute_WithOverwrite(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 
@@ -631,6 +724,7 @@ A-->B`
 	}
 }
 
+// TestWhiteboardUpdateExecute_RawWithOverwrite verifies raw updates send overwrite=true when requested.
 func TestWhiteboardUpdateExecute_RawWithOverwrite(t *testing.T) {
 	factory, stdout, reg := newUpdateExecuteFactory(t)
 

skills/lark-apps/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: lark-apps
 version: 1.0.0
-description: "妙搭（Spark/Miaoda）应用开发与托管：应用创建、HTML静态站点发布、本地全栈开发、云端生成迭代。当用户要开发/新建一个系统·工具·平台·应用，或要本地开发 / 云端开发 / 修改 / 部署 / 发布 / 上线 / 拿可分享链接，或用 HTML 做页面·网站给人看，或提到妙搭/Spark/Miaoda、应用数据库、可见范围时使用。不负责普通云盘文件上传（lark-drive）、飞书文档编辑（lark-doc）、原生幻灯片创建（lark-slides）。"
+description: "妙搭（Spark/Miaoda）应用开发与托管：应用创建、HTML静态站点发布、本地全栈开发、云端生成迭代。当用户要开发/新建一个系统·工具·平台·应用，或要本地开发 / 云端开发 / 修改 / 部署 / 发布 / 上线 / 拿可分享链接，或用 HTML 做页面·网站·部署到妙搭，或提到妙搭/Spark/Miaoda（应用运行时域名形如 *.aiforce.cloud）、应用数据库、可见范围时使用。不负责普通云盘文件上传（lark-drive）、飞书文档编辑（lark-doc）、原生幻灯片创建（lark-slides）。"
 metadata:
   requires:
     bins: ["lark-cli"]
@@ -48,8 +48,14 @@ metadata:
 
 - **发布意图判定**：用户要"可访问 / 线上 / 分享 / 新链接 / 上线" = 发布意图，先走发布链路、确认完成再给链接。
 - 完成 ≠ 发布：云端会话完成 / `+list is_published=true` 都不代表最新内容已部署。
-- 开发态链接 `https://miaoda.feishu.cn/app/{app_id}` 仅进编辑态，不能顶替发布当分享链接。
+- 开发态链接 `https://miaoda.feishu.cn/app/{app_id}`：进应用编辑/开发态、管理与继续开发应用的入口。发布成功后，连同发布态链接一并提供给用户（说明"管理 / 继续开发去这里"）；但它仅进编辑态，**不能**顶替发布态链接当分享链接。
 - 发布态链接来源：html → `+html-publish` 的 `data.url`；全栈 → `+release-get` 轮询 `finished` 给 `online_url` / `failed` 给 `error_logs`。
+- **可见范围**：发布态链接（html 的 `data.url`、全栈的 `online_url`）默认仅**创建者可见**，发给他人对方会无权限打不开。当可分享链接交付给用户前，先告知当前仅本人可见，再询问是否用 `+access-scope-set`（`tenant`/`public`/`specific`）放开（可先 `+access-scope-get` 查当前范围）。
+
+## 能力边界
+
+- lark-cli **不支持**配置应用的权限（应用内 RBAC、成员角色、协作者权限）/ 自动化 / 插件。`+access-scope-*` 只管运行时可见范围（谁能打开应用），不是角色权限。
+- 用户要配置权限 / 自动化 / 插件时，引导其使用开发态连接前往云端开发（妙搭 web）处理。
 
 ## app_id 获取
 
@@ -69,4 +75,4 @@ metadata:
 ## 高影响动作：确认与预授权
 
 - **预授权判定**：判断用户是否表达了"放手做完、不用中途逐步问我"的意图——明确免确认（如"别问 / 直接做 / 自己定"），或要求一气呵成做到完成（如"做完部署上线给我"）。是 → 整个流程按合理默认往下走、不再逐步确认（含 clone 到派生目录、发布等）；否 → 缺失参数（如目录）该问就问、高影响动作先确认。
-- **不豁免底线**：会删/丢数据或不可逆的 DB 操作（判据见 [`lark-apps-db-execute.md`](references/lark-apps-db-execute.md)）即便已预授权，也先 `--dry-run` 确认。
+- **禁止预授权判定底线**（即便已预授权也不豁免）：① 会删/丢数据或不可逆的 DB 操作（判据见 [`lark-apps-db-execute.md`](references/lark-apps-db-execute.md)）先 `--dry-run` 确认；② `+html-publish` 体积超限时（判据见 [`lark-apps-html-publish.md`](references/lark-apps-html-publish.md)），立即停止并转述超限项。

skills/lark-apps/references/lark-apps-html-publish.md

@@ -11,7 +11,7 @@
 - 必填：`--app-id`、`--path`。
 - `--path` 可以是单个文件或目录；入口必须是 `index.html`。
 - 可选：`--allow-sensitive`，跳过凭据文件扫描。
-- 客户端会打包 tar.gz 并上传发布；压缩包上限当前为 20MB，未压缩候选文件总量也有保护上限。
+- 客户端打包 tar.gz 上传发布。三条硬性大小限制，任一超限即被客户端拒绝、无法发布：单个 `.html` 文件 ≤ 10MB、打包后 tar.gz ≤ 20MB、未压缩候选文件总量 ≤ 200MB。
 
 ## 示例
 
@@ -33,12 +33,19 @@ lark-cli apps +html-publish --app-id app_xxx --path ./index.html --dry-run
 - 发布态访问链接以本命令成功返回的 `data.url` 为准。
 - 重新发布前，`+list` 的 `is_published=true` 只能说明历史上发布过，不代表当前本地产物已经部署。
 
+## 发布前置门（第一步，先于任何其他动作）
+
+收到发布意图后，第一个动作是量三个尺寸，不是读文件内容、不是打包：
+1. 单个 `.html` ≤ 10MB / tar.gz ≤ 20MB / 未压缩总量 ≤ 200MB。
+2. 任一超限 → 立即 STOP，把超限数字转述给用户，交还决定权。
+3. 三项都通过 → 才进入下面的命令骨架。
+
 ## 预览与发布边界
 
 - 用户只说“用 HTML 写个 PPT/页面给我看看”时，先生成本地文件或目录，返回路径并问是否发布到妙搭分享；不要默认创建应用或部署。
 - 用户明确说“部署出去/发链接/可分享”时，才创建 `html` 应用并用 `+html-publish`。
 - 用户要发布但没有 app_id 时，先 `+create --app-type html` 创建应用；应用名可从页面/站点主题生成，不要让用户手动提供 app_id。
-- 若产物首页不是 `index.html`，发布前改名或复制为 `index.html`；目录发布时只传干净产物目录，例如 `./dist`。`.git` 目录会被自动排除，不会进入压缩包；`node_modules`、源码缓存等仍建议手动精简以控制包体。
+- 若产物首页不是 `index.html`，发布前改名或复制为 `index.html`；目录发布时只传干净产物目录，例如 `./dist`。`.git` 目录会被自动排除，不会进入压缩包。
 - 重新部署同一个 HTML 应用时复用原 `app_id`，只重新执行 `+html-publish --app-id <id> --path <dir-or-index.html>`。
 
 ## 安全规则
@@ -48,4 +55,3 @@ lark-cli apps +html-publish --app-id app_xxx --path ./index.html --dry-run
 ## 常见失败
 
 - 缺少 `index.html`：目录根放置 `index.html`，或单文件路径直接指向名为 `index.html` 的文件。
-- 包体过大：让用户精简 `--path`，不要把源码、依赖目录、构建缓存一起发布。

skills/lark-apps/references/lark-apps-init.md

@@ -31,6 +31,7 @@ lark-cli apps +init --app-id app_xxx --dir ./my-app --dry-run
 
 ## Agent 规则
 
-- 目标目录必须不存在、为空目录，或已含 `.spark/meta.json` 的已初始化仓库。
+- 目标目录必须不存在、为空目录，或已含 `.spark/meta.json` 且其 app_id 与 `--app-id` 一致的已初始化仓库。
 - 目标目录已含 `.spark/meta.json` 时，`+init` 会跳过 clone/scaffold，但仍执行一次 env-pull 刷新本地环境变量；告知用户“仓库已初始化，本地环境变量已刷新，可直接开发”，不要误报失败或重复 clone。
 - `+init` 输出没有必要原样复述；告诉用户 clone path、分支和下一步即可。
+- 新建应用做本地初始化时，若选定的目标目录已存在，不要复用，改用一个不冲突的目录名（已预授权”放手做”时自动追加后缀如 `-2`；否则向用户确认目录名）。

skills/lark-doc/SKILL.md

@@ -26,7 +26,7 @@ lark-cli docs +update --api-version v2 --doc "文档URL或token" --command appen
 **CRITICAL — 执行对应操作前，MUST 先用 Read 工具读取以下文件，缺一不可：**
 1. [`../lark-shared/SKILL.md`](../lark-shared/SKILL.md) — 认证、权限处理、全局参数（所有操作通用）
 2. **读取文档（`docs +fetch --api-version v2`）** → 必读 [`lark-doc-fetch.md`](references/lark-doc-fetch.md)（`--scope` / `--detail` 选择、局部读取策略、`<fragment>` / `<excerpt>` 输出结构）
-3. **创建或编辑文档内容** → 必读 [`lark-doc-xml.md`](references/lark-doc-xml.md)（XML 语法规则，仅当用户明确要求 Markdown 时改读 [`lark-doc-md.md`](references/lark-doc-md.md)）；从零创建时加读 [`lark-doc-create-workflow.md`](references/style/lark-doc-create-workflow.md)；编辑已有文档时加读 [`lark-doc-update-workflow.md`](references/style/lark-doc-update-workflow.md)
+3. **创建或编辑文档内容** → 必读 [`lark-doc-xml.md`](references/lark-doc-xml.md)（XML 语法规则，仅当用户明确要求 Markdown 时改读 [`lark-doc-md.md`](references/lark-doc-md.md)）；从零创建时加读 [`lark-doc-create-workflow.md`](references/style/lark-doc-create-workflow.md)；编辑已有文档时加读 [`lark-doc-update.md`](references/lark-doc-update.md) 和 [`lark-doc-update-workflow.md`](references/style/lark-doc-update-workflow.md)
 4. **需要使用 callout、grid、table、whiteboard 等富 block 时** → 参考 [`lark-doc-style.md`](references/style/lark-doc-style.md) 的元素能力说明。该文件不是固定模板或强制排版规范；除非用户明确要求美化、重排版或特定风格，不要为了“达标”主动套用固定结构。
 
 **未读完以上文件就执行相应操作会导致参数选择错误或格式错误。**
@@ -36,11 +36,9 @@ lark-cli docs +update --api-version v2 --doc "文档URL或token" --command appen
 > - **精准编辑场景**（`docs +update` 的 `str_replace` / `block_insert_after` / `block_replace` / `block_delete` / `block_move_after` 等局部精修指令）：优先使用 XML（`--doc-format xml`，即默认值）。XML 能稳定表达 block 结构和样式，局部精修更可控；不要因为 Markdown 更简单就自行切换。
 
 ## 快速决策
-- 用户需要“某个 block 的直达链接 / 锚点链接”时：返回 `文档基础 URL#block_id`。如果当前只有文档 URL 没有 block_id，先用 `docs +fetch --detail with-ids` 拿到目标 block 的 id
-- 例：
-  - 已知文档 URL = `https://xxx.feishu.cn/docx/doxcn123`
-  - 已知 block_id = `blkcn456`
-  - 应返回 `https://xxx.feishu.cn/docx/doxcn123#blkcn456`
+- 先判定任务路径：找文档 / 导入导出走 [`lark-drive`](../lark-drive/SKILL.md)；只读 / 摘要用 `docs +fetch` 默认 `simple`；明确旧文本 → 新文本直接 `str_replace`；只有 block 链接、评论锚点、插入 / 替换 / 删除 / 移动才局部 fetch `with-ids`；保真改写已有内容才读 `full`
+- block 直达链接格式：`文档基础 URL#block_id`；没有 block_id 时局部 fetch `with-ids`
+- 连续执行多个文档写操作时，必须按 [`lark-doc-update.md`](references/lark-doc-update.md) 的「Block ID 生命周期」判断旧 block ID 是否还能复用；`overwrite` / `block_replace` / `block_delete` 后不要复用受影响的旧 ID，插入 / 复制后要重新 fetch 才能拿到新 block ID
 - 用户需要在文档内**创建、复制或移动**资源块（画板、电子表格、多维表格等）时，必须先读取 [`lark-doc-xml.md`](references/lark-doc-xml.md) 的「三、资源块」章节
 - 写文档时，由内容和用户意图决定表达形式；流程、架构、路线图、关键指标等信息可以使用画板，但不要默认把重要信息都画板化
 - 新增画板必须隔离到 SubAgent：简单图由 SubAgent 直接插入 `<whiteboard type="svg">完整 SVG</whiteboard>`，不读 `lark-whiteboard`；复杂图才由主 Agent 先建 `<whiteboard type="blank"></whiteboard>`，再启动 SubAgent 读取 `lark-whiteboard` 写入

skills/lark-doc/references/lark-doc-create.md

@@ -4,7 +4,7 @@
 > 1. [`lark-doc-xml.md`](lark-doc-xml.md) — XML 语法规则（使用 Markdown 格式时改读 [`lark-doc-md.md`](lark-doc-md.md)）
 > 2. [`lark-doc-create-workflow.md`](style/lark-doc-create-workflow.md) — 从零创作工作流（Code-Act Loop、并行执行策略）
 >
-> **需要使用 callout、grid、table、whiteboard 等富 block，或用户明确要求美化/重排版时，再参考 [`lark-doc-style.md`](style/lark-doc-style.md)。该文件是表达组件参考，不是固定模板。**
+> **需要富 block 或用户明确要求美化/重排版时，再参考 [`lark-doc-style.md`](style/lark-doc-style.md)。**
 >
 > **未读完以上文件就生成内容会导致格式错误。**
 
@@ -74,7 +74,7 @@ lark-cli docs +create --api-version v2 --doc-format markdown --content $'# 项
 ## 最佳实践
 
 - 文档标题从内容中自动提取：XML 使用 `<title>`；Markdown 使用文档开头唯一的一级标题（`# 标题`），正文从 `##` 开始。不要在内容开头重复写标题，也不要在 Markdown 正文中使用多个一级标题。
-- **创建较长的文档时只建骨架**：`--content` 仅传标题 + 各级 heading + 简短占位摘要；正文留给后续 `block_insert_after --block-id <章节标题 block_id>` 分段追加。一次性塞超长 `--content` 既容易触发参数限制，调试也更难。
+- **较长文档**：参考 [`lark-doc-create-workflow.md`](style/lark-doc-create-workflow.md) 先建骨架再分段写入；短文档可一次写完整内容。
 - **表达形式**：由用户目标和内容决定。需要结构化表达时可参考 [`lark-doc-style.md`](style/lark-doc-style.md)，但不要默认套用固定开头、固定富 block 比例或固定图表
 
 ## 参考

skills/lark-doc/references/lark-doc-update.md

@@ -5,7 +5,7 @@
 > 1. [`lark-doc-xml.md`](lark-doc-xml.md) — XML 语法规则（使用 Markdown 格式时改读 [`lark-doc-md.md`](lark-doc-md.md)）
 > 2. [`lark-doc-update-workflow.md`](style/lark-doc-update-workflow.md) — 改写增强工作流（Code-Act Loop、并行执行策略）
 >
-> **需要使用 callout、grid、table、whiteboard 等富 block，或用户明确要求美化/重排版时，再参考 [`lark-doc-style.md`](style/lark-doc-style.md)。该文件是表达组件参考，不是固定模板。**
+> **需要富 block 或用户明确要求美化/重排版时，再参考 [`lark-doc-style.md`](style/lark-doc-style.md)。**
 >
 > **未读完以上文件就生成内容会导致格式错误。**
 
@@ -44,6 +44,15 @@
 | `append` | ⚠️ 在文档**末尾**追加内容（等价于 `block_insert_after --block-id -1`）。**不适用于逐章填充**——逐章写入请用 `block_insert_after` 并指定对应标题的 `--block-id` | `--content` |
 | `block_move_after` | 移动已有 block 到指定位置 | `--block-id` `--src-block-ids` |
 
+## Block ID 生命周期
+
+写操作后不要默认复用之前 fetch 到的 block ID：
+
+- `overwrite` / `block_replace` / `block_delete`：受影响旧 ID 失效，继续 block 级操作前重新 fetch
+- `block_insert_after` / `append` / `block_copy_insert_after`：锚点 / 源 ID 通常保留，新内容是新 ID；要操作新内容先重新 fetch
+- `block_move_after`：被移动 ID 通常保留，但位置、章节、range 语义变化；后续依赖位置时重新 fetch
+- `str_replace`：简单行内替换通常不改变 ID；跨行 / 大段替换后如继续 block 级操作，先重新 fetch
+
 ## 指令示例
 
 ### str_replace — 全文文本替换
@@ -114,8 +123,6 @@ lark-cli docs +update --api-version v2 --doc "<doc_id>" --command block_replace
   --content '<p>替换后的段落内容</p>'
 ```
 
-> `block_replace` 由服务端执行整块替换，目标 block 的 ID 不保证在替换后继续可用。后续如果还要在替换后的块附近继续 `block_insert_after`、`range` 或其他 block 级操作，先重新 `docs +fetch --detail with-ids` 获取最新 block ID，不要复用旧 ID。
-
 ### block_delete — 删除指定 block
 
 ```bash
@@ -237,7 +244,6 @@ lark-cli docs +update --api-version v2 --doc "<doc_id>" --command str_replace \
 - **保护不可重建的内容**：图片、画板、电子表格等以 token 形式存储，替换时避开这些 block
 - **str_replace 的 replacement 支持富文本**：可以用行内标签 `<b>`、`<a>`、`<cite>`、`<latex>` 等替换普通文本为富文本
 - **同一 block 只能被 replace 一次**：多次修改同一 block 请合并为一次 block_replace
-- **block_replace 后重新获取 ID**：`block_replace` 成功后旧 block ID 不保证继续可用；继续做相邻块操作前，重新 `docs +fetch --detail with-ids`
 - **block_delete 支持批量**：用逗号分隔多个 block_id 一次删除
 - **复杂结构重组**：将多个段落转换为 grid / table 等复杂布局时，分步操作比 overwrite 更安全：
   1. 用 `block_insert_after` 在目标位置插入新的富文本结构

skills/lark-doc/references/lark-doc-whiteboard.md

@@ -9,11 +9,11 @@
 | `lark-doc`        | 识别画板机会、使用 Mermaid/SVG 创建图表、调度 SubAgent、插入简单 SVG 画板或复杂空白画板 | 主 Agent 不直接创作画板内容；              |
 | `lark-whiteboard` | 查询/导出已有画板；复杂图表生成（Mermaid/DSL/SVG 路由、场景选型、渲染验证）；写入已有/空白画板  | 仅特别复杂的图表或已有画板更新时由独立 SubAgent 读取 |
 
-## 画板优先规则
+## 画板适用规则
 
-写文档时，重要信息优先画板化。遇到核心流程、系统架构、方案对比、风险链路、里程碑、指标趋势、因果归因、组织关系、能力分层等内容，不要只用段落或表格承载；除非内容只是一次性补充说明，否则应规划为画板。
+写文档时，核心流程、系统架构、方案对比、风险链路、里程碑、指标趋势、因果归因、组织关系、能力分层等内容，如果图示能明显降低理解成本，可以规划为画板；结构简单或文字更清楚的内容不必强行画板化。
 
-同一篇文档可以有多个画板。优先设计多个聚焦画板，而不是把所有信息塞进一张大图。
+同一篇文档可以有多个画板。确有多个独立图示点时，可拆成多个聚焦画板，而不是把所有信息塞进一张大图。
 
 ## 文档与画板协同流程
 

skills/lark-doc/references/style/lark-doc-create-workflow.md

@@ -43,7 +43,7 @@
 8. **优先处理步骤三识别出的画板需求**：
    参考 [lark-doc-whiteboard.md](../lark-doc-whiteboard.md)中的方式，插入图表画板。
 9. Spawn 内容改写 Agent 定向润色：
-   - 文字密集且不易读的章节可转为 `<table>`/`<grid>`/`<callout>`，也可以拆段、改列表或保留纯文本
+   - 文字密集且不易读时，优先拆段、改列表、增加小标题或调整顺序；只有确实存在行列数据、并列对比或强提醒信息时，才考虑 `<table>` / `<grid>` / `<callout>`
    - 需要明显分隔的主题可补充 `<hr/>`，不强制章节间都使用
    - 本地图片使用 `docs +media-insert` 插入
 

skills/lark-doc/references/style/lark-doc-style.md

@@ -10,18 +10,18 @@
 2. **尊重用户风格**：用户给出样例、语气、结构或已有文档时，优先沿用；没有要求时不强行使用固定开头、固定章节或固定视觉组件
 3. **适度结构化**：结构化 block 用于降低理解成本，不为了“丰富”而堆叠
 4. **保持一致但不过度统一**：同类信息可使用相近表达，但允许因内容差异采用不同形式
-5. **重要信息画板化**：核心流程、架构、对比、风险、路线图、指标趋势等重要信息优先使用画板表达
+5. **图示服务理解**：流程、架构、对比、风险、路线图、指标趋势等内容在图示明显降低理解成本时，可使用画板表达
 
 ## 二、元素选择指南
 
 需要图表时，按类型选择插入方式：思维导图/时序图/类图/饼图/甘特图可用 `<whiteboard type="mermaid">` 直接内嵌；其他新图表可启动 SubAgent 插入 `<whiteboard type="svg">完整 SVG</whiteboard>`；只有编辑**已有**画板时才调用 **lark-whiteboard** skill。
 
-| 场景                                         | 推荐方案                                  |
+| 场景                                         | 可选表达方式                                |
 |--------------------------------------------|---------------------------------------|
-| 需要突出的一小段结论 / 摘要 / 注意事项                    | `<callout>`；是否使用 emoji 和颜色由文档语气决定 |
-| 方案对比 / 优劣势 / Before vs After                | `<grid>` 2 列分栏、`<table>` 或画板，按复杂度选择 |
+| 少数需要视觉提醒的短句，如风险、限制、待确认事项或关键提醒          | 需要视觉提醒时可用 `<callout>`；普通结论、摘要或章节导语优先使用段落、列表、小标题或加粗 |
+| 方案对比 / 优劣势 / Before vs After                | 简短对比可用段落、列表或 `<grid>`；维度较多且需要逐项比较时再考虑 `<table>` 或画板 |
 | 简短低风险对比                                    | `<grid>` 2 列分栏                        |
-| 3+ 属性的结构化数据 / 指标表                          | `<table>` + 表头背景色                     |
+| 需要按行列精确比较或查阅的数据，如指标、清单、字段说明、排期           | 可用 `<table>`；短要点、步骤、摘要或普通说明优先使用段落、列表或小标题 |
 | 任务清单 / 检查项                                 | `<checkbox>`                          |
 | 代码片段                                       | `<pre lang="x" caption="说明">`         |
 | 引用 / 公式                                    | `<blockquote>` / `<latex>`            |

skills/lark-doc/references/style/lark-doc-update-workflow.md

@@ -34,7 +34,7 @@
    参考 [lark-doc-whiteboard.md](../lark-doc-whiteboard.md)中的方式，插入图表画板。
 6. Spawn 内容改写 Agent 在不重叠的章节上并行改进，各 Agent 收到文档 token 和特定 block ID：
    - 沿用或轻微调整已有文档风格，除非用户要求彻底重排版
-   - 可以通过重写段落、调整标题、拆分列表、补表格/分栏/callout 等方式提升可读性
+   - 优先通过重写段落、调整标题、拆分列表或补充小标题提升可读性
    - 富 block 是可选表达手段，不因固定比例而添加；画板类需求只走第 5 步
 
 ### 步骤三：验证（串行）

skills/lark-event/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: lark-event
 version: 1.0.0
-description: "Lark/Feishu real-time event listening / subscribing / consuming: stream events as NDJSON via `lark-cli event consume <EventKey>` (covers IM messages/reactions/chat changes, VC meeting ended, Minutes generated, Whiteboard updated, etc.). Use for Lark bots, real-time message processing, long-running subscribers, streaming webhook/push handlers. Supports `--max-events` / `--timeout` bounded runs and a stderr ready-marker contract — designed for AI agents running as subprocesses."
+description: "Lark/Feishu real-time event listening / subscribing / consuming: stream events as NDJSON via `lark-cli event consume <EventKey>` (covers IM messages/reactions/chat changes, Task updates, VC meeting ended, Minutes generated, Whiteboard updated, etc.). Use for Lark bots, real-time message processing, long-running subscribers, streaming webhook/push handlers. Supports `--max-events` / `--timeout` bounded runs and a stderr ready-marker contract — designed for AI agents running as subprocesses."
 metadata:
   requires:
     bins: ["lark-cli"]
@@ -148,6 +148,7 @@ Lark-defined semantic tags (**not** JSON Schema's standard `format`). Common val
 | Topic      | Reference                                                                    | Coverage |
 |------------|------------------------------------------------------------------------------|---|
 | IM         | [`references/lark-event-im.md`](references/lark-event-im.md)                 | Catalog of 11 IM EventKeys + shape notes (flat vs V2 envelope) + `im.message.receive_v1` field gotchas (`sender_id` is open_id only; `.content` is plain text except for `interactive` cards) + common jq recipes (filter by chat_type / message_type / sender) |
+| Task       | [`references/lark-event-task.md`](references/lark-event-task.md)             | Catalog of 1 Task EventKey (`task.task.update_user_access_v2`) + Native V2 envelope shape + task commit types + user/bot subscription notes |
 | VC         | [`references/lark-event-vc.md`](references/lark-event-vc.md)                 | Catalog of 2 VC EventKeys (`vc.meeting.participant_meeting_ended_v1`, `vc.note.generated_v1`) + field reference + source type semantics (meeting only) |
 | Minutes    | [`references/lark-event-minutes.md`](references/lark-event-minutes.md)       | Catalog of 1 Minutes EventKey (`minutes.minute.generated_v1`) + field reference + source type semantics (meeting only) |
 | Whiteboard | [`references/lark-event-whiteboard.md`](references/lark-event-whiteboard.md) | Catalog of 1 Board EventKey (`board.whiteboard.updated_v1`) + per-whiteboard subscription model (requires `-p whiteboard_id=<token>`) + payload field reference (whiteboard_id / operator_ids triple-id) |

skills/lark-event/references/lark-event-task.md

@@ -0,0 +1,78 @@
+# Task Events
+
+> **Prerequisite:** Read [`../SKILL.md`](../SKILL.md) first for the `event consume` essentials (commands, subprocess contract, jq usage).
+
+## Key catalog (1)
+
+| EventKey | Purpose |
+|---|---|
+| `task.task.update_user_access_v2` | A visible task has been created, deleted, or updated |
+
+This key uses a **Native schema** (V2 envelope; output rooted at `.event`) and carries a **PreConsume hook** that calls the Task event subscription API before listening.
+
+## Scopes & auth
+
+| EventKey | Scope | Auth |
+|---|---|---|
+| `task.task.update_user_access_v2` | `task:task:read` | user, bot |
+
+Supports `--as user` or `--as bot`.
+
+- `--as user`: receive task updates visible to the current user through authorship, assignment, following, or other access.
+- `--as bot`: receive task updates for tasks the application is responsible for.
+
+## `task.task.update_user_access_v2`
+
+### Subscription behavior
+
+On startup, `event consume` calls:
+
+```text
+POST /open-apis/task/v2/task_v2/task_subscription?user_id_type=open_id
+```
+
+The Task subscription API has no matching unsubscribe endpoint in the current CLI metadata, so graceful exit has no cleanup call for this EventKey. Re-running the consumer repeats the subscribe call for the selected identity.
+
+This EventKey is single-consumer per local bus subscription: start one `event consume task.task.update_user_access_v2` process for a given app/profile/identity at a time.
+
+### Output fields (V2 envelope; root path `.event`)
+
+| Field | Type | Description |
+|---|---|---|
+| `.header.event_id` | string | Globally unique event ID; safe for deduplication |
+| `.header.create_time` | string (timestamp_ms) | Event creation time in milliseconds |
+| `.event.event_types[]` | string enum | Task commit types included in this event |
+| `.event.task_guid` | string (kind=task_guid) | Task GUID that changed |
+
+Commit types:
+
+```text
+task_assignees_update
+task_completed_update
+task_create
+task_deleted
+task_desc_update
+task_followers_update
+task_reminders_update
+task_start_due_update
+task_summary_update
+```
+
+### Example
+
+```bash
+# Stream task update events for the current user
+lark-cli event consume task.task.update_user_access_v2 --as user
+
+# Sample one event for payload inspection
+lark-cli event consume task.task.update_user_access_v2 \
+  --as user --max-events 1 --timeout 2m
+
+# Project to a compact task-update record
+lark-cli event consume task.task.update_user_access_v2 \
+  --as user \
+  --jq '{event_id: .header.event_id, task_guid: .event.task_guid, event_types: .event.event_types, timestamp: .header.create_time}'
+
+# Consume as the app identity
+lark-cli event consume task.task.update_user_access_v2 --as bot
+```

skills/lark-im/SKILL.md

@@ -61,6 +61,10 @@ The four message-pulling shortcuts (`+messages-mget`, `+chat-messages-list`, `+m
 
 Card messages (`interactive` type) are not yet supported for compact conversion in event subscriptions. The raw event data will be returned instead, with a hint printed to stderr.
 
+### Audio Messages
+
+`--audio` sends a voice message and supports only Opus audio files, for example `.opus` files or Ogg Opus (`.ogg`) files. For `mp3`, `wav`, or other non-Opus audio, either convert to `.opus` first and keep using `--audio`, or send the original file as an attachment with `--file`.
+
 ### Flag Types
 
 Flags support two layers:
@@ -110,16 +114,122 @@ 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) |
 
-## 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:
+## API Resources
 
 ```bash
-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
+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
 ```
 
-> **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.
+> **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` |

skills/lark-im/references/lark-im-chat-create.md

@@ -12,24 +12,43 @@ This skill maps to the shortcut: `lark-cli im +chat-create` (internally calls `P
 ## Commands
 
 ```bash
-# Private group (default)
+# Create a private group (default)
 lark-cli im +chat-create --name "My Group"
 
-# Public group (--name required, min 2 chars)
+# Create a public group (name is required and must be at least 2 characters)
 lark-cli im +chat-create --name "Public Group" --type public
 
-# Topic chat (a 话题群; see note under Parameters)
+# Create a topic chat
 lark-cli im +chat-create --name "Topic Group" --chat-mode topic
 
-# 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"
+# Specify the group owner
+lark-cli im +chat-create --name "My Group" --owner ou_xxx
 
-# Bot identity, making the creating bot a manager
-lark-cli im +chat-create --name "My Group" --users "ou_aaa" --as bot --set-bot-manager
+# 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
 ```
 
-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 |
@@ -87,13 +106,6 @@ 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 |
@@ -105,13 +117,43 @@ lark-cli im +messages-send --chat-id "$CHAT_ID" --text "Welcome, everyone!"
 | `external` | Whether the group is external |
 | `share_link` | Group share link (omitted if retrieval fails) |
 
-## Common Errors and Troubleshooting
+## Usage Scenarios
 
-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:
+### 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
 
 | 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

skills/lark-im/references/lark-im-messages-reply.md

@@ -147,6 +147,9 @@ lark-cli im +messages-reply --message-id om_xxx --file ./report.pdf
 # Reply with a local video (--video-cover is required as the video cover)
 lark-cli im +messages-reply --message-id om_xxx --video ./demo.mp4 --video-cover ./cover.png
 
+# Reply with a voice message
+lark-cli im +messages-reply --message-id om_xxx --audio ./voice.opus
+
 # With an idempotency key
 lark-cli im +messages-reply --message-id om_xxx --text "Received" --idempotency-key my-unique-id
 
@@ -159,6 +162,7 @@ lark-cli im +messages-reply --message-id om_xxx --markdown $'## Test\n\nhello' -
 - 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.
+- `--audio` sends a voice message and accepts only Opus audio (`.opus` or Ogg Opus `.ogg`) for local paths and URLs. For `mp3`, `wav`, or other non-Opus audio, convert to `.opus` before using `--audio`, or use `--file` to send the original audio as an attachment.
 
 ## Parameters
 
@@ -173,7 +177,7 @@ lark-cli im +messages-reply --message-id om_xxx --markdown $'## Test\n\nhello' -
 | `--file <path\|url\|key>` | One content option | Cwd-relative local file path, URL, or `file_key` (`file_xxx`)                                                                                                                                 |
 | `--video <path\|url\|key>` | One content option | Cwd-relative local video path, URL, or `file_key` (`file_xxx`); **must be used together with `--video-cover`**                                                                                |
 | `--video-cover <path\|url\|key>` | **Required with `--video`** | Cwd-relative local cover image path, URL, or `image_key` (`img_xxx`)                                                                                                                          |
-| `--audio <path\|url\|key>` | One content option | Cwd-relative local audio path, URL, or `file_key` (`file_xxx`)                                                                                                                                            |
+| `--audio <path\|url\|key>` | One content option | Voice-message audio key, URL, or cwd-relative local path. Local paths and URLs must be Opus (`.opus` or Ogg Opus `.ogg`) |
 | `--reply-in-thread` | No | Reply inside the thread. The reply appears in the target message's thread instead of the main chat stream                                                                                     |
 | `--idempotency-key <key>` | No | Idempotency key; the same key sends only one reply within 1 hour                                                                                                                              |
 | `--as <identity>` | No | Identity type: `bot` or `user` (default `bot`)                                                                                                                                                |

skills/lark-im/references/lark-im-messages-send.md

@@ -1,8 +1,10 @@
 # im +messages-send
 
-> **Prerequisite:** Read [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) first for authentication, global parameters, and safety rules.
+> **Prerequisite:** Read [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) first to understand authentication, global parameters, and safety rules.
 
-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`).
+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`).
 
 ## Safety Constraints
 
@@ -14,94 +16,250 @@ Messages sent by this tool are visible to other people. Before calling it, you *
 
 **Do not** send messages without explicit user approval.
 
-- `--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.
+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.
 
 ## Choose The Right Content Flag
 
-| 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 |
+### Default Selection Rule For Agents
 
-These content flags (and the media flags) are **mutually exclusive** — pass exactly one. Media flags are also mutually exclusive with each other.
+- 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.
 
-## `--markdown` Gotchas
+| 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 |
 
-`--markdown` always forces `msg_type=post` (single `zh_cn` locale) and normalizes input for Feishu post rendering. Key boundaries (not full CommonMark/GFM):
+### `--text` vs `--markdown`
 
-- **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.
+- 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.
 
-## Preserving Exact Formatting
+## What `--markdown` Really Does
 
-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:
+`--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.
+
+### Image Constraint for `--markdown`
+
+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.
+
+**Steps:**
 
 ```bash
-lark-cli im +messages-send --chat-id oc_xxx --text $'Build failed\nBranch: feature/x\nAction: check logs'
+# 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.'
 ```
 
+## Preserving 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'
+```
+
+```bash
+lark-cli im +messages-send --chat-id oc_xxx --text $'```bash\nmake test\nmake lint\n```'
+```
+
+Use this path when you want the receiver to see the text exactly as entered, not a converted Markdown post.
+
 ## Commands
 
 ```bash
-# Formatted update (Markdown → post)
+# Send a formatted update
 lark-cli im +messages-send --chat-id oc_xxx --markdown $'## Update\n\n- item 1\n- item 2'
 
-# Plain one-line text
+# Send a plain one-line message
 lark-cli im +messages-send --chat-id oc_xxx --text "Hello"
 
-# Direct message (pass open_id)
+# Equivalent manual JSON
+lark-cli im +messages-send --chat-id oc_xxx --content '{"text":"Hello"}'
+
+# Send to a direct message (pass open_id)
 lark-cli im +messages-send --user-id ou_xxx --text "Hello"
 
-# Exact post structure with a title
+# 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
 lark-cli im +messages-send --chat-id oc_xxx --msg-type post --content '{"zh_cn":{"title":"Title","content":[[{"tag":"text","text":"Body"}]]}}'
 
-# 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.'
-
-# Media (local files uploaded automatically; --video requires --video-cover)
+# 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
+
+# Send a local file (uploaded automatically before sending)
 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 a voice message
 lark-cli im +messages-send --chat-id oc_xxx --audio ./voice.opus
 
-# 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
+# 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
 ```
 
-Run `lark-cli im +messages-send --help` for the full flag list and types. Load-bearing rules that `--help` may not make obvious:
+## Media Input Rules
 
-- **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.
+- 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.
+- `--audio` sends a voice message and accepts only Opus audio (`.opus` or Ogg Opus `.ogg`) for local paths and URLs. For `mp3`, `wav`, or other non-Opus audio, convert to `.opus` before using `--audio`, or use `--file` to send the original audio as an attachment.
 
-## `content` Format Reference (for `--content`)
+## 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 | Voice-message audio key, URL, or cwd-relative local path. Local paths and URLs must be Opus (`.opus` or Ogg Opus `.ogg`) |
+| `--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
 
 | `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` / `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) |
-
-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 normalizes mentions for `text` and `post` only — `interactive` cards are passed through verbatim.
-
-- **`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>`.
+| `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) |
 
 ## Return Value
 
 ```json
-{"message_id": "om_xxx", "chat_id": "oc_xxx", "create_time": "1234567890"}
+{
+  "message_id": "om_xxx",
+  "chat_id": "oc_xxx",
+  "create_time": "1234567890"
+}
 ```
+
+## @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`
+
+- 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)
+
+### `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
+
+- `--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