opt(round-003): 复测 效果+0.200 token+17 耗时-1612ms → 采纳

* feat: surface search API notices

sa: safe
doc: none
cfg: none
test: unit test

* fix: surface search notices in default output

* docs: add search notice doc comments

* docs: expand search notice doc comments

cmd/service/paramflags_test.go

@@ -347,89 +347,6 @@ func boolIntQueryMethod(required bool) meta.Method {
 	})
 }
 
-func slidesSpec() meta.Service {
-	return meta.ServiceFromMap(map[string]interface{}{
-		"name":        "slides",
-		"servicePath": "/open-apis/slides_ai/v1",
-	})
-}
-
-func slidesXMLPresentationSlideGetMethod() meta.Method {
-	return meta.FromMap(map[string]interface{}{
-		"path":       "xml_presentations/{xml_presentation_id}/slide",
-		"httpMethod": "GET",
-		"parameters": map[string]interface{}{
-			"xml_presentation_id": map[string]interface{}{"type": "string", "location": "path", "required": true},
-			"slide_id":            map[string]interface{}{"type": "string", "location": "query"},
-			"slide_number":        map[string]interface{}{"type": "integer", "location": "query"},
-		},
-	})
-}
-
-func slidesXMLPresentationsGetMethod() meta.Method {
-	return meta.FromMap(map[string]interface{}{
-		"path":       "xml_presentations/{xml_presentation_id}",
-		"httpMethod": "GET",
-		"parameters": map[string]interface{}{
-			"xml_presentation_id": map[string]interface{}{"type": "string", "location": "path", "required": true},
-			"revision_id":         map[string]interface{}{"type": "integer", "location": "query"},
-			"remove_attr_id":      map[string]interface{}{"type": "boolean", "location": "query"},
-		},
-	})
-}
-
-func TestSlidesXMLPresentationsGet_RemoveAttrID(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
-	cmd := NewCmdServiceMethod(f, slidesSpec(), slidesXMLPresentationsGetMethod(), "get", "xml_presentations", nil)
-	cmd.SetArgs([]string{"--xml-presentation-id", "pid_123", "--remove-attr-id", "--dry-run"})
-
-	if err := cmd.Execute(); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	out := stdout.String()
-	for _, want := range []string{"/open-apis/slides_ai/v1/xml_presentations/pid_123", `"remove_attr_id": true`} {
-		if !strings.Contains(out, want) {
-			t.Errorf("expected dry-run output to contain %q, got:\n%s", want, out)
-		}
-	}
-}
-
-func TestSlidesXMLPresentationSlideGet_BySlideNumber(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
-	cmd := NewCmdServiceMethod(f, slidesSpec(), slidesXMLPresentationSlideGetMethod(), "get", "xml_presentation.slide", nil)
-	cmd.SetArgs([]string{"--xml-presentation-id", "pid_123", "--slide-number", "2", "--dry-run"})
-
-	if err := cmd.Execute(); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	out := stdout.String()
-	for _, want := range []string{"/open-apis/slides_ai/v1/xml_presentations/pid_123/slide", `"slide_number": 2`} {
-		if !strings.Contains(out, want) {
-			t.Errorf("expected dry-run output to contain %q, got:\n%s", want, out)
-		}
-	}
-	if strings.Contains(out, "slide_id") {
-		t.Errorf("slide_number-only request should not include slide_id, got:\n%s", out)
-	}
-}
-
-func TestSlidesXMLPresentationSlideGet_SlideIDWinsOverSlideNumber(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
-	cmd := NewCmdServiceMethod(f, slidesSpec(), slidesXMLPresentationSlideGetMethod(), "get", "xml_presentation.slide", nil)
-	cmd.SetArgs([]string{"--xml-presentation-id", "pid_123", "--slide-id", "sld_123", "--slide-number", "2", "--dry-run"})
-
-	if err := cmd.Execute(); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	out := stdout.String()
-	if !strings.Contains(out, `"slide_id": "sld_123"`) {
-		t.Errorf("expected slide_id in dry-run output, got:\n%s", out)
-	}
-	if strings.Contains(out, "slide_number") {
-		t.Errorf("slide_id should take precedence over slide_number, got:\n%s", out)
-	}
-}
-
 // Presence is intent: a typed flag is only overlaid when explicitly Changed,
 // so --flag=false / --flag 0 are real values and must be sent — not silently
 // dropped as "empty", which would let the API default win over an explicit

cmd/service/service.go

@@ -566,7 +566,6 @@ func buildServiceRequest(opts *ServiceMethodOptions) (client.RawApiRequest, *cmd
 	for name, value := range params {
 		queryParams[name] = value
 	}
-	normalizeServiceQueryParams(opts.SchemaPath, queryParams)
 
 	request := client.RawApiRequest{
 		Method: httpMethod,
@@ -624,15 +623,6 @@ func buildServiceRequest(opts *ServiceMethodOptions) (client.RawApiRequest, *cmd
 	return request, nil, nil
 }
 
-func normalizeServiceQueryParams(schemaPath string, queryParams map[string]interface{}) {
-	if schemaPath != "slides.xml_presentation.slide.get" {
-		return
-	}
-	if slideID, ok := queryParams["slide_id"]; ok && !unusableParamValue(slideID) {
-		delete(queryParams, "slide_number")
-	}
-}
-
 func serviceDryRun(f *cmdutil.Factory, request client.RawApiRequest, config *core.CliConfig, format string) error {
 	return cmdutil.PrintDryRun(f.IOStreams.Out, request, config, format)
 }

harness-opt/.gitignore

@@ -0,0 +1,5 @@
+# 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

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

harness-opt/baseline/noise-floor.json

@@ -0,0 +1,50 @@
+{
+  "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

@@ -0,0 +1,33 @@
+{
+  "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

@@ -0,0 +1,869 @@
+{
+  "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

@@ -0,0 +1,48 @@
+{
+  "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

@@ -0,0 +1,60 @@
+{
+  "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

@@ -0,0 +1,13 @@
+# 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

@@ -0,0 +1,12 @@
+{
+  "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

@@ -0,0 +1,35 @@
+{
+  "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

@@ -0,0 +1,35 @@
+{
+  "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

@@ -0,0 +1,35 @@
+{
+  "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

@@ -0,0 +1,35 @@
+{
+  "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

@@ -0,0 +1,67 @@
+[
+  {
+    "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

@@ -0,0 +1,24 @@
+{
+  "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

@@ -0,0 +1,29 @@
+{
+  "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

@@ -0,0 +1,97 @@
+# 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

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

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

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

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

@@ -0,0 +1,222 @@
+{
+  "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

@@ -0,0 +1,15 @@
+{
+  "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

@@ -0,0 +1,404 @@
+{
+  "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

@@ -0,0 +1,44 @@
+# 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

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

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

@@ -0,0 +1,35 @@
+# 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

@@ -0,0 +1,65 @@
+[
+  {
+    "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

@@ -0,0 +1,27 @@
+{
+  "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

@@ -0,0 +1,11 @@
+{
+  "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

@@ -0,0 +1,113 @@
+# 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

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

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

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

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

@@ -0,0 +1,220 @@
+{
+  "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

@@ -0,0 +1,15 @@
+{
+  "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

@@ -0,0 +1,380 @@
+{
+  "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

@@ -0,0 +1,48 @@
+# 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

@@ -0,0 +1,11 @@
+[
+  {
+    "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

@@ -0,0 +1,43 @@
+# 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

@@ -0,0 +1,59 @@
+[
+  {
+    "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

@@ -0,0 +1,27 @@
+{
+  "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

@@ -0,0 +1,20 @@
+{
+  "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

@@ -0,0 +1,119 @@
+# 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

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

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

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

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

@@ -0,0 +1,220 @@
+{
+  "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

@@ -0,0 +1,18 @@
+{
+  "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

@@ -0,0 +1,394 @@
+{
+  "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

@@ -0,0 +1,43 @@
+# 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

@@ -0,0 +1,20 @@
+[
+  {
+    "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

@@ -0,0 +1,44 @@
+# 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

@@ -0,0 +1,20 @@
+[
+  {
+    "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

@@ -0,0 +1,152 @@
+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

@@ -0,0 +1,334 @@
+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

@@ -0,0 +1,135 @@
+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)
+

internal/cmdutil/secheader.go

@@ -75,8 +75,6 @@ func BaseSecurityHeaders() http.Header {
 	h.Set(HeaderVersion, build.Version)
 	h.Set(HeaderBuild, DetectBuildKind())
 	h.Set(HeaderUserAgent, UserAgentValue())
-	h.Set("x-tt-env", "ppe_pdf2slide")
-	h.Set("x-use-ppe", "1")
 	if v := AgentTraceValue(); v != "" {
 		h.Set(HeaderAgentTrace, v)
 	}

internal/core/types.go

@@ -30,10 +30,10 @@ const OAuthTokenV3Path = "/oauth/v3/token"
 
 // Endpoints holds resolved endpoint URLs for different Lark services.
 type Endpoints struct {
-	Open     string // e.g. "https://open.feishu-pre.cn"
-	Accounts string // e.g. "https://accounts.feishu-pre.cn"
+	Open     string // e.g. "https://open.feishu.cn"
+	Accounts string // e.g. "https://accounts.feishu.cn"
 	MCP      string // e.g. "https://mcp.feishu.cn"
-	AppLink  string // e.g. "https://applink.feishu-pre.cn"
+	AppLink  string // e.g. "https://applink.feishu.cn"
 }
 
 // ResolveEndpoints resolves endpoint URLs based on brand.
@@ -48,10 +48,10 @@ func ResolveEndpoints(brand LarkBrand) Endpoints {
 		}
 	default:
 		return Endpoints{
-			Open:     "https://open.feishu-pre.cn",
-			Accounts: "https://accounts.feishu-pre.cn",
+			Open:     "https://open.feishu.cn",
+			Accounts: "https://accounts.feishu.cn",
 			MCP:      "https://mcp.feishu.cn",
-			AppLink:  "https://applink.feishu-pre.cn",
+			AppLink:  "https://applink.feishu.cn",
 		}
 	}
 }

internal/meta/normalize.go

@@ -113,7 +113,8 @@ type EnumOption struct {
 }
 
 // EnumOptions returns the field's allowed values paired with their descriptions
-// — from enum, or from options when enum is absent — coerced to the canonical
+// — from enum (with descriptions backfilled from options when the field carries
+// both forms), or from options when enum is absent — coerced to the canonical
 // type and ordered: numeric and boolean values are sorted; string values keep
 // source order (which can encode priority). Uncoercible literals are dropped.
 // Returns nil when the field declares no enum constraint.
@@ -122,9 +123,14 @@ func (f Field) EnumOptions() []EnumOption {
 	var out []EnumOption
 	switch {
 	case len(f.Enum) > 0:
+		// key by raw literal so enum "1" and option 1 align across JSON types
+		desc := make(map[string]string, len(f.Options))
+		for _, o := range f.Options {
+			desc[fmt.Sprintf("%v", o.Value)] = o.Description
+		}
 		for _, e := range f.Enum {
 			if v, ok := coerceLiteral(ct, e); ok {
-				out = append(out, EnumOption{Value: v})
+				out = append(out, EnumOption{Value: v, Description: desc[fmt.Sprintf("%v", e)]})
 			}
 		}
 	case len(f.Options) > 0:

internal/meta/normalize_test.go

@@ -80,6 +80,39 @@ func TestField_EnumOptions(t *testing.T) {
 	}
 }
 
+func TestField_EnumOptions_BothEnumAndOptions(t *testing.T) {
+	// enum is the value set; descriptions backfilled from options, empty where absent
+	f := Field{Type: "string", Enum: []any{"1", "2", "3", "4", "6"}, Options: []Option{
+		{Value: "1", Description: "from"},
+		{Value: "2", Description: "to"},
+		{Value: "6", Description: "subject"},
+	}}
+	want := []EnumOption{
+		{Value: "1", Description: "from"},
+		{Value: "2", Description: "to"},
+		{Value: "3", Description: ""},
+		{Value: "4", Description: ""},
+		{Value: "6", Description: "subject"},
+	}
+	if got := f.EnumOptions(); !reflect.DeepEqual(got, want) {
+		t.Errorf("EnumOptions(enum+options) = %+v, want %+v", got, want)
+	}
+
+	// enum values stored as strings match option values stored as numbers
+	fi := Field{Type: "integer", Enum: []any{"10", "2", "1"}, Options: []Option{
+		{Value: 1, Description: "one"},
+		{Value: 2, Description: "two"},
+	}}
+	wantI := []EnumOption{
+		{Value: int64(1), Description: "one"},
+		{Value: int64(2), Description: "two"},
+		{Value: int64(10), Description: ""},
+	}
+	if got := fi.EnumOptions(); !reflect.DeepEqual(got, wantI) {
+		t.Errorf("EnumOptions(integer enum+options) = %+v, want %+v", got, wantI)
+	}
+}
+
 func TestField_Enum_NumberAndBoolean(t *testing.T) {
 	// number: string-stored floats coerced to float64 and numerically sorted
 	if got := (Field{Type: "number", Enum: []any{"2.5", "1.5", "10"}}).EnumValues(); !reflect.DeepEqual(got, []any{1.5, 2.5, float64(10)}) {

internal/schema/assembler_test.go

@@ -472,6 +472,18 @@ func TestConvert_EnumDescriptions(t *testing.T) {
 	if bare.EnumDescriptions != nil {
 		t.Errorf("bare enum must have nil EnumDescriptions, got %v", bare.EnumDescriptions)
 	}
+
+	// enum + options both present -> enumDescriptions backfilled, aligned, "" where absent
+	both := Convert(meta.Field{Type: "string", Enum: []any{"1", "2", "3"}, Options: []meta.Option{
+		{Value: "1", Description: "from"},
+		{Value: "2", Description: "to"},
+	}})
+	if !reflect.DeepEqual(both.Enum, []interface{}{"1", "2", "3"}) {
+		t.Errorf("both Enum = %v", both.Enum)
+	}
+	if !reflect.DeepEqual(both.EnumDescriptions, []string{"from", "to", ""}) {
+		t.Errorf("both EnumDescriptions = %v, want [from to \"\"] aligned with enum", both.EnumDescriptions)
+	}
 }
 
 func TestBuildMeta_AffordanceFromMethod(t *testing.T) {

shortcuts/apps/git_credential.go

@@ -17,6 +17,7 @@ import (
 	"text/tabwriter"
 	"time"
 
+	"github.com/google/uuid"
 	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
 	"github.com/spf13/cobra"
 
@@ -32,6 +33,7 @@ import (
 )
 
 const gitCredentialIssuePath = apiBasePath + "/apps/:app_id/git_info"
+const gitCredentialHelperReportedShortcut = appsService + ":+git-credential-helper"
 
 // gitCredentialIssueHint is the actionable next-step attached to a failed
 // Git-credential issuance. A 5xx is flagged retryable separately at the call site.
@@ -302,7 +304,12 @@ func (i factoryIssuer) Issue(ctx context.Context, appID string, profile gitcred.
 		HttpMethod: http.MethodGet,
 		ApiPath:    issuePath(appID),
 	}
-	resp, err := ac.DoSDKRequest(ctx, req, core.AsUser)
+	ctx = contextWithGitCredentialHelperShortcut(ctx)
+	var opts []larkcore.RequestOptionFunc
+	if optFn := cmdutil.ShortcutHeaderOpts(ctx); optFn != nil {
+		opts = append(opts, optFn)
+	}
+	resp, err := ac.DoSDKRequest(ctx, req, core.AsUser, opts...)
 	data, err := parseIssueCredentialData(resp, err, errclass.ClassifyContext{
 		Brand:    string(cfg.Brand),
 		AppID:    cfg.AppID,
@@ -314,6 +321,13 @@ func (i factoryIssuer) Issue(ctx context.Context, appID string, profile gitcred.
 	return issuedFromData(appID, data)
 }
 
+func contextWithGitCredentialHelperShortcut(ctx context.Context) context.Context {
+	if _, ok := cmdutil.ShortcutNameFromContext(ctx); ok {
+		return ctx
+	}
+	return cmdutil.ContextWithShortcut(ctx, gitCredentialHelperReportedShortcut, uuid.New().String())
+}
+
 func runGitCredentialHelper(ctx context.Context, f *cmdutil.Factory, appID, action string) error {
 	if f == nil || f.IOStreams == nil {
 		return nil

shortcuts/apps/git_credential_test.go

@@ -825,7 +825,7 @@ func TestRunGitCredentialHelperActions(t *testing.T) {
 func TestFactoryIssuerBranches(t *testing.T) {
 	factory, _, reg := newAppsExecuteFactory(t)
 	expiresAt := time.Now().Add(24 * time.Hour).Unix()
-	reg.Register(&httpmock.Stub{
+	issueStub := &httpmock.Stub{
 		Method: "GET",
 		URL:    "/open-apis/spark/v1/apps/app_xxx/git_info",
 		Body: map[string]interface{}{
@@ -836,7 +836,8 @@ func TestFactoryIssuerBranches(t *testing.T) {
 				"StatusCode": 0,
 			},
 		},
-	})
+	}
+	reg.Register(issueStub)
 	issued, err := (factoryIssuer{f: factory}).Issue(context.Background(), "app_xxx", gitcred.ProfileContext{})
 	if err != nil {
 		t.Fatalf("factory issuer returned error: %v", err)
@@ -844,6 +845,12 @@ func TestFactoryIssuerBranches(t *testing.T) {
 	if issued.PAT != "pat-token" {
 		t.Fatalf("PAT = %q", issued.PAT)
 	}
+	if got := issueStub.CapturedHeaders.Get(cmdutil.HeaderShortcut); got != gitCredentialHelperReportedShortcut {
+		t.Fatalf("%s = %q, want %q", cmdutil.HeaderShortcut, got, gitCredentialHelperReportedShortcut)
+	}
+	if got := issueStub.CapturedHeaders.Get(cmdutil.HeaderExecutionId); got == "" {
+		t.Fatalf("%s header missing", cmdutil.HeaderExecutionId)
+	}
 
 	factory.Config = func() (*core.CliConfig, error) { return nil, errors.New("config failed") }
 	if _, err := (factoryIssuer{f: factory}).Issue(context.Background(), "app_xxx", gitcred.ProfileContext{}); err == nil {
@@ -880,6 +887,20 @@ func TestFactoryIssuerBranches(t *testing.T) {
 	}
 }
 
+func TestContextWithGitCredentialHelperShortcutPreservesExistingShortcut(t *testing.T) {
+	ctx := cmdutil.ContextWithShortcut(context.Background(), "apps:+git-credential-init", "exec-existing")
+	got := contextWithGitCredentialHelperShortcut(ctx)
+
+	name, ok := cmdutil.ShortcutNameFromContext(got)
+	if !ok || name != "apps:+git-credential-init" {
+		t.Fatalf("shortcut = %q ok=%v, want existing shortcut", name, ok)
+	}
+	executionID, ok := cmdutil.ExecutionIdFromContext(got)
+	if !ok || executionID != "exec-existing" {
+		t.Fatalf("execution id = %q ok=%v, want existing execution id", executionID, ok)
+	}
+}
+
 func TestGitCredentialHelpersAndParsers(t *testing.T) {
 	if issuePath(" app/with space ") != "/open-apis/spark/v1/apps/app%2Fwith%20space/git_info" {
 		t.Fatalf("issuePath escaped incorrectly: %s", issuePath(" app/with space "))

shortcuts/contact/contact_search_user.go

@@ -85,6 +85,7 @@ type searchUserAPIData struct {
 	Items     []searchUserAPIItem `json:"items"`
 	HasMore   bool                `json:"has_more"`
 	PageToken string              `json:"page_token"`
+	Notice    string              `json:"notice"`
 }
 
 type searchUserAPIItem struct {
@@ -126,6 +127,7 @@ type searchUser struct {
 type searchUserResponse struct {
 	Users   []searchUser `json:"users"`
 	HasMore bool         `json:"has_more"`
+	Notice  string       `json:"notice,omitempty"`
 }
 
 var ContactSearchUser = common.Shortcut{
@@ -189,6 +191,7 @@ var ContactSearchUser = common.Shortcut{
 	Execute: executeSearchUser,
 }
 
+// executeSearchUser dispatches contact search to single-query or fanout mode.
 func executeSearchUser(ctx context.Context, runtime *common.RuntimeContext) error {
 	if strings.TrimSpace(runtime.Str("queries")) != "" {
 		return executeSearchUserFanout(ctx, runtime)
@@ -196,6 +199,7 @@ func executeSearchUser(ctx context.Context, runtime *common.RuntimeContext) erro
 	return executeSearchUserSingle(ctx, runtime)
 }
 
+// executeSearchUserSingle performs one contact search and preserves server notices.
 func executeSearchUserSingle(ctx context.Context, runtime *common.RuntimeContext) error {
 	body, err := buildSearchUserBody(runtime)
 	if err != nil {
@@ -222,7 +226,7 @@ func executeSearchUserSingle(ctx context.Context, runtime *common.RuntimeContext
 	}
 
 	users, hasMore := projectUsers(respData, runtime.Str("lang"), runtime.Config.Brand)
-	out := searchUserResponse{Users: users, HasMore: hasMore}
+	out := searchUserResponse{Users: users, HasMore: hasMore, Notice: respData.Notice}
 
 	runtime.OutFormat(out, &output.Meta{Count: len(users)}, func(w io.Writer) {
 		if len(users) == 0 {

shortcuts/contact/contact_search_user_fanout.go

@@ -45,22 +45,17 @@ type fanoutResult struct {
 	Query   string
 	Users   []searchUser
 	HasMore bool
+	Notice  string
 	ErrMsg  string // empty = success
 	Err     error  // original failure, kept for typed all-failed propagation
 }
 
-// isFanoutSummaryFormat gates the per-fanout stderr summary line. Includes csv
-// because that summary lives on stderr and never corrupts the csv stream on
-// stdout — single-query mode keeps the narrower isHumanReadableFormat predicate
-// for its refine hint, so adding csv here doesn't regress that path.
+// isFanoutSummaryFormat gates the per-fanout stderr summary line.
 func isFanoutSummaryFormat(format string) bool {
 	return format == "pretty" || format == "table" || format == "csv"
 }
 
-// runOneQuery converts every failure mode (transport, HTTP status, parse,
-// API code) into an ErrMsg string instead of returning a Go error. The
-// fanout dispatcher (Task 6) relies on this so a single failed query never
-// short-circuits the remaining workers.
+// runOneQuery converts one fanout request into either users or an error summary.
 func runOneQuery(ctx context.Context, runtime *common.RuntimeContext, index int, query string,
 	filter *searchUserAPIFilter) fanoutResult {
 	// Pre-check ctx so queued workers see cancellation before issuing a
@@ -94,9 +89,10 @@ func runOneQuery(ctx context.Context, runtime *common.RuntimeContext, index int,
 	}
 
 	users, hasMore := projectUsers(respData, runtime.Str("lang"), runtime.Config.Brand)
-	return fanoutResult{Index: index, Query: query, Users: users, HasMore: hasMore}
+	return fanoutResult{Index: index, Query: query, Users: users, HasMore: hasMore, Notice: respData.Notice}
 }
 
+// fanoutErrorResult records a failed fanout query without stopping other workers.
 func fanoutErrorResult(index int, query string, err error) fanoutResult {
 	if err == nil {
 		return fanoutResult{Index: index, Query: query}
@@ -113,17 +109,16 @@ type querySummary struct {
 	Query   string `json:"query"`
 	Error   string `json:"error,omitempty"`
 	HasMore bool   `json:"has_more"`
+	Notice  string `json:"notice,omitempty"`
 }
 
 type fanoutResponse struct {
 	Users   []fanoutUser   `json:"users"`
 	Queries []querySummary `json:"queries"`
+	Notice  string         `json:"notice,omitempty"`
 }
 
-// buildFanoutResponse walks results by Index (input order), flattens users[]
-// with matched_query, lists every input in queries[] (including successes),
-// and returns an error only when every query failed. The error wraps the
-// first failing query's ErrMsg so the CLI exits non-zero on full failure.
+// buildFanoutResponse flattens ordered fanout results and fails only when all queries fail.
 func buildFanoutResponse(queries []string, results []fanoutResult) (*fanoutResponse, error) {
 	indexed := make([]fanoutResult, len(queries))
 	for _, r := range results {
@@ -142,6 +137,7 @@ func buildFanoutResponse(queries []string, results []fanoutResult) (*fanoutRespo
 			Query:   queries[i],
 			Error:   r.ErrMsg,
 			HasMore: r.HasMore,
+			Notice:  r.Notice,
 		})
 		if r.ErrMsg != "" {
 			failed++
@@ -152,6 +148,9 @@ func buildFanoutResponse(queries []string, results []fanoutResult) (*fanoutRespo
 			}
 			continue
 		}
+		if out.Notice == "" {
+			out.Notice = r.Notice
+		}
 		for _, u := range r.Users {
 			out.Users = append(out.Users, fanoutUser{searchUser: u, MatchedQuery: queries[i]})
 		}

shortcuts/contact/contact_search_user_test.go

@@ -562,6 +562,7 @@ func mountAndRun(t *testing.T, s common.Shortcut, args []string, f *cmdutil.Fact
 	return parent.Execute()
 }
 
+// searchUserStub returns a representative user search response with a notice.
 func searchUserStub() *httpmock.Stub {
 	return &httpmock.Stub{
 		Method: "POST",
@@ -569,6 +570,7 @@ func searchUserStub() *httpmock.Stub {
 		Body: map[string]interface{}{
 			"code": 0, "msg": "ok",
 			"data": map[string]interface{}{
+				"notice": "The query is too long and has been truncated to the first 50 characters for search.",
 				"items": []interface{}{
 					map[string]interface{}{
 						"id": "ou_a",
@@ -590,6 +592,7 @@ func searchUserStub() *httpmock.Stub {
 	}
 }
 
+// TestSearchUser_Integration_PrettyRendersExpectedColumns verifies human output columns.
 func TestSearchUser_Integration_PrettyRendersExpectedColumns(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, searchUserDefaultConfig())
 	reg.Register(searchUserStub())
@@ -614,6 +617,7 @@ func TestSearchUser_Integration_PrettyRendersExpectedColumns(t *testing.T) {
 	}
 }
 
+// TestSearchUser_Integration_JSONStructuredFields verifies normalized JSON and notices.
 func TestSearchUser_Integration_JSONStructuredFields(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, searchUserDefaultConfig())
 	reg.Register(searchUserStub())
@@ -631,6 +635,9 @@ func TestSearchUser_Integration_JSONStructuredFields(t *testing.T) {
 	if !ok {
 		t.Fatalf("envelope.data: expected object, got %v\nraw=%s", got["data"], stdout.String())
 	}
+	if data["notice"] != "The query is too long and has been truncated to the first 50 characters for search." {
+		t.Fatalf("data.notice = %v", data["notice"])
+	}
 	users, _ := data["users"].([]interface{})
 	if len(users) != 1 {
 		t.Fatalf("users: expected 1, got %d (output=%s)", len(users), stdout.String())
@@ -1358,6 +1365,7 @@ func TestSearchUser_Integration_NoAutoPaginationFlags(t *testing.T) {
 	}
 }
 
+// TestFanout_FilterAppliedToEachQuery verifies shared fanout filters reach every request.
 func TestFanout_FilterAppliedToEachQuery(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, searchUserDefaultConfig())
 	stub := &httpmock.Stub{
@@ -1399,6 +1407,7 @@ func TestFanout_FilterAppliedToEachQuery(t *testing.T) {
 	}
 }
 
+// TestFanout_PartialFailure_ExitZero verifies partial fanout failures keep notices.
 func TestFanout_PartialFailure_ExitZero(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, searchUserDefaultConfig())
 	reg.Register(&httpmock.Stub{
@@ -1406,6 +1415,7 @@ func TestFanout_PartialFailure_ExitZero(t *testing.T) {
 		BodyFilter: func(b []byte) bool { return strings.Contains(string(b), `"alice"`) },
 		Body: map[string]interface{}{"code": 0, "msg": "ok",
 			"data": map[string]interface{}{
+				"notice":   "The query is too long and has been truncated to the first 50 characters for search.",
 				"items":    []interface{}{map[string]interface{}{"id": "ou_a"}},
 				"has_more": false,
 			}},
@@ -1432,10 +1442,17 @@ func TestFanout_PartialFailure_ExitZero(t *testing.T) {
 	if len(users) != 1 {
 		t.Errorf("users: expected 1 (alice), got %d; stdout=%s", len(users), stdout.String())
 	}
+	if data["notice"] != "The query is too long and has been truncated to the first 50 characters for search." {
+		t.Fatalf("data.notice = %v", data["notice"])
+	}
 	queries := data["queries"].([]interface{})
 	if len(queries) != 2 {
 		t.Fatalf("queries: expected 2, got %d", len(queries))
 	}
+	q0 := queries[0].(map[string]interface{})
+	if q0["notice"] != "The query is too long and has been truncated to the first 50 characters for search." {
+		t.Fatalf("queries[0].notice = %v", q0["notice"])
+	}
 	q1 := queries[1].(map[string]interface{})
 	if !strings.HasPrefix(q1["error"].(string), "HTTP 500") {
 		t.Errorf("queries[1].error: got %q", q1["error"])

shortcuts/doc/docs_search.go

@@ -74,6 +74,9 @@ var DocsSearch = common.Shortcut{
 			"page_token": data["page_token"],
 			"results":    normalizedItems,
 		}
+		if notice, _ := data["notice"].(string); notice != "" {
+			resultData["notice"] = notice
+		}
 
 		runtime.OutFormat(resultData, &output.Meta{Count: len(normalizedItems)}, func(w io.Writer) {
 			if len(normalizedItems) == 0 {

shortcuts/doc/docs_search_test.go

@@ -7,8 +7,48 @@ import (
 	"encoding/json"
 	"strings"
 	"testing"
+
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/httpmock"
 )
 
+// TestDocsSearchExecutePassesThroughNotice verifies docs +search preserves notices.
+func TestDocsSearchExecutePassesThroughNotice(t *testing.T) {
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, docsTestConfigWithAppID("docs-search-notice"))
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/search/v2/doc_wiki/search",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "ok",
+			"data": map[string]interface{}{
+				"notice":     notice,
+				"res_units":  []interface{}{},
+				"total":      0,
+				"has_more":   false,
+				"page_token": "",
+			},
+		},
+	})
+
+	if err := mountAndRunDocs(t, DocsSearch, []string{"+search", "--query", "incident", "--format", "json", "--as", "user"}, f, stdout); err != nil {
+		t.Fatalf("DocsSearch.Execute() error = %v", err)
+	}
+	reg.Verify(t)
+
+	var env map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &env); err != nil {
+		t.Fatalf("json.Unmarshal(stdout) error = %v\nstdout=%s", err, stdout.String())
+	}
+	data, _ := env["data"].(map[string]interface{})
+	if got, _ := data["notice"].(string); got != notice {
+		t.Fatalf("data.notice = %q, want %q; data=%#v", got, notice, data)
+	}
+}
+
+// TestAddIsoTimeFieldsSupportsJSONNumber verifies JSON numbers get ISO fields.
 func TestAddIsoTimeFieldsSupportsJSONNumber(t *testing.T) {
 	t.Parallel()
 

shortcuts/drive/drive_add_comment.go

@@ -121,7 +121,7 @@ const (
 var DriveAddComment = common.Shortcut{
 	Service:     "drive",
 	Command:     "+add-comment",
-	Description: "Add a comment to doc/docx/file/sheet/slides; file targets support selected extensions and full comments only",
+	Description: "Add a comment to doc/docx/file/sheet/slides/base(bitable); file targets support selected extensions and full comments only",
 	Risk:        "write",
 	Scopes: []string{
 		"drive:drive.metadata:readonly",
@@ -131,12 +131,12 @@ var DriveAddComment = common.Shortcut{
 	},
 	AuthTypes: []string{"user", "bot"},
 	Flags: []common.Flag{
-		{Name: "doc", Desc: "document URL/token, file URL/token, sheet/slides URL, or wiki URL that resolves to doc/docx/file/sheet/slides", Required: true},
-		{Name: "type", Desc: "document type: doc, docx, file, sheet, slides (required when --doc is a bare token; auto-detected for URLs)", Enum: []string{"doc", "docx", "file", "sheet", "slides"}},
+		{Name: "doc", Desc: "document URL/token, file URL/token, sheet/slides/base/bitable URL, or wiki URL that resolves to doc/docx/file/sheet/slides/base(bitable)", Required: true},
+		{Name: "type", Desc: "document type: doc, docx, file, sheet, slides, bitable, base (required when --doc is a bare token; auto-detected for URLs; use bitable as the wire value, base is accepted as a compatibility alias)", Enum: []string{"doc", "docx", "file", "sheet", "slides", "bitable", "base"}},
 		{Name: "content", Desc: "reply_elements JSON string", Required: true, Input: []string{common.File, common.Stdin}},
 		{Name: "full-comment", Type: "bool", Desc: "create a full-document comment; also the default when no location is provided"},
 		{Name: "selection-with-ellipsis", Desc: "target content locator (plain text or 'start...end')"},
-		{Name: "block-id", Desc: "for docx: anchor block ID; for sheet: <sheetId>!<cell> (e.g. a281f9!D6); for slides: <slide-block-type>!<xml-id> (e.g. shape!bPq)"},
+		{Name: "block-id", Desc: "for docx: anchor block ID; for sheet: <sheetId>!<cell>; for slides: <slide-block-type>!<xml-id>; for base(bitable): <table-id>!<record-id>!<view-id>"},
 	},
 	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
 		docRef, err := parseCommentDocRef(runtime.Str("doc"), runtime.Str("type"))
@@ -148,6 +148,17 @@ var DriveAddComment = common.Shortcut{
 			return err
 		}
 
+		if docRef.Kind == "base" {
+			if runtime.Bool("full-comment") {
+				return errs.NewValidationError(errs.SubtypeInvalidArgument, "--full-comment is not applicable for base(bitable) comments; use --block-id <table-id>!<record-id>!<view-id>").WithParam("--full-comment")
+			}
+			if strings.TrimSpace(runtime.Str("selection-with-ellipsis")) != "" {
+				return errs.NewValidationError(errs.SubtypeInvalidArgument, "--selection-with-ellipsis is not applicable for base(bitable) comments; use --block-id <table-id>!<record-id>!<view-id>").WithParam("--selection-with-ellipsis")
+			}
+			_, err := parseBaseCommentAnchor(runtime)
+			return err
+		}
+
 		// Sheet comment validation.
 		if docRef.Kind == "sheet" {
 			blockID := strings.TrimSpace(runtime.Str("block-id"))
@@ -188,7 +199,7 @@ var DriveAddComment = common.Shortcut{
 			return validateFileCommentMode(mode, "")
 		}
 		if mode == commentModeLocal && docRef.Kind == "doc" {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "local comments only support docx, sheet, and slides; old doc format only supports full comments")
+			return errs.NewValidationError(errs.SubtypeInvalidArgument, "local comments only support docx, sheet, slides, and base(bitable); old doc format only supports full comments")
 		}
 
 		return nil
@@ -215,6 +226,23 @@ var DriveAddComment = common.Shortcut{
 			resolvedToken = target.FileToken
 		}
 
+		if resolvedKind == "base" {
+			anchor, err := parseBaseCommentAnchor(runtime)
+			if err != nil {
+				return common.NewDryRunAPI().Set("error", err.Error())
+			}
+			commentBody := buildBaseCommentCreateV2Request(replyElements, anchor)
+			desc := "1-step request: create base(bitable) record-local comment"
+			if isWiki {
+				desc = "2-step orchestration: resolve wiki -> create base(bitable) record-local comment"
+			}
+			return common.NewDryRunAPI().
+				Desc(desc).
+				POST("/open-apis/drive/v1/files/:file_token/new_comments").
+				Body(commentBody).
+				Set("file_token", resolvedToken)
+		}
+
 		// Sheet comment dry-run.
 		if resolvedKind == "sheet" {
 			anchor, _ := parseSheetCellRef(blockID)
@@ -352,6 +380,14 @@ var DriveAddComment = common.Shortcut{
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
 		// Sheet comment: direct URL or token fast path.
 		docRef, _ := parseCommentDocRef(runtime.Str("doc"), runtime.Str("type"))
+		if docRef.Kind == "base" {
+			return executeBaseComment(runtime, resolvedCommentTarget{
+				DocID:      docRef.Token,
+				FileToken:  docRef.Token,
+				FileType:   "base",
+				ResolvedBy: "base",
+			})
+		}
 		if docRef.Kind == "sheet" {
 			return executeSheetComment(runtime, docRef)
 		}
@@ -375,6 +411,9 @@ var DriveAddComment = common.Shortcut{
 		if target.FileType == "slides" {
 			return executeSlidesComment(runtime, commentDocRef{Kind: "slides", Token: target.FileToken})
 		}
+		if target.FileType == "base" {
+			return executeBaseComment(runtime, target)
+		}
 		if target.FileType == "file" {
 			return executeFileComment(runtime, target)
 		}
@@ -482,6 +521,12 @@ func parseCommentDocRef(input, docType string) (commentDocRef, error) {
 	if token, ok := extractURLToken(raw, "/sheets/"); ok {
 		return commentDocRef{Kind: "sheet", Token: token}, nil
 	}
+	if token, ok := extractURLToken(raw, "/base/"); ok {
+		return commentDocRef{Kind: "base", Token: token}, nil
+	}
+	if token, ok := extractURLToken(raw, "/bitable/"); ok {
+		return commentDocRef{Kind: "base", Token: token}, nil
+	}
 	if token, ok := extractURLToken(raw, "/file/"); ok {
 		return commentDocRef{Kind: "file", Token: token}, nil
 	}
@@ -495,7 +540,7 @@ func parseCommentDocRef(input, docType string) (commentDocRef, error) {
 		return commentDocRef{Kind: "doc", Token: token}, nil
 	}
 	if strings.Contains(raw, "://") {
-		return commentDocRef{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "unsupported --doc input %q: use a doc/docx/file/sheet/slides URL, a token with --type, or a wiki URL that resolves to doc/docx/file/sheet/slides", raw).WithParam("--doc")
+		return commentDocRef{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "unsupported --doc input %q: use a doc/docx/file/sheet/slides/base/bitable URL, a token with --type, or a wiki URL that resolves to doc/docx/file/sheet/slides/base(bitable)", raw).WithParam("--doc")
 	}
 	if strings.ContainsAny(raw, "/?#") {
 		return commentDocRef{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "unsupported --doc input %q: use a token with --type, or a wiki URL", raw).WithParam("--doc")
@@ -504,7 +549,10 @@ func parseCommentDocRef(input, docType string) (commentDocRef, error) {
 	// Bare token: --type is required.
 	docType = strings.TrimSpace(docType)
 	if docType == "" {
-		return commentDocRef{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "--type is required when --doc is a bare token (allowed values: doc, docx, file, sheet, slides)").WithParam("--type")
+		return commentDocRef{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "--type is required when --doc is a bare token (allowed values: doc, docx, file, sheet, slides, bitable, base; use bitable as the wire value, base is accepted as a compatibility alias)").WithParam("--type")
+	}
+	if docType == "bitable" || docType == "base" {
+		return commentDocRef{Kind: "base", Token: raw}, nil
 	}
 	return commentDocRef{Kind: docType, Token: raw}, nil
 }
@@ -515,11 +563,11 @@ func resolveCommentTarget(ctx context.Context, runtime *common.RuntimeContext, i
 		return resolvedCommentTarget{}, err
 	}
 
-	if docRef.Kind == "docx" || docRef.Kind == "doc" || docRef.Kind == "file" || docRef.Kind == "sheet" || docRef.Kind == "slides" {
+	if docRef.Kind == "docx" || docRef.Kind == "doc" || docRef.Kind == "file" || docRef.Kind == "sheet" || docRef.Kind == "slides" || docRef.Kind == "base" {
 		if mode == commentModeLocal {
 			switch docRef.Kind {
 			case "doc":
-				return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "local comments only support docx, sheet, and slides; old doc format only supports full comments")
+				return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "local comments only support docx, sheet, slides, and base(bitable); old doc format only supports full comments")
 			case "file":
 				if err := validateFileCommentMode(mode, ""); err != nil {
 					return resolvedCommentTarget{}, err
@@ -557,6 +605,22 @@ func resolveCommentTarget(ctx context.Context, runtime *common.RuntimeContext, i
 	if objType == "slides" && strings.TrimSpace(runtime.Str("selection-with-ellipsis")) != "" {
 		return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but --selection-with-ellipsis is not applicable for slide comments; use --block-id <slide-block-type>!<xml-id>", objType)
 	}
+	if objType == "bitable" || objType == "base" {
+		if runtime.Bool("full-comment") {
+			return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but --full-comment is not applicable for base(bitable) comments; use --block-id <table-id>!<record-id>!<view-id>", objType).WithParam("--full-comment")
+		}
+		if strings.TrimSpace(runtime.Str("selection-with-ellipsis")) != "" {
+			return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but --selection-with-ellipsis is not applicable for base(bitable) comments; use --block-id <table-id>!<record-id>!<view-id>", objType).WithParam("--selection-with-ellipsis")
+		}
+		fmt.Fprintf(runtime.IO().ErrOut, "Resolved wiki to base: %s\n", common.MaskToken(objToken))
+		return resolvedCommentTarget{
+			DocID:      objToken,
+			FileToken:  objToken,
+			FileType:   "base",
+			ResolvedBy: "wiki",
+			WikiToken:  docRef.Token,
+		}, nil
+	}
 	if objType == "sheet" {
 		// Sheet comments are handled via the sheet fast path in Execute.
 		fmt.Fprintf(runtime.IO().ErrOut, "Resolved wiki to %s: %s\n", objType, common.MaskToken(objToken))
@@ -592,10 +656,10 @@ func resolveCommentTarget(ctx context.Context, runtime *common.RuntimeContext, i
 		}, nil
 	}
 	if mode == commentModeLocal && objType != "docx" {
-		return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but local comments only support docx, sheet, and slides; for sheet use --block-id <sheetId>!<cell>, for slides use --block-id <slide-block-type>!<xml-id>", objType)
+		return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but local comments only support docx, sheet, slides, and base(bitable); for sheet use --block-id <sheetId>!<cell>, for slides use --block-id <slide-block-type>!<xml-id>, for base use --block-id <table-id>!<record-id>!<view-id>", objType)
 	}
 	if mode == commentModeFull && objType != "docx" && objType != "doc" {
-		return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but comments only support doc/docx/file/sheet/slides", objType)
+		return resolvedCommentTarget{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "wiki resolved to %q, but comments only support doc/docx/file/sheet/slides/base(bitable)", objType)
 	}
 
 	fmt.Fprintf(runtime.IO().ErrOut, "Resolved wiki to %s: %s\n", objType, common.MaskToken(objToken))
@@ -787,6 +851,12 @@ type sheetAnchor struct {
 	Row     int
 }
 
+type baseAnchor struct {
+	BlockID      string
+	BaseRecordID string
+	BaseViewID   string
+}
+
 func buildCommentCreateV2Request(fileType, blockID, slideBlockType string, replyElements []map[string]interface{}, sheet *sheetAnchor) map[string]interface{} {
 	body := map[string]interface{}{
 		"file_type":      fileType,
@@ -813,6 +883,18 @@ func buildCommentCreateV2Request(fileType, blockID, slideBlockType string, reply
 	return body
 }
 
+func buildBaseCommentCreateV2Request(replyElements []map[string]interface{}, anchor baseAnchor) map[string]interface{} {
+	return map[string]interface{}{
+		"file_type":      "bitable",
+		"reply_elements": replyElements,
+		"anchor": map[string]interface{}{
+			"block_id":       anchor.BlockID,
+			"base_record_id": anchor.BaseRecordID,
+			"base_view_id":   anchor.BaseViewID,
+		},
+	}
+}
+
 func anchorBlockIDForDryRun(blockID string) string {
 	if strings.TrimSpace(blockID) != "" {
 		return strings.TrimSpace(blockID)
@@ -820,6 +902,26 @@ func anchorBlockIDForDryRun(blockID string) string {
 	return "<anchor_block_id>"
 }
 
+func parseBaseCommentAnchor(runtime *common.RuntimeContext) (baseAnchor, error) {
+	blockID := strings.TrimSpace(runtime.Str("block-id"))
+	if blockID == "" {
+		return baseAnchor{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "--block-id is required for base(bitable) record-local comments (format: <table-id>!<record-id>!<view-id>, e.g. tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R)").WithParam("--block-id")
+	}
+	return parseBaseBlockRef(blockID)
+}
+
+func parseBaseBlockRef(blockID string) (baseAnchor, error) {
+	parts := strings.Split(strings.TrimSpace(blockID), "!")
+	if len(parts) != 3 || strings.TrimSpace(parts[0]) == "" || strings.TrimSpace(parts[1]) == "" || strings.TrimSpace(parts[2]) == "" {
+		return baseAnchor{}, errs.NewValidationError(errs.SubtypeInvalidArgument, "base(bitable) record-local comments require --block-id in <table-id>!<record-id>!<view-id> format, e.g. tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R").WithParam("--block-id")
+	}
+	return baseAnchor{
+		BlockID:      strings.TrimSpace(parts[0]),
+		BaseRecordID: strings.TrimSpace(parts[1]),
+		BaseViewID:   strings.TrimSpace(parts[2]),
+	}, nil
+}
+
 func parseSlidesBlockRef(blockID string) (string, string, error) {
 	blockID = strings.TrimSpace(blockID)
 	if blockID == "" {
@@ -1030,6 +1132,53 @@ func executeSheetComment(runtime *common.RuntimeContext, docRef commentDocRef) e
 	return nil
 }
 
+func executeBaseComment(runtime *common.RuntimeContext, target resolvedCommentTarget) error {
+	replyElements, err := parseCommentReplyElements(runtime.Str("content"))
+	if err != nil {
+		return err
+	}
+	anchor, err := parseBaseCommentAnchor(runtime)
+	if err != nil {
+		return err
+	}
+
+	requestPath := fmt.Sprintf("/open-apis/drive/v1/files/%s/new_comments", validate.EncodePathSegment(target.FileToken))
+	requestBody := buildBaseCommentCreateV2Request(replyElements, anchor)
+
+	fmt.Fprintf(runtime.IO().ErrOut, "Creating base(bitable) record-local comment in %s (table=%s, record=%s, view=%s)\n",
+		common.MaskToken(target.FileToken), anchor.BlockID, anchor.BaseRecordID, anchor.BaseViewID)
+
+	data, err := runtime.CallAPITyped("POST", requestPath, nil, requestBody)
+	if err != nil {
+		return err
+	}
+
+	out := map[string]interface{}{
+		"file_token":     target.FileToken,
+		"file_type":      "bitable",
+		"resolved_by":    target.ResolvedBy,
+		"comment_mode":   "base_record",
+		"base_block_id":  anchor.BlockID,
+		"base_record_id": anchor.BaseRecordID,
+		"base_view_id":   anchor.BaseViewID,
+	}
+	if commentID := data["comment_id"]; commentID != nil {
+		out["comment_id"] = commentID
+	}
+	if replyID := data["reply_id"]; replyID != nil {
+		out["reply_id"] = replyID
+	}
+	if createdAt := firstPresentValue(data, "created_at", "create_time"); createdAt != nil {
+		out["created_at"] = createdAt
+	}
+	if target.WikiToken != "" {
+		out["wiki_token"] = target.WikiToken
+	}
+
+	runtime.Out(out, nil)
+	return nil
+}
+
 func executeFileComment(runtime *common.RuntimeContext, target resolvedCommentTarget) error {
 	replyElements, err := parseCommentReplyElements(runtime.Str("content"))
 	if err != nil {

shortcuts/drive/drive_add_comment_test.go

@@ -133,6 +133,20 @@ func TestParseCommentDocRef(t *testing.T) {
 			wantKind:  "file",
 			wantToken: "fileToken",
 		},
+		{
+			name:      "raw token with type bitable",
+			input:     "baseToken",
+			docType:   "bitable",
+			wantKind:  "base",
+			wantToken: "baseToken",
+		},
+		{
+			name:      "raw token with type base alias",
+			input:     "baseToken",
+			docType:   "base",
+			wantKind:  "base",
+			wantToken: "baseToken",
+		},
 		{
 			name:    "raw token without type",
 			input:   "xxxxxx",
@@ -156,6 +170,18 @@ func TestParseCommentDocRef(t *testing.T) {
 			wantKind:  "file",
 			wantToken: "boxcn123",
 		},
+		{
+			name:      "base url",
+			input:     "https://example.larksuite.com/base/baseToken123?table=tbl1",
+			wantKind:  "base",
+			wantToken: "baseToken123",
+		},
+		{
+			name:      "bitable url",
+			input:     "https://example.larksuite.com/bitable/baseToken456?table=tbl1",
+			wantKind:  "base",
+			wantToken: "baseToken456",
+		},
 		{
 			name:    "unsupported url",
 			input:   "https://example.com/not-a-doc",
@@ -726,6 +752,35 @@ func TestBuildCommentCreateV2RequestSheetOverridesBlockID(t *testing.T) {
 	}
 }
 
+func TestBuildBaseCommentCreateV2Request(t *testing.T) {
+	t.Parallel()
+	replyElements := []map[string]interface{}{
+		{"type": "text", "text": "base comment"},
+	}
+	got := buildBaseCommentCreateV2Request(replyElements, baseAnchor{
+		BlockID:      "tbl9mp6fj9kDKHQV",
+		BaseRecordID: "recBIBgGmb",
+		BaseViewID:   "vewc46MG1R",
+	})
+
+	if got["file_type"] != "bitable" {
+		t.Fatalf("expected file_type bitable, got %#v", got["file_type"])
+	}
+	anchor, ok := got["anchor"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("expected anchor map, got %#v", got["anchor"])
+	}
+	if anchor["block_id"] != "tbl9mp6fj9kDKHQV" {
+		t.Fatalf("expected block_id tbl9mp6fj9kDKHQV, got %#v", anchor["block_id"])
+	}
+	if anchor["base_record_id"] != "recBIBgGmb" {
+		t.Fatalf("expected base_record_id recBIBgGmb, got %#v", anchor["base_record_id"])
+	}
+	if anchor["base_view_id"] != "vewc46MG1R" {
+		t.Fatalf("expected base_view_id vewc46MG1R, got %#v", anchor["base_view_id"])
+	}
+}
+
 // ── Sheet cell ref parsing tests ────────────────────────────────────────────
 
 func TestParseSheetCellRef(t *testing.T) {
@@ -985,6 +1040,78 @@ func TestFileCommentValidateRejectsSelectionWithEllipsis(t *testing.T) {
 	}
 }
 
+func TestBaseCommentValidateMissingBlockID(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, driveTestConfig())
+	err := mountAndRunDrive(t, DriveAddComment, []string{
+		"+add-comment",
+		"--doc", "https://example.larksuite.com/base/baseToken",
+		"--content", `[{"type":"text","text":"test"}]`,
+		"--as", "user",
+	}, f, stdout)
+	if err == nil || !strings.Contains(err.Error(), "--block-id is required") {
+		t.Fatalf("expected block-id required error, got: %v", err)
+	}
+}
+
+func TestBaseCommentValidateMalformedBlockID(t *testing.T) {
+	cases := []string{
+		"tbl9mp6fj9kDKHQV",
+		"tbl9mp6fj9kDKHQV!recBIBgGmb",
+		"tbl9mp6fj9kDKHQV!!vewc46MG1R",
+	}
+	for _, blockID := range cases {
+		t.Run(blockID, func(t *testing.T) {
+			f, stdout, _, _ := cmdutil.TestFactory(t, driveTestConfig())
+			err := mountAndRunDrive(t, DriveAddComment, []string{
+				"+add-comment",
+				"--doc", "https://example.larksuite.com/base/baseToken",
+				"--content", `[{"type":"text","text":"test"}]`,
+				"--block-id", blockID,
+				"--as", "user",
+			}, f, stdout)
+			if err == nil || !strings.Contains(err.Error(), "<table-id>!<record-id>!<view-id>") {
+				t.Fatalf("expected block-id format error, got: %v", err)
+			}
+		})
+	}
+}
+
+func TestBaseCommentValidateRejectsIncompatibleFlags(t *testing.T) {
+	cases := []struct {
+		name    string
+		args    []string
+		wantErr string
+	}{
+		{
+			name:    "full comment",
+			args:    []string{"--full-comment"},
+			wantErr: "--full-comment is not applicable for base(bitable) comments",
+		},
+		{
+			name:    "selection",
+			args:    []string{"--selection-with-ellipsis", "some text"},
+			wantErr: "--selection-with-ellipsis is not applicable for base(bitable) comments",
+		},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			f, stdout, _, _ := cmdutil.TestFactory(t, driveTestConfig())
+			args := []string{
+				"+add-comment",
+				"--doc", "https://example.larksuite.com/base/baseToken",
+				"--content", `[{"type":"text","text":"test"}]`,
+				"--block-id", "tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R",
+				"--as", "user",
+			}
+			args = append(args, tc.args...)
+			err := mountAndRunDrive(t, DriveAddComment, args, f, stdout)
+			if err == nil || !strings.Contains(err.Error(), tc.wantErr) {
+				t.Fatalf("expected %q error, got: %v", tc.wantErr, err)
+			}
+		})
+	}
+}
+
 // ── Slides comment execute tests ────────────────────────────────────────────
 
 func TestSlidesCommentExecuteSuccess(t *testing.T) {
@@ -1195,6 +1322,87 @@ func TestSheetCommentViaWikiMissingBlockID(t *testing.T) {
 	}
 }
 
+func TestBaseCommentExecuteSuccess(t *testing.T) {
+	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+	createStub := &httpmock.Stub{
+		Method: "POST", URL: "/open-apis/drive/v1/files/baseToken/new_comments",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "success",
+			"data": map[string]interface{}{
+				"comment_id": "baseComment123",
+				"reply_id":   "baseReply123",
+				"created_at": 1700000000,
+			},
+		},
+	}
+	reg.Register(createStub)
+	err := mountAndRunDrive(t, DriveAddComment, []string{
+		"+add-comment",
+		"--doc", "https://example.larksuite.com/base/baseToken",
+		"--content", `[{"type":"text","text":"请看这条记录"}]`,
+		"--block-id", "tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R",
+		"--as", "user",
+	}, f, stdout)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	var requestBody map[string]interface{}
+	if err := json.Unmarshal(createStub.CapturedBody, &requestBody); err != nil {
+		t.Fatalf("failed to decode captured body: %v\nbody:\n%s", err, string(createStub.CapturedBody))
+	}
+	if got := mustStringField(t, requestBody, "file_type", "request.file_type"); got != "bitable" {
+		t.Fatalf("request file_type = %q, want bitable", got)
+	}
+	anchor := mustMapValue(t, requestBody["anchor"], "request.anchor")
+	if got := mustStringField(t, anchor, "block_id", "request.anchor.block_id"); got != "tbl9mp6fj9kDKHQV" {
+		t.Fatalf("request block_id = %q, want tbl9mp6fj9kDKHQV", got)
+	}
+	if got := mustStringField(t, anchor, "base_record_id", "request.anchor.base_record_id"); got != "recBIBgGmb" {
+		t.Fatalf("request base_record_id = %q, want recBIBgGmb", got)
+	}
+	if got := mustStringField(t, anchor, "base_view_id", "request.anchor.base_view_id"); got != "vewc46MG1R" {
+		t.Fatalf("request base_view_id = %q, want vewc46MG1R", got)
+	}
+
+	out := decodeJSONMap(t, stdout.String())
+	data := mustMapValue(t, out["data"], "data")
+	if got := mustStringField(t, data, "file_type", "data.file_type"); got != "bitable" {
+		t.Fatalf("stdout file_type = %q, want bitable\nstdout:\n%s", got, stdout.String())
+	}
+	if got := mustStringField(t, data, "comment_mode", "data.comment_mode"); got != "base_record" {
+		t.Fatalf("stdout comment_mode = %q, want base_record\nstdout:\n%s", got, stdout.String())
+	}
+	if got := mustStringField(t, data, "reply_id", "data.reply_id"); got != "baseReply123" {
+		t.Fatalf("stdout reply_id = %q, want baseReply123\nstdout:\n%s", got, stdout.String())
+	}
+}
+
+func TestBaseCommentExecuteBareToken(t *testing.T) {
+	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+	reg.Register(&httpmock.Stub{
+		Method: "POST", URL: "/open-apis/drive/v1/files/baseBareToken/new_comments",
+		Body: map[string]interface{}{
+			"code": 0, "msg": "success",
+			"data": map[string]interface{}{"comment_id": "baseBareComment"},
+		},
+	})
+	err := mountAndRunDrive(t, DriveAddComment, []string{
+		"+add-comment",
+		"--doc", "baseBareToken",
+		"--type", "bitable",
+		"--content", `[{"type":"text","text":"ok"}]`,
+		"--block-id", "tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R",
+		"--as", "user",
+	}, f, stdout)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !strings.Contains(stdout.String(), "baseBareComment") {
+		t.Fatalf("stdout missing comment_id: %s", stdout.String())
+	}
+}
+
 func TestFileCommentExecuteSuccess(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
 	reg.Register(&httpmock.Stub{
@@ -1433,6 +1641,40 @@ func TestDryRunSlidesDirectURL(t *testing.T) {
 	}
 }
 
+func TestDryRunBaseDirectURL(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, driveTestConfig())
+	err := mountAndRunDrive(t, DriveAddComment, []string{
+		"+add-comment",
+		"--doc", "https://example.larksuite.com/base/baseToken",
+		"--content", `[{"type":"text","text":"test"}]`,
+		"--block-id", "tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R",
+		"--dry-run", "--as", "user",
+	}, f, stdout)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !strings.Contains(stdout.String(), "record-local comment") {
+		t.Fatalf("dry-run output missing record-local comment: %s", stdout.String())
+	}
+	out := decodeJSONMap(t, stdout.String())
+	api := mustSliceValue(t, out["api"], "api")
+	call := mustMapValue(t, api[0], "api[0]")
+	body := mustMapValue(t, call["body"], "api[0].body")
+	anchor := mustMapValue(t, body["anchor"], "api[0].body.anchor")
+	if got := mustStringField(t, body, "file_type", "api[0].body.file_type"); got != "bitable" {
+		t.Fatalf("dry-run body.file_type = %q, want bitable\nstdout:\n%s", got, stdout.String())
+	}
+	if got := mustStringField(t, anchor, "block_id", "api[0].body.anchor.block_id"); got != "tbl9mp6fj9kDKHQV" {
+		t.Fatalf("dry-run body.anchor.block_id = %q, want tbl9mp6fj9kDKHQV\nstdout:\n%s", got, stdout.String())
+	}
+	if got := mustStringField(t, anchor, "base_record_id", "api[0].body.anchor.base_record_id"); got != "recBIBgGmb" {
+		t.Fatalf("dry-run body.anchor.base_record_id = %q, want recBIBgGmb\nstdout:\n%s", got, stdout.String())
+	}
+	if got := mustStringField(t, anchor, "base_view_id", "api[0].body.anchor.base_view_id"); got != "vewc46MG1R" {
+		t.Fatalf("dry-run body.anchor.base_view_id = %q, want vewc46MG1R\nstdout:\n%s", got, stdout.String())
+	}
+}
+
 func TestDryRunWikiResolvesToSlides(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
 	reg.Register(&httpmock.Stub{
@@ -1636,25 +1878,92 @@ func TestResolveWikiToDocxFullComment(t *testing.T) {
 	}
 }
 
-func TestResolveWikiToUnsupportedType(t *testing.T) {
-	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
-	reg.Register(&httpmock.Stub{
-		Method: "GET", URL: "/open-apis/wiki/v2/spaces/get_node",
-		Body: map[string]interface{}{
-			"code": 0, "msg": "success",
-			"data": map[string]interface{}{
-				"node": map[string]interface{}{"obj_type": "bitable", "obj_token": "bitToken"},
-			},
+func TestResolveWikiToBaseComment(t *testing.T) {
+	for _, objType := range []string{"bitable", "base"} {
+		t.Run(objType, func(t *testing.T) {
+			f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+			reg.Register(&httpmock.Stub{
+				Method: "GET", URL: "/open-apis/wiki/v2/spaces/get_node",
+				Body: map[string]interface{}{
+					"code": 0, "msg": "success",
+					"data": map[string]interface{}{
+						"node": map[string]interface{}{"obj_type": objType, "obj_token": "bitToken"},
+					},
+				},
+			})
+			reg.Register(&httpmock.Stub{
+				Method: "POST", URL: "/open-apis/drive/v1/files/bitToken/new_comments",
+				Body: map[string]interface{}{
+					"code": 0, "msg": "success",
+					"data": map[string]interface{}{"comment_id": "wikiBaseComment", "reply_id": "wikiBaseReply"},
+				},
+			})
+			err := mountAndRunDrive(t, DriveAddComment, []string{
+				"+add-comment",
+				"--doc", "https://example.larksuite.com/wiki/wikiToken",
+				"--content", `[{"type":"text","text":"test"}]`,
+				"--block-id", "tbl9mp6fj9kDKHQV!recBIBgGmb!vewc46MG1R",
+				"--as", "user",
+			}, f, stdout)
+			if err != nil {
+				t.Fatalf("unexpected error: %v", err)
+			}
+			if !strings.Contains(stdout.String(), "wikiBaseComment") {
+				t.Fatalf("stdout missing comment_id: %s", stdout.String())
+			}
+			out := decodeJSONMap(t, stdout.String())
+			data := mustMapValue(t, out["data"], "data")
+			if got := mustStringField(t, data, "file_type", "data.file_type"); got != "bitable" {
+				t.Fatalf("stdout file_type = %q, want bitable\nstdout:\n%s", got, stdout.String())
+			}
+			if got := mustStringField(t, data, "wiki_token", "data.wiki_token"); got != "wikiToken" {
+				t.Fatalf("stdout wiki_token = %q, want wikiToken\nstdout:\n%s", got, stdout.String())
+			}
+		})
+	}
+}
+
+func TestResolveWikiToBaseRejectsIncompatibleFlags(t *testing.T) {
+	cases := []struct {
+		name    string
+		args    []string
+		wantErr string
+	}{
+		{
+			name:    "full comment",
+			args:    []string{"--full-comment"},
+			wantErr: "--full-comment is not applicable for base(bitable) comments",
 		},
-	})
-	err := mountAndRunDrive(t, DriveAddComment, []string{
-		"+add-comment",
-		"--doc", "https://example.larksuite.com/wiki/wikiToken",
-		"--content", `[{"type":"text","text":"test"}]`,
-		"--as", "user",
-	}, f, stdout)
-	if err == nil || !strings.Contains(err.Error(), "only support doc/docx/file/sheet/slides") {
-		t.Fatalf("expected unsupported type error, got: %v", err)
+		{
+			name:    "selection",
+			args:    []string{"--selection-with-ellipsis", "some text"},
+			wantErr: "--selection-with-ellipsis is not applicable for base(bitable) comments",
+		},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+			reg.Register(&httpmock.Stub{
+				Method: "GET", URL: "/open-apis/wiki/v2/spaces/get_node",
+				Body: map[string]interface{}{
+					"code": 0, "msg": "success",
+					"data": map[string]interface{}{
+						"node": map[string]interface{}{"obj_type": "bitable", "obj_token": "bitToken"},
+					},
+				},
+			})
+			args := []string{
+				"+add-comment",
+				"--doc", "https://example.larksuite.com/wiki/wikiToken",
+				"--content", `[{"type":"text","text":"test"}]`,
+				"--as", "user",
+			}
+			args = append(args, tc.args...)
+			err := mountAndRunDrive(t, DriveAddComment, args, f, stdout)
+			if err == nil || !strings.Contains(err.Error(), tc.wantErr) {
+				t.Fatalf("expected %q error, got: %v", tc.wantErr, err)
+			}
+		})
 	}
 }
 
@@ -1735,7 +2044,7 @@ func TestDocOldFormatLocalCommentRejected(t *testing.T) {
 		"--block-id", "blk_123",
 		"--as", "user",
 	}, f, stdout)
-	if err == nil || !strings.Contains(err.Error(), "only support docx, sheet, and slides") {
+	if err == nil || !strings.Contains(err.Error(), "only support docx, sheet, slides, and base(bitable)") {
 		t.Fatalf("expected local comment rejection for old doc, got: %v", err)
 	}
 }

shortcuts/drive/drive_import.go

@@ -28,7 +28,7 @@ var DriveImport = common.Shortcut{
 	ConditionalScopes: []string{"wiki:node:retrieve"},
 	AuthTypes:         []string{"user", "bot"},
 	Flags: []common.Flag{
-		{Name: "file", Desc: "local file path (e.g. .docx, .xlsx, .md, .base, .pptx, .pdf; large files auto use multipart upload; .base is capped at 20MB, .pptx/.pdf at 500MB)", Required: true},
+		{Name: "file", Desc: "local file path (e.g. .docx, .xlsx, .md, .base, .pptx; large files auto use multipart upload; .base is capped at 20MB, .pptx at 500MB)", Required: true},
 		{Name: "type", Desc: "target document type (docx, sheet, bitable, slides)", Required: true},
 		{Name: "folder-token", Desc: "target folder token (omit for root folder; API accepts empty mount_key as root)"},
 		{Name: "name", Desc: "imported file name (default: local file name without extension)"},

shortcuts/drive/drive_import_common.go

@@ -45,7 +45,6 @@ var driveImportExtToDocTypes = map[string][]string{
 	"csv":      {"sheet", "bitable"},
 	"base":     {"bitable"},
 	"pptx":     {"slides"},
-	"pdf":      {"slides"},
 }
 
 // driveImportSpec contains the user-facing import inputs after normalization.
@@ -154,7 +153,7 @@ func driveImportFileSizeLimit(filePath, docType string) (int64, bool) {
 	switch strings.TrimPrefix(strings.ToLower(filepath.Ext(filePath)), ".") {
 	case "docx", "doc":
 		return driveImport600MBFileSizeLimit, true
-	case "pptx", "pdf":
+	case "pptx":
 		return driveImport500MBFileSizeLimit, true
 	case "txt", "md", "mark", "markdown", "html", "xls", "base":
 		return driveImport20MBFileSizeLimit, true
@@ -200,7 +199,7 @@ func validateDriveImportFileSize(filePath, docType string, fileSize int64) error
 func validateDriveImportSpec(spec driveImportSpec) error {
 	ext := spec.FileExtension()
 	if ext == "" {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "file must have an extension (e.g. .md, .docx, .xlsx, .pptx, .pdf)").WithParam("--file")
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "file must have an extension (e.g. .md, .docx, .xlsx, .pptx)").WithParam("--file")
 	}
 
 	switch spec.DocType {
@@ -211,7 +210,7 @@ func validateDriveImportSpec(spec driveImportSpec) error {
 
 	supportedTypes, ok := driveImportExtToDocTypes[ext]
 	if !ok {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "unsupported file extension: %s. Supported extensions are: docx, doc, txt, md, mark, markdown, html, xlsx, xls, csv, base, pptx, pdf", ext).WithParam("--file")
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "unsupported file extension: %s. Supported extensions are: docx, doc, txt, md, mark, markdown, html, xlsx, xls, csv, base, pptx", ext).WithParam("--file")
 	}
 
 	typeAllowed := false
@@ -232,8 +231,8 @@ func validateDriveImportSpec(spec driveImportSpec) error {
 			hint = fmt.Sprintf(".xls files can only be imported as 'sheet', not '%s'", spec.DocType)
 		case "base":
 			hint = fmt.Sprintf(".base files can only be imported as 'bitable', not '%s'", spec.DocType)
-		case "pptx", "pdf":
-			hint = fmt.Sprintf(".%s files can only be imported as 'slides', not '%s'", ext, spec.DocType)
+		case "pptx":
+			hint = fmt.Sprintf(".pptx files can only be imported as 'slides', not '%s'", spec.DocType)
 		default:
 			hint = fmt.Sprintf(".%s files can only be imported as 'docx', not '%s'", ext, spec.DocType)
 		}

shortcuts/drive/drive_import_common_test.go

@@ -41,10 +41,6 @@ func TestValidateDriveImportSpec(t *testing.T) {
 			name: "pptx slides ok",
 			spec: driveImportSpec{FilePath: "./deck.pptx", DocType: "slides"},
 		},
-		{
-			name: "pdf slides ok",
-			spec: driveImportSpec{FilePath: "./deck.pdf", DocType: "slides"},
-		},
 		{
 			name:    "base non bitable rejected",
 			spec:    driveImportSpec{FilePath: "./snapshot.base", DocType: "sheet"},
@@ -55,11 +51,6 @@ func TestValidateDriveImportSpec(t *testing.T) {
 			spec:    driveImportSpec{FilePath: "./deck.pptx", DocType: "docx"},
 			wantErr: ".pptx files can only be imported as 'slides'",
 		},
-		{
-			name:    "pdf non slides rejected",
-			spec:    driveImportSpec{FilePath: "./deck.pdf", DocType: "docx"},
-			wantErr: ".pdf files can only be imported as 'slides'",
-		},
 		{
 			name:    "unknown extension rejected",
 			spec:    driveImportSpec{FilePath: "./data.rtf", DocType: "docx"},
@@ -147,19 +138,6 @@ func TestValidateDriveImportFileSize(t *testing.T) {
 			docType:  "slides",
 			fileSize: driveImport500MBFileSizeLimit,
 		},
-		{
-			name:     "pdf exceeds 500mb limit",
-			filePath: "./deck.pdf",
-			docType:  "slides",
-			fileSize: driveImport500MBFileSizeLimit + 1,
-			wantText: "exceeds 500.0 MB import limit for .pdf",
-		},
-		{
-			name:     "pdf within 500mb limit",
-			filePath: "./deck.pdf",
-			docType:  "slides",
-			fileSize: driveImport500MBFileSizeLimit,
-		},
 		{
 			name:     "base exceeds 20mb limit",
 			filePath: "./snapshot.base",

shortcuts/drive/drive_search.go

@@ -149,6 +149,9 @@ var DriveSearch = common.Shortcut{
 			"page_token": data["page_token"],
 			"results":    normalizedItems,
 		}
+		if notice, _ := data["notice"].(string); notice != "" {
+			resultData["notice"] = notice
+		}
 
 		runtime.OutFormat(resultData, &output.Meta{Count: len(normalizedItems)}, func(w io.Writer) {
 			renderDriveSearchTable(w, data, normalizedItems)

shortcuts/drive/drive_search_test.go

@@ -14,12 +14,49 @@ import (
 	"time"
 
 	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/errclass"
+	"github.com/larksuite/cli/internal/httpmock"
 	"github.com/larksuite/cli/internal/output"
 )
 
-// TestClampOpenedTimeWindow covers the 3-month / 1-year boundary logic that
-// narrows --opened-since / --opened-until and generates the multi-slice notice.
+// TestDriveSearchExecutePassesThroughNotice verifies drive +search preserves notices.
+func TestDriveSearchExecutePassesThroughNotice(t *testing.T) {
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/search/v2/doc_wiki/search",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "ok",
+			"data": map[string]interface{}{
+				"notice":     notice,
+				"res_units":  []interface{}{},
+				"total":      0,
+				"has_more":   false,
+				"page_token": "",
+			},
+		},
+	})
+
+	if err := mountAndRunDrive(t, DriveSearch, []string{"+search", "--query", "incident", "--format", "json", "--as", "user"}, f, stdout); err != nil {
+		t.Fatalf("DriveSearch.Execute() error = %v", err)
+	}
+	reg.Verify(t)
+
+	var env map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &env); err != nil {
+		t.Fatalf("json.Unmarshal(stdout) error = %v\nstdout=%s", err, stdout.String())
+	}
+	data, _ := env["data"].(map[string]interface{})
+	if got, _ := data["notice"].(string); got != notice {
+		t.Fatalf("data.notice = %q, want %q; data=%#v", got, notice, data)
+	}
+}
+
+// TestClampOpenedTimeWindow covers opened-time clamping and slice notices.
 func TestClampOpenedTimeWindow(t *testing.T) {
 	t.Parallel()
 

shortcuts/im/builders_test.go

@@ -26,9 +26,7 @@ func mustMarshalDryRun(t *testing.T, v interface{}) string {
 	return string(b)
 }
 
-// newTestRuntimeContext builds a *common.RuntimeContext backed by a cobra
-// command whose flags are populated from the provided string and bool maps,
-// for unit-testing shortcut bodies, validators, and dry-run shapes.
+// newTestRuntimeContext builds a RuntimeContext with string and bool test flags.
 func newTestRuntimeContext(t *testing.T, stringFlags map[string]string, boolFlags map[string]bool) *common.RuntimeContext {
 	t.Helper()
 
@@ -59,9 +57,38 @@ func newTestRuntimeContext(t *testing.T, stringFlags map[string]string, boolFlag
 	return &common.RuntimeContext{Cmd: cmd}
 }
 
-// newMessagesSearchTestRuntimeContext is the messages-search variant of
-// newTestRuntimeContext: registers the search-specific --page-size flag
-// before applying caller-provided values.
+// newChatSearchTestRuntimeContext builds a chat-search RuntimeContext with typed flags.
+func newChatSearchTestRuntimeContext(t *testing.T, stringFlags map[string]string, boolFlags map[string]bool) *common.RuntimeContext {
+	t.Helper()
+
+	cmd := &cobra.Command{Use: "test"}
+	cmd.Flags().Int("page-size", 20, "")
+	for name := range stringFlags {
+		if name == "page-size" {
+			continue
+		}
+		cmd.Flags().String(name, "", "")
+	}
+	for name := range boolFlags {
+		cmd.Flags().Bool(name, false, "")
+	}
+	if err := cmd.ParseFlags(nil); err != nil {
+		t.Fatalf("ParseFlags() error = %v", err)
+	}
+	for name, val := range stringFlags {
+		if err := cmd.Flags().Set(name, val); err != nil {
+			t.Fatalf("Flags().Set(%q) error = %v", name, err)
+		}
+	}
+	for name, val := range boolFlags {
+		if err := cmd.Flags().Set(name, map[bool]string{true: "true", false: "false"}[val]); err != nil {
+			t.Fatalf("Flags().Set(%q) error = %v", name, err)
+		}
+	}
+	return &common.RuntimeContext{Cmd: cmd}
+}
+
+// newMessagesSearchTestRuntimeContext builds a messages-search RuntimeContext.
 func newMessagesSearchTestRuntimeContext(t *testing.T, stringFlags map[string]string, boolFlags map[string]bool) *common.RuntimeContext {
 	t.Helper()
 
@@ -231,6 +258,7 @@ func TestIsMediaKey(t *testing.T) {
 	}
 }
 
+// TestShortcutValidateBranches covers direct shortcut validation branches.
 func TestShortcutValidateBranches(t *testing.T) {
 
 	t.Run("ImChatCreate valid", func(t *testing.T) {
@@ -297,7 +325,7 @@ func TestShortcutValidateBranches(t *testing.T) {
 	})
 
 	t.Run("ImChatSearch invalid page size", func(t *testing.T) {
-		runtime := newTestRuntimeContext(t, map[string]string{
+		runtime := newChatSearchTestRuntimeContext(t, map[string]string{
 			"query":     "ok",
 			"page-size": "0",
 		}, nil)
@@ -307,12 +335,13 @@ func TestShortcutValidateBranches(t *testing.T) {
 		}
 	})
 
-	t.Run("ImChatSearch query too long", func(t *testing.T) {
-		runtime := newTestRuntimeContext(t, map[string]string{
-			"query": strings.Repeat("q", 65),
+	t.Run("ImChatSearch allows long query for server-side notice", func(t *testing.T) {
+		runtime := newChatSearchTestRuntimeContext(t, map[string]string{
+			"query":     strings.Repeat("q", 81),
+			"page-size": "20",
 		}, nil)
 		err := ImChatSearch.Validate(context.Background(), runtime)
-		if err == nil || !strings.Contains(err.Error(), "--query exceeds the maximum of 64 characters") {
+		if err != nil {
 			t.Fatalf("ImChatSearch.Validate() error = %v", err)
 		}
 	})
@@ -607,6 +636,7 @@ func TestShortcutValidateBranches(t *testing.T) {
 	})
 }
 
+// TestMessagesSearchPaginationConfig verifies page-all and page-limit behavior.
 func TestMessagesSearchPaginationConfig(t *testing.T) {
 	t.Run("default single page", func(t *testing.T) {
 		runtime := newMessagesSearchTestRuntimeContext(t, nil, nil)
@@ -650,8 +680,7 @@ func TestMessagesSearchPaginationConfig(t *testing.T) {
 	})
 }
 
-// TestShortcutDryRunShapes verifies that each shortcut's DryRun function
-// produces the expected API path, query parameters, and request body.
+// TestShortcutDryRunShapes verifies shortcut dry-run API paths and payloads.
 func TestShortcutDryRunShapes(t *testing.T) {
 	t.Run("ImChatCreate dry run includes params and body", func(t *testing.T) {
 		cmd := &cobra.Command{Use: "test"}
@@ -674,19 +703,19 @@ func TestShortcutDryRunShapes(t *testing.T) {
 	})
 
 	t.Run("ImChatSearch dry run includes built params", func(t *testing.T) {
-		runtime := newTestRuntimeContext(t, map[string]string{
+		runtime := newChatSearchTestRuntimeContext(t, map[string]string{
 			"query":      "team-alpha",
 			"page-size":  "50",
 			"page-token": "next_page",
 		}, nil)
 		got := mustMarshalDryRun(t, ImChatSearch.DryRun(context.Background(), runtime))
-		if !strings.Contains(got, `"/open-apis/im/v2/chats/search"`) || !strings.Contains(got, `"page_size":20`) || !strings.Contains(got, `"query":"\"team-alpha\""`) {
+		if !strings.Contains(got, `"/open-apis/im/v2/chats/search"`) || !strings.Contains(got, `"page_size":50`) || !strings.Contains(got, `"query":"\"team-alpha\""`) {
 			t.Fatalf("ImChatSearch.DryRun() = %s", got)
 		}
 	})
 
 	t.Run("ImChatSearch dry run still works with --exclude-muted set", func(t *testing.T) {
-		runtime := newTestRuntimeContext(t, map[string]string{
+		runtime := newChatSearchTestRuntimeContext(t, map[string]string{
 			"query": "team-alpha",
 		}, map[string]bool{
 			"exclude-muted": true,

shortcuts/im/im_chat_search.go

@@ -29,7 +29,7 @@ var ImChatSearch = common.Shortcut{
 	AuthTypes:   []string{"user", "bot"},
 	HasFormat:   true,
 	Flags: []common.Flag{
-		{Name: "query", Desc: "search keyword (max 64 chars)"},
+		{Name: "query", Desc: "search keyword (server may return data.notice for overly long input)"},
 		{Name: "search-types", Desc: "chat types, comma-separated (private, external, public_joined, public_not_joined)"},
 		{Name: "chat-modes", Desc: "filter by chat mode, comma-separated (group, topic)"},
 		{Name: "member-ids", Desc: "filter by member open_ids, comma-separated"},
@@ -50,7 +50,7 @@ var ImChatSearch = common.Shortcut{
 			Params(params).
 			Body(body)
 	},
-	// Validate enforces query/member-ids presence, --query rune cap, search-types
+	// Validate enforces query/member-ids presence, search-types
 	// enum, --member-ids count and format, and --page-size bounds.
 	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
 		query := runtime.Str("query")
@@ -58,9 +58,6 @@ var ImChatSearch = common.Shortcut{
 		if query == "" && memberIDs == "" {
 			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--query and --member-ids cannot both be empty; provide at least one (e.g. --query \"team-name\" or --member-ids \"ou_xxx\")")
 		}
-		if query != "" && len([]rune(query)) > 64 {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--query exceeds the maximum of 64 characters (got %d)", len([]rune(query))).WithParam("--query")
-		}
 		if st := runtime.Str("search-types"); st != "" {
 			allowed := map[string]struct{}{
 				"private":           {},
@@ -151,6 +148,9 @@ var ImChatSearch = common.Shortcut{
 			"has_more":   hasMore,
 			"page_token": pageToken,
 		}
+		if notice, _ := resData["notice"].(string); notice != "" {
+			outData["notice"] = notice
+		}
 		if mfOut.Meta.Applied != "" {
 			outData["filter"] = MuteFilterMetaToMap(mfOut.Meta)
 		}

shortcuts/im/im_messages_search.go

@@ -91,7 +91,7 @@ var ImMessagesSearch = common.Shortcut{
 			return err
 		}
 
-		rawItems, hasMore, nextPageToken, truncatedByLimit, pageLimit, err := searchMessages(runtime, req)
+		rawItems, hasMore, nextPageToken, truncatedByLimit, pageLimit, notice, err := searchMessages(runtime, req)
 		if err != nil {
 			return err
 		}
@@ -103,6 +103,9 @@ var ImMessagesSearch = common.Shortcut{
 				"has_more":   hasMore,
 				"page_token": nextPageToken,
 			}
+			if notice != "" {
+				outData["notice"] = notice
+			}
 			runtime.OutFormat(outData, nil, func(w io.Writer) {
 				fmt.Fprintln(w, "No matching messages found.")
 			})
@@ -131,6 +134,9 @@ var ImMessagesSearch = common.Shortcut{
 				"page_token":  nextPageToken,
 				"note":        "failed to fetch message details, returning ID list only",
 			}
+			if notice != "" {
+				outData["notice"] = notice
+			}
 			runtime.OutFormat(outData, nil, func(w io.Writer) {
 				fmt.Fprintf(w, "Found %d messages (failed to fetch details):\n", len(messageIds))
 				for _, id := range messageIds {
@@ -206,6 +212,9 @@ var ImMessagesSearch = common.Shortcut{
 			"has_more":   hasMore,
 			"page_token": nextPageToken,
 		}
+		if notice != "" {
+			outData["notice"] = notice
+		}
 		runtime.OutFormat(outData, nil, func(w io.Writer) {
 			if len(enriched) == 0 {
 				fmt.Fprintln(w, "No matching messages found.")
@@ -377,6 +386,7 @@ func buildMessagesSearchRequest(runtime *common.RuntimeContext) (*messagesSearch
 	}, nil
 }
 
+// messagesSearchPaginationConfig derives auto-pagination mode and page limit.
 func messagesSearchPaginationConfig(runtime *common.RuntimeContext) (autoPaginate bool, pageLimit int) {
 	autoPaginate = runtime.Bool("page-all")
 	if runtime.Cmd != nil && runtime.Cmd.Flags().Changed("page-limit") {
@@ -392,7 +402,8 @@ func messagesSearchPaginationConfig(runtime *common.RuntimeContext) (autoPaginat
 	return autoPaginate, pageLimit
 }
 
-func searchMessages(runtime *common.RuntimeContext, req *messagesSearchRequest) ([]interface{}, bool, string, bool, int, error) {
+// searchMessages fetches message search pages and returns the first server notice.
+func searchMessages(runtime *common.RuntimeContext, req *messagesSearchRequest) ([]interface{}, bool, string, bool, int, string, error) {
 	autoPaginate, pageLimit := messagesSearchPaginationConfig(runtime)
 	pageToken := ""
 	if tokens := req.params["page_token"]; len(tokens) > 0 {
@@ -410,6 +421,7 @@ func searchMessages(runtime *common.RuntimeContext, req *messagesSearchRequest)
 		lastPageToken    string
 		truncatedByLimit bool
 		pageCount        int
+		notice           string
 	)
 
 	for {
@@ -423,9 +435,12 @@ func searchMessages(runtime *common.RuntimeContext, req *messagesSearchRequest)
 
 		searchData, err := runtime.DoAPIJSONTyped(http.MethodPost, "/open-apis/im/v1/messages/search", params, req.body)
 		if err != nil {
-			return nil, false, "", false, pageLimit, err
+			return nil, false, "", false, pageLimit, "", err
 		}
 
+		if notice == "" {
+			notice, _ = searchData["notice"].(string)
+		}
 		items, _ := searchData["items"].([]interface{})
 		allItems = append(allItems, items...)
 		lastHasMore, lastPageToken = common.PaginationMeta(searchData)
@@ -441,9 +456,10 @@ func searchMessages(runtime *common.RuntimeContext, req *messagesSearchRequest)
 		pageToken = lastPageToken
 	}
 
-	return allItems, lastHasMore, lastPageToken, truncatedByLimit, pageLimit, nil
+	return allItems, lastHasMore, lastPageToken, truncatedByLimit, pageLimit, notice, nil
 }
 
+// batchMGetMessages fetches message details in API-sized batches.
 func batchMGetMessages(runtime *common.RuntimeContext, messageIds []string) ([]interface{}, error) {
 	var items []interface{}
 	for _, batch := range chunkStrings(messageIds, messagesSearchMGetBatchSize) {
@@ -457,6 +473,7 @@ func batchMGetMessages(runtime *common.RuntimeContext, messageIds []string) ([]i
 	return items, nil
 }
 
+// batchQueryChatContexts fetches chat metadata best-effort for message rows.
 func batchQueryChatContexts(runtime *common.RuntimeContext, chatIds []string) map[string]map[string]interface{} {
 	chatContexts := map[string]map[string]interface{}{}
 	// Best-effort: a failed chunk only loses its own entries.
@@ -466,6 +483,7 @@ func batchQueryChatContexts(runtime *common.RuntimeContext, chatIds []string) ma
 	return chatContexts
 }
 
+// chunkStrings splits a string slice into fixed-size batches.
 func chunkStrings(items []string, chunkSize int) [][]string {
 	if len(items) == 0 || chunkSize <= 0 {
 		return nil

shortcuts/im/im_search_notice_test.go

@@ -0,0 +1,129 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package im
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/shortcuts/common"
+	"github.com/spf13/cobra"
+)
+
+// TestImChatSearchExecutePassesThroughNotice verifies chat search notice output.
+func TestImChatSearchExecutePassesThroughNotice(t *testing.T) {
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+	longQuery := strings.Repeat("q", 81)
+
+	runtime := newBotShortcutRuntime(t, shortcutRoundTripFunc(func(req *http.Request) (*http.Response, error) {
+		if !strings.Contains(req.URL.Path, "/open-apis/im/v2/chats/search") {
+			return nil, fmt.Errorf("unexpected request: %s", req.URL.String())
+		}
+		var body map[string]interface{}
+		if err := json.NewDecoder(req.Body).Decode(&body); err != nil {
+			return nil, fmt.Errorf("decode request body: %w", err)
+		}
+		if got, _ := body["query"].(string); got != longQuery {
+			return nil, fmt.Errorf("body.query = %q, want %q", got, longQuery)
+		}
+		return shortcutJSONResponse(200, map[string]interface{}{
+			"code": 0,
+			"data": map[string]interface{}{
+				"notice":     notice,
+				"items":      []interface{}{},
+				"total":      0,
+				"has_more":   false,
+				"page_token": "",
+			},
+		}), nil
+	}))
+	runtime.Cmd = newChatSearchNoticeTestCommand(t, longQuery)
+	runtime.Format = "json"
+
+	if err := ImChatSearch.Execute(context.Background(), runtime); err != nil {
+		t.Fatalf("ImChatSearch.Execute() error = %v", err)
+	}
+
+	data := decodeShortcutData(t, runtime)
+	if got, _ := data["notice"].(string); got != notice {
+		t.Fatalf("data.notice = %q, want %q; data=%#v", got, notice, data)
+	}
+}
+
+// TestImMessagesSearchExecutePassesThroughNotice verifies message search notice output.
+func TestImMessagesSearchExecutePassesThroughNotice(t *testing.T) {
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+
+	runtime := newMessagesSearchRuntime(t, map[string]string{
+		"query": "incident",
+	}, nil, shortcutRoundTripFunc(func(req *http.Request) (*http.Response, error) {
+		if !strings.Contains(req.URL.Path, "/open-apis/im/v1/messages/search") {
+			return nil, fmt.Errorf("unexpected request: %s", req.URL.String())
+		}
+		return shortcutJSONResponse(200, map[string]interface{}{
+			"code": 0,
+			"data": map[string]interface{}{
+				"notice":     notice,
+				"items":      []interface{}{},
+				"has_more":   false,
+				"page_token": "",
+			},
+		}), nil
+	}))
+	runtime.Format = "json"
+
+	if err := ImMessagesSearch.Execute(context.Background(), runtime); err != nil {
+		t.Fatalf("ImMessagesSearch.Execute() error = %v", err)
+	}
+
+	data := decodeShortcutData(t, runtime)
+	if got, _ := data["notice"].(string); got != notice {
+		t.Fatalf("data.notice = %q, want %q; data=%#v", got, notice, data)
+	}
+}
+
+// newChatSearchNoticeTestCommand builds a typed chat-search command for notice tests.
+func newChatSearchNoticeTestCommand(t *testing.T, query string) *cobra.Command {
+	t.Helper()
+
+	cmd := &cobra.Command{Use: "test"}
+	for _, name := range []string{"query", "search-types", "member-ids", "sort-by", "page-token"} {
+		cmd.Flags().String(name, "", "")
+	}
+	for _, name := range []string{"is-manager", "disable-search-by-user", "exclude-muted"} {
+		cmd.Flags().Bool(name, false, "")
+	}
+	cmd.Flags().Int("page-size", 20, "")
+	if err := cmd.ParseFlags(nil); err != nil {
+		t.Fatalf("ParseFlags() error = %v", err)
+	}
+	if err := cmd.Flags().Set("query", query); err != nil {
+		t.Fatalf("Flags().Set(query) error = %v", err)
+	}
+	return cmd
+}
+
+// decodeShortcutData extracts the JSON envelope data object from shortcut output.
+func decodeShortcutData(t *testing.T, runtime *common.RuntimeContext) map[string]interface{} {
+	t.Helper()
+
+	out, ok := runtime.Factory.IOStreams.Out.(*bytes.Buffer)
+	if !ok {
+		t.Fatalf("stdout buffer has type %T", runtime.Factory.IOStreams.Out)
+	}
+	var env map[string]interface{}
+	if err := json.Unmarshal(out.Bytes(), &env); err != nil {
+		t.Fatalf("json.Unmarshal(stdout) error = %v\nstdout=%s", err, out.String())
+	}
+	data, ok := env["data"].(map[string]interface{})
+	if !ok {
+		t.Fatalf("envelope data missing or wrong type: %#v", env)
+	}
+	return data
+}

shortcuts/mail/mail_triage.go

@@ -159,6 +159,7 @@ var MailTriage = common.Shortcut{
 		var messages []map[string]interface{}
 		var hasMore bool
 		var nextPageToken string
+		var notice string
 
 		useSearch, err := resolveTriagePath(parsed, query, filter)
 		if err != nil {
@@ -189,6 +190,9 @@ var MailTriage = common.Shortcut{
 				if err != nil {
 					return err
 				}
+				if notice == "" {
+					notice, _ = searchData["notice"].(string)
+				}
 				pageMessages := buildTriageMessagesFromSearchItems(searchData["items"])
 				messages = append(messages, pageMessages...)
 				pageHasMore, _ := searchData["has_more"].(bool)
@@ -282,8 +286,14 @@ var MailTriage = common.Shortcut{
 				"has_more":   hasMore,
 				"page_token": nextPageToken,
 			}
+			if notice != "" {
+				outData["notice"] = notice
+			}
 			output.PrintJson(runtime.IO().Out, outData)
 		default: // "table"
+			if notice != "" {
+				fmt.Fprintf(runtime.IO().ErrOut, "notice: %s\n", notice)
+			}
 			if len(messages) == 0 {
 				fmt.Fprintln(runtime.IO().ErrOut, "No messages found.")
 				return nil

shortcuts/mail/mail_triage_test.go

@@ -1478,14 +1478,16 @@ func boolPtr(v bool) *bool { return &v }
 
 // --- mailbox_id preservation tests ---
 
+// TestMailTriageStructuredOutputPreservesMailboxID verifies mailbox and notice metadata.
 func TestMailTriageStructuredOutputPreservesMailboxID(t *testing.T) {
 	tests := []struct {
-		name      string
-		mailbox   string
-		format    string
-		args      []string
-		register  func(*httpmock.Registry, string)
-		wantCount int
+		name       string
+		mailbox    string
+		format     string
+		args       []string
+		register   func(*httpmock.Registry, string)
+		wantCount  int
+		wantNotice string
 	}{
 		{
 			name:    "list json default mailbox",
@@ -1522,9 +1524,10 @@ func TestMailTriageStructuredOutputPreservesMailboxID(t *testing.T) {
 			register: func(reg *httpmock.Registry, mailbox string) {
 				registerMailTriageSearchStub(reg, mailbox, []interface{}{
 					mailTriageSearchItem("search_pub_001", "Shared search"),
-				}, false, "")
+				}, false, "", "The query is too long and has been truncated to the first 50 characters for search.")
 			},
-			wantCount: 1,
+			wantCount:  1,
+			wantNotice: "The query is too long and has been truncated to the first 50 characters for search.",
 		},
 		{
 			name:    "empty list json keeps top-level mailbox",
@@ -1559,6 +1562,9 @@ func TestMailTriageStructuredOutputPreservesMailboxID(t *testing.T) {
 			if data["mailbox_id"] != tt.mailbox {
 				t.Fatalf("top-level mailbox_id mismatch: got %v, want %q", data["mailbox_id"], tt.mailbox)
 			}
+			if tt.wantNotice != "" && data["notice"] != tt.wantNotice {
+				t.Fatalf("notice mismatch: got %v, want %q", data["notice"], tt.wantNotice)
+			}
 			messages := mailTriageMessagesFromOutput(t, data)
 			if len(messages) != tt.wantCount {
 				t.Fatalf("message count mismatch: got %d, want %d", len(messages), tt.wantCount)
@@ -1572,6 +1578,7 @@ func TestMailTriageStructuredOutputPreservesMailboxID(t *testing.T) {
 	}
 }
 
+// TestMailTriageMissingMessageMetadataStillGetsMailboxID verifies fallback rows keep mailbox IDs.
 func TestMailTriageMissingMessageMetadataStillGetsMailboxID(t *testing.T) {
 	f, stdout, _, reg := mailShortcutTestFactory(t)
 	defer reg.Verify(t)
@@ -1604,6 +1611,7 @@ func TestMailTriageMissingMessageMetadataStillGetsMailboxID(t *testing.T) {
 	}
 }
 
+// TestMailTriageTableOutputPreservesMailboxContext verifies public mailbox table hints.
 func TestMailTriageTableOutputPreservesMailboxContext(t *testing.T) {
 	tests := []struct {
 		name              string
@@ -1654,6 +1662,33 @@ func TestMailTriageTableOutputPreservesMailboxContext(t *testing.T) {
 	}
 }
 
+// TestMailTriageDefaultTableOutputPrintsSearchNoticeToStderr verifies stderr notices.
+func TestMailTriageDefaultTableOutputPrintsSearchNoticeToStderr(t *testing.T) {
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+
+	f, stdout, stderr, reg := mailShortcutTestFactory(t)
+	defer reg.Verify(t)
+
+	registerMailTriageSearchStub(reg, "me", []interface{}{
+		mailTriageSearchItem("msg_search_notice", "Search notice result"),
+	}, false, "", notice)
+
+	if err := runMountedMailShortcut(t, MailTriage, []string{
+		"+triage",
+		"--query", strings.Repeat("q", 81),
+	}, f, stdout); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	if out := stdout.String(); !strings.Contains(out, "msg_search_notice") {
+		t.Fatalf("stdout should contain table row, got:\n%s", out)
+	}
+	if errOut := stderr.String(); !strings.Contains(errOut, "notice: "+notice) {
+		t.Fatalf("stderr should contain search notice, got:\n%s", errOut)
+	}
+}
+
+// decodeMailTriageJSONOutput decodes structured triage output for assertions.
 func decodeMailTriageJSONOutput(t *testing.T, stdout interface{ Bytes() []byte }) map[string]interface{} {
 	t.Helper()
 	var data map[string]interface{}
@@ -1663,6 +1698,7 @@ func decodeMailTriageJSONOutput(t *testing.T, stdout interface{ Bytes() []byte }
 	return data
 }
 
+// mailTriageMessagesFromOutput extracts triage messages as object maps.
 func mailTriageMessagesFromOutput(t *testing.T, data map[string]interface{}) []map[string]interface{} {
 	t.Helper()
 	rawMessages, ok := data["messages"].([]interface{})
@@ -1715,7 +1751,8 @@ func registerMailTriageBatchStub(reg *httpmock.Registry, mailbox string, message
 	})
 }
 
-func registerMailTriageSearchStub(reg *httpmock.Registry, mailbox string, items []interface{}, hasMore bool, pageToken string) {
+// registerMailTriageSearchStub registers a mailbox search response for triage tests.
+func registerMailTriageSearchStub(reg *httpmock.Registry, mailbox string, items []interface{}, hasMore bool, pageToken string, notices ...string) {
 	data := map[string]interface{}{
 		"items":    items,
 		"has_more": hasMore,
@@ -1723,6 +1760,9 @@ func registerMailTriageSearchStub(reg *httpmock.Registry, mailbox string, items
 	if pageToken != "" {
 		data["page_token"] = pageToken
 	}
+	if len(notices) > 0 && notices[0] != "" {
+		data["notice"] = notices[0]
+	}
 	reg.Register(&httpmock.Stub{
 		Method: "POST",
 		URL:    mailboxPath(mailbox, "search"),

shortcuts/minutes/minutes_search.go

@@ -308,6 +308,9 @@ var MinutesSearch = common.Shortcut{
 			"has_more":   data["has_more"],
 			"page_token": data["page_token"],
 		}
+		if notice, _ := data["notice"].(string); notice != "" {
+			outData["notice"] = notice
+		}
 
 		runtime.OutFormat(outData, &output.Meta{Count: len(rows)}, func(w io.Writer) {
 			if len(rows) == 0 {

shortcuts/minutes/minutes_search_test.go

@@ -609,6 +609,8 @@ func TestMinutesSearchExecuteShowsPaginationHintForTableFormat(t *testing.T) {
 func TestMinutesSearchExecuteJSONCountUsesRenderedRows(t *testing.T) {
 	t.Parallel()
 
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+
 	f, stdout, _, reg := cmdutil.TestFactory(t, defaultConfig())
 	reg.Register(&httpmock.Stub{
 		Method: "POST",
@@ -617,6 +619,7 @@ func TestMinutesSearchExecuteJSONCountUsesRenderedRows(t *testing.T) {
 			"code": 0,
 			"msg":  "ok",
 			"data": map[string]interface{}{
+				"notice": notice,
 				"items": []interface{}{
 					nil,
 					map[string]interface{}{
@@ -641,6 +644,9 @@ func TestMinutesSearchExecuteJSONCountUsesRenderedRows(t *testing.T) {
 	reg.Verify(t)
 
 	var envelope struct {
+		Data struct {
+			Notice string `json:"notice"`
+		} `json:"data"`
 		Meta struct {
 			Count int `json:"count"`
 		} `json:"meta"`
@@ -651,6 +657,9 @@ func TestMinutesSearchExecuteJSONCountUsesRenderedRows(t *testing.T) {
 	if envelope.Meta.Count != 1 {
 		t.Fatalf("meta.count = %d, want 1", envelope.Meta.Count)
 	}
+	if envelope.Data.Notice != notice {
+		t.Fatalf("data.notice = %q, want %q", envelope.Data.Notice, notice)
+	}
 }
 
 // TestMinuteSearchFieldExtractors verifies field extractors read populated metadata correctly.

shortcuts/slides/shortcuts.go

@@ -11,8 +11,6 @@ func Shortcuts() []common.Shortcut {
 		SlidesCreate,
 		SlidesMediaUpload,
 		SlidesReplaceSlide,
-		SlidesReplacePages,
 		SlidesScreenshot,
-		SlidesXMLGet,
 	}
 }

shortcuts/slides/slides_replace_pages.go

@@ -1,413 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package slides
-
-import (
-	"context"
-	"encoding/json"
-	"encoding/xml"
-	"fmt"
-	"io"
-	"strings"
-
-	"github.com/larksuite/cli/errs"
-	"github.com/larksuite/cli/internal/validate"
-	"github.com/larksuite/cli/shortcuts/common"
-)
-
-// SlidesReplacePages rebuilds multiple pages inside an existing presentation.
-// It deliberately creates the new page before deleting the old one so a create
-// failure cannot remove existing user content. The operation is not atomic.
-const replacePagesInitialRevisionID = -1
-
-var SlidesReplacePages = common.Shortcut{
-	Service:     "slides",
-	Command:     "+replace-pages",
-	Description: "Batch rebuild pages inside an existing Slides presentation (create before old page, then delete old page; not atomic)",
-	Risk:        "write",
-	Scopes:      []string{"slides:presentation:update", "slides:presentation:write_only", "wiki:node:read"},
-	AuthTypes:   []string{"user", "bot"},
-	Flags: []common.Flag{
-		{Name: "presentation", Desc: "xml_presentation_id, slides URL, or wiki URL that resolves to slides", Required: true},
-		{Name: "pages", Desc: "JSON array of page replacements (each: {slide_id, content}); supports @file or -", Required: true, Input: []string{common.File, common.Stdin}},
-		{Name: "continue-on-error", Type: "bool", Desc: "continue with later pages after a create/delete failure; default false"},
-		{Name: "validate-only", Type: "bool", Desc: "validate input and build the create/delete plan without write calls"},
-	},
-	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		if _, err := parsePresentationRef(runtime.Str("presentation")); err != nil {
-			return err
-		}
-		pages, err := parseReplacePages(runtime.Str("pages"))
-		if err != nil {
-			return err
-		}
-		return validateReplacePagesInput(pages)
-	},
-	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		dry := common.NewDryRunAPI()
-		resolved, err := prepareReplacePages(runtime)
-		if err != nil {
-			return dry.Set("error", err.Error())
-		}
-		appendReplacePagesDryRunCalls(dry, resolved)
-		return dry.
-			Set("xml_presentation_id", resolved.PresentationID).
-			Set("pages_count", len(resolved.Plan)).
-			Set("plan", replacePagesPlanOutput(resolved.Plan)).
-			Set("note", "dry-run built a create/delete plan from slide_id inputs; no Slides presentation get/create/delete calls were executed")
-	},
-	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		resolved, err := prepareReplacePages(runtime)
-		if err != nil {
-			return err
-		}
-		if runtime.Bool("validate-only") {
-			runtime.Out(map[string]interface{}{
-				"xml_presentation_id": resolved.PresentationID,
-				"pages_count":         len(resolved.Plan),
-				"plan":                replacePagesPlanOutput(resolved.Plan),
-				"status":              "validated",
-				"note":                "validate-only checked input and built the create/delete plan; no Slides presentation get/create/delete calls were executed",
-			}, nil)
-			return nil
-		}
-
-		revisionID := replacePagesInitialRevisionID
-		results := make([]replacePageResult, 0, len(resolved.Plan))
-		for i, item := range resolved.Plan {
-			result, err := replaceOnePage(runtime, resolved.PresentationID, item, revisionID)
-			results = append(results, result)
-			if result.RevisionID != nil {
-				revisionID = *result.RevisionID
-			}
-			if err != nil {
-				if runtime.Bool("continue-on-error") {
-					continue
-				}
-				return appendSlidesProgressHint(err, fmt.Sprintf("slides +replace-pages stopped at item %d/%d; %d page(s) completed before failure; old page is kept when create failed", i+1, len(resolved.Plan), countReplacedPages(results)))
-			}
-		}
-
-		out := map[string]interface{}{
-			"xml_presentation_id": resolved.PresentationID,
-			"pages_count":         len(resolved.Plan),
-			"results":             replacePageResultsOutput(results),
-			"status":              "completed",
-			"summary":             replacePagesSummaryOutput(results),
-			"note":                "batch replace is not atomic; each page was created before its old page was deleted",
-		}
-		if revisionID != replacePagesInitialRevisionID {
-			out["revision_id"] = revisionID
-		}
-		if hasReplacePageFailures(results) {
-			out["status"] = "partial_failure"
-			return runtime.OutPartialFailure(out, nil)
-		}
-		runtime.Out(out, nil)
-		return nil
-	},
-}
-
-type replacePageInput struct {
-	SlideID string
-	Content string
-}
-
-type replacePagePlanItem struct {
-	OldSlideID string
-	Content    string
-	Locator    string
-}
-
-type replacePagesPrepared struct {
-	PresentationID string
-	Plan           []replacePagePlanItem
-}
-
-type replacePageResult struct {
-	OldSlideID string
-	NewSlideID string
-	Status     string
-	Error      string
-	RevisionID *int
-}
-
-func prepareReplacePages(runtime *common.RuntimeContext) (*replacePagesPrepared, error) {
-	ref, err := parsePresentationRef(runtime.Str("presentation"))
-	if err != nil {
-		return nil, err
-	}
-	presentationID, err := resolvePresentationID(runtime, ref)
-	if err != nil {
-		return nil, err
-	}
-	pages, err := parseReplacePages(runtime.Str("pages"))
-	if err != nil {
-		return nil, err
-	}
-	if err := validateReplacePagesInput(pages); err != nil {
-		return nil, err
-	}
-
-	plan, err := buildReplacePagesPlan(pages)
-	if err != nil {
-		return nil, err
-	}
-	return &replacePagesPrepared{PresentationID: presentationID, Plan: plan}, nil
-}
-
-func parseReplacePages(raw string) ([]replacePageInput, error) {
-	s := strings.TrimSpace(raw)
-	if s == "" {
-		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages cannot be empty").WithParam("--pages")
-	}
-	var decoded []map[string]interface{}
-	if err := json.Unmarshal([]byte(s), &decoded); err != nil {
-		return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages invalid JSON, must be an array of objects: %v", err).WithParam("--pages").WithCause(err)
-	}
-	out := make([]replacePageInput, 0, len(decoded))
-	for i, m := range decoded {
-		p := replacePageInput{}
-		if v, ok := m["slide_number"]; ok {
-			_ = v
-			return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages[%d].slide_number is no longer supported; use slide_id", i).WithParam("--pages").WithHint("read current slide IDs first, then pass slide_id for each page replacement")
-		}
-		if v, ok := m["slide_id"]; ok {
-			s, ok := v.(string)
-			if !ok {
-				return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages[%d].slide_id must be a string", i).WithParam("--pages")
-			}
-			p.SlideID = s
-		}
-		if v, ok := m["content"]; ok {
-			s, ok := v.(string)
-			if !ok {
-				return nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages[%d].content must be a string", i).WithParam("--pages")
-			}
-			p.Content = s
-		}
-		out = append(out, p)
-	}
-	return out, nil
-}
-
-func validateReplacePagesInput(pages []replacePageInput) error {
-	if len(pages) == 0 {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages must contain at least 1 item").WithParam("--pages")
-	}
-	seenIDs := map[string]bool{}
-	for i, p := range pages {
-		id := strings.TrimSpace(p.SlideID)
-		if id == "" {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages[%d].slide_id is required", i).WithParam("--pages")
-		}
-		if seenIDs[id] {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages contains duplicate slide_id %q", id).WithParam("--pages")
-		}
-		seenIDs[id] = true
-		if strings.TrimSpace(p.Content) == "" {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages[%d].content cannot be empty", i).WithParam("--pages")
-		}
-		if err := validateCompleteSlideXML(p.Content); err != nil {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--pages[%d].content must be a complete <slide> XML element: %v", i, err).WithParam("--pages").WithCause(err)
-		}
-	}
-	return nil
-}
-
-func validateCompleteSlideXML(content string) error {
-	dec := xml.NewDecoder(strings.NewReader(content))
-	depth := 0
-	seenRoot := false
-	for {
-		tok, err := dec.Token()
-		if err == io.EOF {
-			break
-		}
-		if err != nil {
-			return err
-		}
-		switch t := tok.(type) {
-		case xml.StartElement:
-			if depth == 0 {
-				if seenRoot {
-					return fmt.Errorf("multiple root elements")
-				}
-				if t.Name.Local != "slide" {
-					return fmt.Errorf("root element is <%s>, want <slide>", t.Name.Local)
-				}
-				seenRoot = true
-			}
-			depth++
-		case xml.EndElement:
-			depth--
-		case xml.CharData:
-			if depth == 0 && strings.TrimSpace(string(t)) != "" {
-				return fmt.Errorf("non-whitespace text outside root element")
-			}
-		}
-	}
-	if !seenRoot {
-		return fmt.Errorf("missing root element")
-	}
-	if depth != 0 {
-		return fmt.Errorf("unclosed XML element")
-	}
-	return nil
-}
-
-func buildReplacePagesPlan(pages []replacePageInput) ([]replacePagePlanItem, error) {
-	plan := make([]replacePagePlanItem, 0, len(pages))
-	for _, page := range pages {
-		id := strings.TrimSpace(page.SlideID)
-		plan = append(plan, replacePagePlanItem{
-			OldSlideID: id,
-			Content:    page.Content,
-			Locator:    "slide_id",
-		})
-	}
-	return plan, nil
-}
-
-func appendReplacePagesDryRunCalls(dry *common.DryRunAPI, resolved *replacePagesPrepared) {
-	dry.Desc("Batch replace pages in-place: create each new page before old page, then delete old page (not atomic)")
-	for i, item := range resolved.Plan {
-		dry.POST(fmt.Sprintf("/open-apis/slides_ai/v1/xml_presentations/%s/slide", validate.EncodePathSegment(resolved.PresentationID))).
-			Desc(fmt.Sprintf("[%d/%d] Create replacement before old slide %s", i*2+1, len(resolved.Plan)*2, item.OldSlideID)).
-			Params(map[string]interface{}{"revision_id": "<latest_or_revision_returned_by_previous_step>"}).
-			Body(map[string]interface{}{
-				"slide":           map[string]interface{}{"content": item.Content},
-				"before_slide_id": item.OldSlideID,
-			})
-		dry.DELETE(fmt.Sprintf("/open-apis/slides_ai/v1/xml_presentations/%s/slide", validate.EncodePathSegment(resolved.PresentationID))).
-			Desc(fmt.Sprintf("[%d/%d] Delete old slide %s after create succeeds", i*2+2, len(resolved.Plan)*2, item.OldSlideID)).
-			Params(map[string]interface{}{
-				"slide_id":    item.OldSlideID,
-				"revision_id": "<revision_returned_by_create>",
-			})
-	}
-}
-
-func replaceOnePage(runtime *common.RuntimeContext, presentationID string, item replacePagePlanItem, revisionID int) (replacePageResult, error) {
-	result := replacePageResult{
-		OldSlideID: item.OldSlideID,
-		Status:     "pending",
-	}
-	slideURL := fmt.Sprintf("/open-apis/slides_ai/v1/xml_presentations/%s/slide", validate.EncodePathSegment(presentationID))
-	createData, err := runtime.CallAPITyped(
-		"POST",
-		slideURL,
-		map[string]interface{}{"revision_id": revisionID},
-		map[string]interface{}{
-			"slide":           map[string]interface{}{"content": item.Content},
-			"before_slide_id": item.OldSlideID,
-		},
-	)
-	if err != nil {
-		result.Status = "create_failed"
-		result.Error = err.Error()
-		return result, err
-	}
-	newSlideID := common.GetString(createData, "slide_id")
-	if newSlideID == "" {
-		err := errs.NewInternalError(errs.SubtypeInvalidResponse, "slide.create returned no slide_id for replacement of slide_id %q", item.OldSlideID)
-		result.Status = "create_failed"
-		result.Error = err.Error()
-		return result, err
-	}
-	result.NewSlideID = newSlideID
-	if rev, ok := revisionFromData(createData); ok {
-		revisionID = rev
-		result.RevisionID = &rev
-	}
-
-	deleteData, err := runtime.CallAPITyped(
-		"DELETE",
-		slideURL,
-		map[string]interface{}{
-			"slide_id":    item.OldSlideID,
-			"revision_id": revisionID,
-		},
-		nil,
-	)
-	if err != nil {
-		result.Status = "delete_failed"
-		result.Error = err.Error()
-		return result, err
-	}
-	if rev, ok := revisionFromData(deleteData); ok {
-		result.RevisionID = &rev
-	}
-	result.Status = "replaced"
-	return result, nil
-}
-
-func revisionFromData(data map[string]interface{}) (int, bool) {
-	if _, ok := data["revision_id"]; !ok {
-		return 0, false
-	}
-	return int(common.GetFloat(data, "revision_id")), true
-}
-
-func replacePagesPlanOutput(plan []replacePagePlanItem) []map[string]interface{} {
-	out := make([]map[string]interface{}, 0, len(plan))
-	for _, item := range plan {
-		out = append(out, map[string]interface{}{
-			"old_slide_id":           item.OldSlideID,
-			"insert_before_slide_id": item.OldSlideID,
-			"locator":                item.Locator,
-			"action":                 "create_before_then_delete_old",
-		})
-	}
-	return out
-}
-
-func replacePageResultsOutput(results []replacePageResult) []map[string]interface{} {
-	out := make([]map[string]interface{}, 0, len(results))
-	for _, result := range results {
-		m := map[string]interface{}{
-			"old_slide_id": result.OldSlideID,
-			"status":       result.Status,
-		}
-		if result.NewSlideID != "" {
-			m["new_slide_id"] = result.NewSlideID
-		}
-		if result.Error != "" {
-			m["error"] = result.Error
-		}
-		if result.RevisionID != nil {
-			m["revision_id"] = *result.RevisionID
-		}
-		out = append(out, m)
-	}
-	return out
-}
-
-func replacePagesSummaryOutput(results []replacePageResult) map[string]interface{} {
-	replaced := countReplacedPages(results)
-	return map[string]interface{}{
-		"replaced": replaced,
-		"failed":   len(results) - replaced,
-		"total":    len(results),
-	}
-}
-
-func countReplacedPages(results []replacePageResult) int {
-	n := 0
-	for _, result := range results {
-		if result.Status == "replaced" {
-			n++
-		}
-	}
-	return n
-}
-
-func hasReplacePageFailures(results []replacePageResult) bool {
-	for _, result := range results {
-		if result.Status == "create_failed" || result.Status == "delete_failed" {
-			return true
-		}
-	}
-	return false
-}

shortcuts/slides/slides_replace_pages_test.go

@@ -1,306 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package slides
-
-import (
-	"encoding/json"
-	"errors"
-	"strings"
-	"testing"
-
-	"github.com/larksuite/cli/errs"
-	"github.com/larksuite/cli/internal/cmdutil"
-	"github.com/larksuite/cli/internal/httpmock"
-	"github.com/larksuite/cli/internal/output"
-)
-
-func TestReplacePagesCreatesBeforeThenDeletesOld(t *testing.T) {
-	t.Parallel()
-
-	f, stdout, _, reg := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-	createStub := &httpmock.Stub{
-		Method: "POST",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{"slide_id": "new2", "revision_id": 11},
-		},
-	}
-	reg.Register(createStub)
-	deleteStub := &httpmock.Stub{
-		Method: "DELETE",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{"revision_id": 12},
-		},
-	}
-	reg.Register(deleteStub)
-
-	pages := `[{"slide_id":"old2","content":"<slide xmlns=\"http://www.larkoffice.com/sml/2.0\"><data></data></slide>"}]`
-	err := runSlidesShortcut(t, f, stdout, SlidesReplacePages, []string{
-		"+replace-pages",
-		"--presentation", "pres_abc",
-		"--pages", pages,
-		"--as", "user",
-	})
-	if err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-
-	var createBody struct {
-		Slide struct {
-			Content string `json:"content"`
-		} `json:"slide"`
-		BeforeSlideID string `json:"before_slide_id"`
-	}
-	if err := json.Unmarshal(createStub.CapturedBody, &createBody); err != nil {
-		t.Fatalf("decode create body: %v\nraw=%s", err, createStub.CapturedBody)
-	}
-	if createBody.BeforeSlideID != "old2" {
-		t.Fatalf("before_slide_id = %q, want old2", createBody.BeforeSlideID)
-	}
-	if !strings.Contains(createBody.Slide.Content, "<slide") {
-		t.Fatalf("create content = %q", createBody.Slide.Content)
-	}
-	deleteURL := string(deleteStub.CapturedBody)
-	if deleteURL != "" {
-		t.Fatalf("delete body = %q, want empty", deleteURL)
-	}
-
-	data := decodeShortcutData(t, stdout)
-	if data["xml_presentation_id"] != "pres_abc" {
-		t.Fatalf("xml_presentation_id = %v", data["xml_presentation_id"])
-	}
-	if data["revision_id"] != float64(12) {
-		t.Fatalf("revision_id = %v, want 12", data["revision_id"])
-	}
-	summary, _ := data["summary"].(map[string]interface{})
-	if summary["failed"] != float64(0) {
-		t.Fatalf("summary.failed = %v, want 0", summary["failed"])
-	}
-	results, _ := data["results"].([]interface{})
-	if len(results) != 1 {
-		t.Fatalf("results len = %d, want 1", len(results))
-	}
-	first, _ := results[0].(map[string]interface{})
-	if first["old_slide_id"] != "old2" || first["new_slide_id"] != "new2" || first["status"] != "replaced" {
-		t.Fatalf("result = %#v", first)
-	}
-}
-
-func TestReplacePagesContinueOnErrorReturnsPartialFailure(t *testing.T) {
-	t.Parallel()
-
-	f, stdout, _, reg := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-	reg.Register(&httpmock.Stub{
-		Method: "POST",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 3350001,
-			"msg":  "invalid param",
-			"data": map[string]interface{}{},
-		},
-	})
-	reg.Register(&httpmock.Stub{
-		Method: "POST",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{"slide_id": "new2", "revision_id": 11},
-		},
-	})
-	reg.Register(&httpmock.Stub{
-		Method: "DELETE",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{"revision_id": 12},
-		},
-	})
-
-	pages := `[
-		{"slide_id":"old1","content":"<slide xmlns=\"http://www.larkoffice.com/sml/2.0\"><data></data></slide>"},
-		{"slide_id":"old2","content":"<slide xmlns=\"http://www.larkoffice.com/sml/2.0\"><data></data></slide>"}
-	]`
-	err := runSlidesShortcut(t, f, stdout, SlidesReplacePages, []string{
-		"+replace-pages",
-		"--presentation", "pres_abc",
-		"--pages", pages,
-		"--continue-on-error",
-		"--as", "user",
-	})
-	var pfErr *output.PartialFailureError
-	if !errors.As(err, &pfErr) {
-		t.Fatalf("err = %T %v, want *output.PartialFailureError", err, err)
-	}
-
-	env := decodeReplacePagesEnvelope(t, stdout)
-	if env.OK {
-		t.Fatalf("stdout ok = true, want false for partial failure")
-	}
-	data := env.Data
-	if data["status"] != "partial_failure" {
-		t.Fatalf("status = %v, want partial_failure", data["status"])
-	}
-	summary, _ := data["summary"].(map[string]interface{})
-	if summary["replaced"] != float64(1) || summary["failed"] != float64(1) || summary["total"] != float64(2) {
-		t.Fatalf("summary = %#v, want replaced=1 failed=1 total=2", summary)
-	}
-	results, _ := data["results"].([]interface{})
-	if len(results) != 2 {
-		t.Fatalf("results len = %d, want 2", len(results))
-	}
-	first, _ := results[0].(map[string]interface{})
-	second, _ := results[1].(map[string]interface{})
-	if first["status"] != "create_failed" {
-		t.Fatalf("first status = %v, want create_failed", first["status"])
-	}
-	if second["status"] != "replaced" || second["new_slide_id"] != "new2" {
-		t.Fatalf("second result = %#v, want replaced with new2", second)
-	}
-}
-
-func TestReplacePagesContinueOnErrorDeleteFailureIncludesNewSlideID(t *testing.T) {
-	t.Parallel()
-
-	f, stdout, _, reg := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-	reg.Register(&httpmock.Stub{
-		Method: "POST",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{"slide_id": "new1", "revision_id": 11},
-		},
-	})
-	reg.Register(&httpmock.Stub{
-		Method: "DELETE",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc/slide",
-		Body: map[string]interface{}{
-			"code": 3350001,
-			"msg":  "invalid param",
-			"data": map[string]interface{}{},
-		},
-	})
-
-	pages := `[{"slide_id":"old1","content":"<slide xmlns=\"http://www.larkoffice.com/sml/2.0\"><data></data></slide>"}]`
-	err := runSlidesShortcut(t, f, stdout, SlidesReplacePages, []string{
-		"+replace-pages",
-		"--presentation", "pres_abc",
-		"--pages", pages,
-		"--continue-on-error",
-		"--as", "user",
-	})
-	var pfErr *output.PartialFailureError
-	if !errors.As(err, &pfErr) {
-		t.Fatalf("err = %T %v, want *output.PartialFailureError", err, err)
-	}
-
-	env := decodeReplacePagesEnvelope(t, stdout)
-	if env.OK {
-		t.Fatalf("stdout ok = true, want false for partial failure")
-	}
-	results, _ := env.Data["results"].([]interface{})
-	if len(results) != 1 {
-		t.Fatalf("results len = %d, want 1", len(results))
-	}
-	first, _ := results[0].(map[string]interface{})
-	if first["status"] != "delete_failed" {
-		t.Fatalf("status = %v, want delete_failed", first["status"])
-	}
-	if first["new_slide_id"] != "new1" {
-		t.Fatalf("new_slide_id = %v, want new1", first["new_slide_id"])
-	}
-}
-
-func TestReplacePagesDryRunPlansOnly(t *testing.T) {
-	t.Parallel()
-
-	f, stdout, _, _ := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-
-	pages := `[{"slide_id":"old2","content":"<slide xmlns=\"http://www.larkoffice.com/sml/2.0\"><data></data></slide>"}]`
-	err := runSlidesShortcut(t, f, stdout, SlidesReplacePages, []string{
-		"+replace-pages",
-		"--presentation", "pres_abc",
-		"--pages", pages,
-		"--dry-run",
-		"--as", "user",
-	})
-	if err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-
-	var out map[string]interface{}
-	if err := json.Unmarshal(stdout.Bytes(), &out); err != nil {
-		t.Fatalf("decode dry-run: %v\nraw=%s", err, stdout.String())
-	}
-	if out["xml_presentation_id"] != "pres_abc" {
-		t.Fatalf("xml_presentation_id = %v", out["xml_presentation_id"])
-	}
-	plan, _ := out["plan"].([]interface{})
-	if len(plan) != 1 {
-		t.Fatalf("plan len = %d, want 1", len(plan))
-	}
-	item, _ := plan[0].(map[string]interface{})
-	if item["old_slide_id"] != "old2" || item["action"] != "create_before_then_delete_old" {
-		t.Fatalf("plan item = %#v", item)
-	}
-	api, _ := out["api"].([]interface{})
-	if len(api) != 2 {
-		t.Fatalf("api len = %d, want create/delete plan", len(api))
-	}
-}
-
-func TestReplacePagesValidationParam(t *testing.T) {
-	t.Parallel()
-
-	tests := []struct {
-		name  string
-		pages string
-	}{
-		{"empty pages", `[]`},
-		{"slide number no longer supported", `[{"slide_number":1,"content":"<slide/>"}]`},
-		{"no locator", `[{"content":"<slide/>"}]`},
-		{"empty content", `[{"slide_id":"s1","content":"  "}]`},
-		{"not slide XML", `[{"slide_id":"s1","content":"<shape/>"}]`},
-		{"duplicate id", `[{"slide_id":"s1","content":"<slide/>"},{"slide_id":"s1","content":"<slide/>"}]`},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-			f, stdout, _, _ := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-			err := runSlidesShortcut(t, f, stdout, SlidesReplacePages, []string{
-				"+replace-pages",
-				"--presentation", "pres_abc",
-				"--pages", tt.pages,
-				"--as", "user",
-			})
-			var ve *errs.ValidationError
-			if !errors.As(err, &ve) {
-				t.Fatalf("err = %v, want *errs.ValidationError", err)
-			}
-			if ve.Param != "--pages" {
-				t.Fatalf("Param = %q, want --pages", ve.Param)
-			}
-		})
-	}
-}
-
-type replacePagesEnvelope struct {
-	OK   bool                   `json:"ok"`
-	Data map[string]interface{} `json:"data"`
-}
-
-func decodeReplacePagesEnvelope(t *testing.T, stdout interface{ Bytes() []byte }) replacePagesEnvelope {
-	t.Helper()
-	var env replacePagesEnvelope
-	if err := json.Unmarshal(stdout.Bytes(), &env); err != nil {
-		t.Fatalf("decode output: %v\nraw=%s", err, string(stdout.Bytes()))
-	}
-	if env.Data == nil {
-		t.Fatalf("missing data: %#v", env)
-	}
-	return env
-}

shortcuts/slides/slides_xml_get.go

@@ -1,144 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package slides
-
-import (
-	"bytes"
-	"context"
-	"fmt"
-	"strings"
-
-	"github.com/larksuite/cli/errs"
-	"github.com/larksuite/cli/extension/fileio"
-	"github.com/larksuite/cli/internal/validate"
-	"github.com/larksuite/cli/shortcuts/common"
-)
-
-// SlidesXMLGet fetches the full XML presentation content and writes it to a
-// local file, keeping the terminal output small for large decks.
-var SlidesXMLGet = common.Shortcut{
-	Service:     "slides",
-	Command:     "+xml-get",
-	Description: "Fetch full presentation XML and save it to a local file",
-	Risk:        "read",
-	Scopes:      []string{"slides:presentation:read"},
-	// wiki:node:read is required only when --presentation is a wiki URL.
-	ConditionalScopes: []string{"wiki:node:read"},
-	AuthTypes:         []string{"user", "bot"},
-	Flags: []common.Flag{
-		{Name: "presentation", Desc: "xml_presentation_id, slides URL, or wiki URL that resolves to slides", Required: true},
-		{Name: "output", Desc: "local XML output path; existing file is overwritten", Required: true},
-		{Name: "revision-id", Type: "int", Default: "-1", Desc: "presentation revision_id; -1 means latest"},
-		{Name: "remove-attr-id", Type: "bool", Desc: "remove XML id attributes in the returned content; useful for read-only inspection, not precise block editing"},
-	},
-	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		ref, err := parsePresentationRef(runtime.Str("presentation"))
-		if err != nil {
-			return err
-		}
-		if ref.Kind == "wiki" {
-			if err := runtime.EnsureScopes([]string{"wiki:node:read"}); err != nil {
-				return err
-			}
-		}
-		if strings.TrimSpace(runtime.Str("output")) == "" {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--output cannot be empty").WithParam("--output")
-		}
-		if _, err := runtime.ResolveSavePath(runtime.Str("output")); err != nil {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--output invalid: %v", err).WithParam("--output").WithCause(err)
-		}
-		if runtime.Int("revision-id") < -1 {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument, "--revision-id must be -1 or a non-negative integer").WithParam("--revision-id")
-		}
-		return nil
-	},
-	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		ref, err := parsePresentationRef(runtime.Str("presentation"))
-		if err != nil {
-			return common.NewDryRunAPI().Set("error", err.Error())
-		}
-		presentationID := ref.Token
-		dry := common.NewDryRunAPI()
-		if ref.Kind == "wiki" {
-			presentationID = "<resolved_slides_token>"
-			dry.Desc("2-step orchestration: resolve wiki → fetch full presentation XML").
-				GET("/open-apis/wiki/v2/spaces/get_node").
-				Desc("[1] Resolve wiki node to slides presentation").
-				Params(map[string]interface{}{"token": ref.Token})
-		} else {
-			dry.Desc("Fetch full presentation XML and save it to a local file")
-		}
-		params := map[string]interface{}{
-			"revision_id": runtime.Int("revision-id"),
-		}
-		if runtime.Bool("remove-attr-id") {
-			params["remove_attr_id"] = true
-		}
-		dry.GET(fmt.Sprintf(
-			"/open-apis/slides_ai/v1/xml_presentations/%s",
-			validate.EncodePathSegment(presentationID),
-		)).
-			Params(params)
-		return dry.Set("output", runtime.Str("output")).Set("stdout_content", "suppressed; XML content is saved to --output during execution")
-	},
-	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		ref, err := parsePresentationRef(runtime.Str("presentation"))
-		if err != nil {
-			return err
-		}
-		presentationID, err := resolvePresentationID(runtime, ref)
-		if err != nil {
-			return err
-		}
-
-		params := map[string]interface{}{
-			"revision_id": runtime.Int("revision-id"),
-		}
-		if runtime.Bool("remove-attr-id") {
-			params["remove_attr_id"] = true
-		}
-		data, err := runtime.CallAPITyped(
-			"GET",
-			fmt.Sprintf("/open-apis/slides_ai/v1/xml_presentations/%s", validate.EncodePathSegment(presentationID)),
-			params,
-			nil,
-		)
-		if err != nil {
-			return err
-		}
-
-		presentation := common.GetMap(data, "xml_presentation")
-		content := common.GetString(presentation, "content")
-		if content == "" {
-			return errs.NewInternalError(errs.SubtypeInvalidResponse, "slides xml get returned empty xml_presentation.content")
-		}
-		outputPath := runtime.Str("output")
-		result, err := runtime.FileIO().Save(outputPath, fileio.SaveOptions{
-			ContentType:   "application/xml",
-			ContentLength: int64(len(content)),
-		}, bytes.NewReader([]byte(content)))
-		if err != nil {
-			return common.WrapSaveErrorTyped(err)
-		}
-		resolvedPath, err := runtime.ResolveSavePath(outputPath)
-		if err != nil {
-			return errs.NewInternalError(errs.SubtypeFileIO, "resolve saved XML path %s: %v", outputPath, err).WithCause(err)
-		}
-
-		out := map[string]interface{}{
-			"xml_presentation_id": presentationID,
-			"path":                resolvedPath,
-			"size":                result.Size(),
-			"content_saved":       true,
-		}
-		if revisionID := common.GetFloat(presentation, "revision_id"); revisionID > 0 {
-			out["revision_id"] = int(revisionID)
-		}
-		if runtime.Bool("remove-attr-id") {
-			out["remove_attr_id"] = true
-		}
-		runtime.Out(out, nil)
-		return nil
-	},
-}

shortcuts/slides/slides_xml_get_test.go

@@ -1,157 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package slides
-
-import (
-	"net/http"
-	"net/url"
-	"os"
-	"path/filepath"
-	"strings"
-	"testing"
-
-	"github.com/larksuite/cli/errs"
-	"github.com/larksuite/cli/internal/cmdutil"
-	"github.com/larksuite/cli/internal/httpmock"
-)
-
-func TestSlidesXMLGetWritesContentToFileAndSuppressesXML(t *testing.T) {
-	dir := t.TempDir()
-	withSlidesTestWorkingDir(t, dir)
-
-	xml := `<presentation><slide id="s1"><shape id="a">hello</shape></slide></presentation>`
-	var capturedQuery url.Values
-	f, stdout, _, reg := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-	reg.Register(&httpmock.Stub{
-		Method: "GET",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_abc",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{
-				"xml_presentation": map[string]interface{}{
-					"presentation_id": "pres_abc",
-					"revision_id":     7,
-					"content":         xml,
-				},
-			},
-		},
-		OnMatch: func(req *http.Request) {
-			capturedQuery = req.URL.Query()
-		},
-	})
-
-	err := runSlidesShortcut(t, f, stdout, SlidesXMLGet, []string{
-		"+xml-get",
-		"--presentation", "pres_abc",
-		"--output", "readback.xml",
-		"--revision-id", "7",
-		"--remove-attr-id",
-		"--as", "user",
-	})
-	if err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-
-	path := filepath.Join(dir, "readback.xml")
-	got, err := os.ReadFile(path)
-	if err != nil {
-		t.Fatalf("read saved XML: %v", err)
-	}
-	if string(got) != xml {
-		t.Fatalf("saved XML = %q, want %q", got, xml)
-	}
-	if strings.Contains(stdout.String(), xml) {
-		t.Fatalf("stdout leaked full XML content: %s", stdout.String())
-	}
-	if got := capturedQuery.Get("revision_id"); got != "7" {
-		t.Fatalf("revision_id query = %q, want 7", got)
-	}
-	if got := capturedQuery.Get("remove_attr_id"); got != "true" {
-		t.Fatalf("remove_attr_id query = %q, want true", got)
-	}
-
-	data := decodeShortcutData(t, stdout)
-	if data["xml_presentation_id"] != "pres_abc" {
-		t.Fatalf("xml_presentation_id = %v, want pres_abc", data["xml_presentation_id"])
-	}
-	if data["revision_id"] != float64(7) {
-		t.Fatalf("revision_id = %v, want 7", data["revision_id"])
-	}
-	if data["size"] != float64(len(xml)) {
-		t.Fatalf("size = %v, want %d", data["size"], len(xml))
-	}
-	gotPath, _ := data["path"].(string)
-	if !filepath.IsAbs(gotPath) {
-		t.Fatalf("path = %v, want absolute path", gotPath)
-	}
-	if !strings.HasSuffix(gotPath, "readback.xml") {
-		t.Fatalf("path = %v, want readback.xml suffix", gotPath)
-	}
-}
-
-func TestSlidesXMLGetResolvesWikiPresentation(t *testing.T) {
-	dir := t.TempDir()
-	withSlidesTestWorkingDir(t, dir)
-
-	f, stdout, _, reg := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-	reg.Register(&httpmock.Stub{
-		Method: "GET",
-		URL:    "/open-apis/wiki/v2/spaces/get_node",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{
-				"node": map[string]interface{}{
-					"obj_type":  "slides",
-					"obj_token": "pres_real",
-				},
-			},
-		},
-	})
-	reg.Register(&httpmock.Stub{
-		Method: "GET",
-		URL:    "/open-apis/slides_ai/v1/xml_presentations/pres_real",
-		Body: map[string]interface{}{
-			"code": 0,
-			"data": map[string]interface{}{
-				"xml_presentation": map[string]interface{}{
-					"content": `<presentation/>`,
-				},
-			},
-		},
-	})
-
-	err := runSlidesShortcut(t, f, stdout, SlidesXMLGet, []string{
-		"+xml-get",
-		"--presentation", "https://example.feishu.cn/wiki/wikcn123",
-		"--output", "wiki.xml",
-		"--as", "user",
-	})
-	if err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	data := decodeShortcutData(t, stdout)
-	if data["xml_presentation_id"] != "pres_real" {
-		t.Fatalf("xml_presentation_id = %v, want pres_real", data["xml_presentation_id"])
-	}
-}
-
-func TestSlidesXMLGetRejectsUnsafeOutputPath(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, slidesTestConfig(t, ""))
-	err := runSlidesShortcut(t, f, stdout, SlidesXMLGet, []string{
-		"+xml-get",
-		"--presentation", "pres_abc",
-		"--output", "../readback.xml",
-		"--as", "user",
-	})
-	if err == nil {
-		t.Fatal("expected unsafe output path error, got nil")
-	}
-	problem, ok := errs.ProblemOf(err)
-	if !ok {
-		t.Fatalf("expected typed error, got %T %v", err, err)
-	}
-	if problem.Param != "--output" {
-		t.Fatalf("param = %q, want --output", problem.Param)
-	}
-}

shortcuts/task/task_search.go

@@ -73,12 +73,16 @@ var SearchTask = common.Shortcut{
 		var rawItems []interface{}
 		var lastPageToken string
 		var lastHasMore bool
+		var notice string
 		currentBody := body
 		for page := 0; page < pageLimit; page++ {
 			data, err := callTaskAPITyped(runtime, http.MethodPost, "/open-apis/task/v2/tasks/search", nil, currentBody)
 			if err != nil {
 				return err
 			}
+			if notice == "" {
+				notice, _ = data["notice"].(string)
+			}
 			items, _ := data["items"].([]interface{})
 			rawItems = append(rawItems, items...)
 			lastHasMore, _ = data["has_more"].(bool)
@@ -115,6 +119,9 @@ var SearchTask = common.Shortcut{
 			"page_token": lastPageToken,
 			"has_more":   lastHasMore,
 		}
+		if notice != "" {
+			outData["notice"] = notice
+		}
 		runtime.OutFormat(outData, &output.Meta{Count: len(enriched)}, func(w io.Writer) {
 			if len(enriched) == 0 {
 				fmt.Fprintln(w, "No tasks found.")

shortcuts/task/task_search_test.go

@@ -153,6 +153,7 @@ func TestSearchTask_DryRun(t *testing.T) {
 	}
 }
 
+// TestSearchTask_Execute verifies task search output, enrichment, and notices.
 func TestSearchTask_Execute(t *testing.T) {
 	tests := []struct {
 		name      string
@@ -171,6 +172,7 @@ func TestSearchTask_Execute(t *testing.T) {
 						"code": 0,
 						"msg":  "success",
 						"data": map[string]interface{}{
+							"notice":     "The query is too long and has been truncated to the first 50 characters for search.",
 							"has_more":   false,
 							"page_token": "",
 							"items": []interface{}{
@@ -191,7 +193,7 @@ func TestSearchTask_Execute(t *testing.T) {
 					},
 				})
 			},
-			wantParts: []string{`"guid": "task-123"`, `"summary": "Search Result"`},
+			wantParts: []string{`"guid": "task-123"`, `"summary": "Search Result"`, `"notice": "The query is too long and has been truncated to the first 50 characters for search."`},
 		},
 		{
 			name: "fallback to app link",

shortcuts/task/task_tasklist_search.go

@@ -70,12 +70,16 @@ var SearchTasklist = common.Shortcut{
 		var rawItems []interface{}
 		var lastPageToken string
 		var lastHasMore bool
+		var notice string
 		currentBody := body
 		for page := 0; page < pageLimit; page++ {
 			data, err := callTaskAPITyped(runtime, http.MethodPost, "/open-apis/task/v2/tasklists/search", nil, currentBody)
 			if err != nil {
 				return err
 			}
+			if notice == "" {
+				notice, _ = data["notice"].(string)
+			}
 			items, _ := data["items"].([]interface{})
 			rawItems = append(rawItems, items...)
 			lastHasMore, _ = data["has_more"].(bool)
@@ -118,6 +122,9 @@ var SearchTasklist = common.Shortcut{
 			"page_token": lastPageToken,
 			"has_more":   lastHasMore,
 		}
+		if notice != "" {
+			outData["notice"] = notice
+		}
 		runtime.OutFormat(outData, &output.Meta{Count: len(tasklists)}, func(w io.Writer) {
 			if len(tasklists) == 0 {
 				fmt.Fprintln(w, "No tasklists found.")

shortcuts/task/task_tasklist_search_test.go

@@ -126,6 +126,7 @@ func TestSearchTasklist_DryRun(t *testing.T) {
 	}
 }
 
+// TestSearchTasklist_Execute verifies tasklist search output, enrichment, and notices.
 func TestSearchTasklist_Execute(t *testing.T) {
 	tests := []struct {
 		name      string
@@ -144,6 +145,7 @@ func TestSearchTasklist_Execute(t *testing.T) {
 						"code": 0,
 						"msg":  "success",
 						"data": map[string]interface{}{
+							"notice":     "The query is too long and has been truncated to the first 50 characters for search.",
 							"has_more":   false,
 							"page_token": "",
 							"items":      []interface{}{map[string]interface{}{"id": "tl-123"}},
@@ -162,7 +164,7 @@ func TestSearchTasklist_Execute(t *testing.T) {
 					},
 				})
 			},
-			wantParts: []string{`"guid": "tl-123"`, `"name": "Q2 Plan"`},
+			wantParts: []string{`"guid": "tl-123"`, `"name": "Q2 Plan"`, `"notice": "The query is too long and has been truncated to the first 50 characters for search."`},
 		},
 		{
 			name: "fallback on detail error",

shortcuts/vc/vc_search.go

@@ -236,6 +236,9 @@ var VCSearch = common.Shortcut{
 			"has_more":   data["has_more"],
 			"page_token": data["page_token"],
 		}
+		if notice, _ := data["notice"].(string); notice != "" {
+			outData["notice"] = notice
+		}
 		hasMore, _ := data["has_more"].(bool)
 		runtime.OutFormat(outData, &output.Meta{Count: len(items)}, func(w io.Writer) {
 			if len(items) == 0 {

shortcuts/vc/vc_search_test.go

@@ -5,6 +5,7 @@ package vc
 
 import (
 	"context"
+	"encoding/json"
 	"errors"
 	"strings"
 	"testing"
@@ -14,6 +15,7 @@ import (
 	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/httpmock"
 	"github.com/larksuite/cli/shortcuts/common"
 )
 
@@ -253,6 +255,7 @@ func TestSearch_Validation_InvalidPageSize(t *testing.T) {
 	}
 }
 
+// TestSearch_DryRun verifies meeting search dry-run includes the API path.
 func TestSearch_DryRun(t *testing.T) {
 	f, stdout, _, _ := cmdutil.TestFactory(t, defaultConfig())
 	err := mountAndRun(t, VCSearch, []string{"+search", "--query", "test", "--dry-run", "--as", "user"}, f, stdout)
@@ -264,6 +267,43 @@ func TestSearch_DryRun(t *testing.T) {
 	}
 }
 
+// TestSearch_ExecutePassesThroughNotice verifies meeting search notice output.
+func TestSearch_ExecutePassesThroughNotice(t *testing.T) {
+	const notice = "The query is too long and has been truncated to the first 50 characters for search."
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, defaultConfig())
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/vc/v1/meetings/search",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "ok",
+			"data": map[string]interface{}{
+				"notice":     notice,
+				"items":      []interface{}{},
+				"total":      0,
+				"has_more":   false,
+				"page_token": "",
+			},
+		},
+	})
+
+	if err := mountAndRun(t, VCSearch, []string{"+search", "--query", "incident", "--format", "json", "--as", "user"}, f, stdout); err != nil {
+		t.Fatalf("VCSearch.Execute() error = %v", err)
+	}
+	reg.Verify(t)
+
+	var env map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &env); err != nil {
+		t.Fatalf("json.Unmarshal(stdout) error = %v\nstdout=%s", err, stdout.String())
+	}
+	data, _ := env["data"].(map[string]interface{})
+	if got, _ := data["notice"].(string); got != notice {
+		t.Fatalf("data.notice = %q, want %q; data=%#v", got, notice, data)
+	}
+}
+
+// TestSearch_InvalidTimeRange verifies invalid meeting search time input fails.
 func TestSearch_InvalidTimeRange(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, defaultConfig())
 	err := mountAndRun(t, VCSearch, []string{"+search", "--start", "bad-time", "--as", "user"}, f, nil)

skill-template/domains/drive.md

@@ -61,7 +61,7 @@ lark-cli drive +inspect --url 'https://xxx.feishu.cn/wiki/wikcnXXX'
    | `docx` | 新版云文档 | `drive file.comments.*`、`docx.*` |
    | `doc` | 旧版云文档 | `drive file.comments.*` |
    | `sheet` | 电子表格 | `sheets.*` |
-   | `bitable` | 多维表格 | `bitable.*` |
+   | `bitable` | 多维表格 / Base | `drive file.comments.*`、`bitable.*` |
    | `slides` | 幻灯片 | `drive.*` |
    | `file` | 文件 | `drive.*` |
    | `mindnote` | 思维导图 | `drive.*` |
@@ -112,8 +112,8 @@ Drive Folder (云空间文件夹)
 | 操作 | 需要的 Token | 说明 |
 |------|-------------|------|
 | 读取文档内容 | `file_token` / 通过 `docs +fetch --api-version v2` 自动处理 | `docs +fetch --api-version v2` 支持直接传入 URL |
-| 添加局部评论（划词评论） | `file_token` | 传 `--block-id` 时，`drive +add-comment` 会创建局部评论；仅支持 `docx`，以及最终解析为 `docx` 的 wiki URL |
-| 添加全文评论 | `file_token` | 不传 `--block-id` 时，`drive +add-comment` 默认创建全文评论；支持 `docx`、旧版 `doc` URL，以及最终解析为 `doc`/`docx` 的 wiki URL |
+| 添加局部评论（划词评论） | `file_token` | 传 `--block-id` 时，`drive +add-comment` 会创建局部评论；`docx` 支持文本定位或 block_id，`sheet` 使用 `<sheetId>!<cell>`，`slides` 使用 `<slide-block-type>!<xml-id>`；Base / bitable 只有记录局部评论，定位为 file_token(base token) + `--block-id <table-id>!<record-id>!<view-id>` |
+| 添加全文评论 | `file_token` | 不传 `--block-id` 时，`drive +add-comment` 默认创建全文评论；支持 `docx`、旧版 `doc` URL、白名单扩展名的 Drive file，以及最终解析为 `doc`/`docx`/`file` 的 wiki URL |
 | 下载文件 | `file_token` | 从文件 URL 中直接提取 |
 | 上传文件 | `folder_token` / `wiki_node_token` | 目标位置的 token |
 | 列出文档评论 | `file_token` | 同添加评论 |
@@ -121,11 +121,15 @@ Drive Folder (云空间文件夹)
 ### 评论能力边界（关键！）
 
 - `drive +add-comment` 支持两种模式。
-- 全文评论：未传 `--block-id` 时默认启用，也可显式传 `--full-comment`；支持 `docx`、旧版 `doc` URL，以及最终解析为 `doc`/`docx` 的 wiki URL。
-- 局部评论：传 `--block-id` 时启用；仅支持 `docx`，以及最终解析为 `docx` 的 wiki URL。block ID 可通过 `docs +fetch --api-version v2 --detail with-ids` 获取。
+- 全文评论：未传 `--block-id` 时默认启用，也可显式传 `--full-comment`；支持 `docx`、旧版 `doc` URL、白名单扩展名的 Drive file，以及最终解析为 `doc`/`docx`/`file` 的 wiki URL。`sheet`、`slides`、Base / bitable 不支持全文评论。
+- 局部评论：传 `--block-id` 时启用；`docx` 支持文本定位或 block id，`sheet` 支持 `<sheetId>!<cell>`，`slides` 支持 `<slide-block-type>!<xml-id>`，Base / bitable 支持 `<table-id>!<record-id>!<view-id>`；wiki URL 解析到这些类型时也支持对应局部评论。Drive file 本次只支持全文评论，不支持局部评论。
+- Drive file 评论仅支持白名单扩展名：`.md`、`.txt`、`.json`、`.csv`、`.go`、`.js`、`.py`、`.pptx`、`.png`、`.jpg`、`.jpeg`、`.zip`、`.mp3`、`.mp4`。`.pdf`、`.docx`、`.xlsx` 等未在白名单内的普通文件暂不支持，CLI 会直接报错提示当前还不支持这种类型的评论。
+- Review / 审阅 / 校对 / 逐条指出问题场景优先使用局部评论，不要把多个可定位问题汇总成一条全文评论；具体参数和定位方式见生成后的 `skills/lark-drive/references/lark-drive-add-comment.md`。
 - `drive +add-comment` 的 `--content` 需要传 `reply_elements` JSON 数组字符串，例如 `--content '[{"type":"text","text":"正文"}]'`。
-- 如果 wiki 解析后不是 `doc`/`docx`，不要用 `+add-comment`。
-- 如果需要更底层地直接调用评论 V2 协议，再走原生 API：先执行 `lark-cli schema drive.file.comments.create_v2`，再执行 `lark-cli drive file.comments create_v2 ...`。全文评论省略 `anchor`，局部评论传 `anchor.block_id`。
+- `slides` 评论要求显式传 `--block-id <slide-block-type>!<xml-id>`；CLI 会将其拆分后写入 `anchor.block_id` 和 `anchor.slide_block_type`。其中 `<xml-id>` 是 PPT XML 协议中的元素 `id`；不支持 `--selection-with-ellipsis` 和 `--full-comment`。
+- Base 记录局部评论使用 `--type bitable` / `--type base` 或 `/base/`、`/bitable/`、wiki Base 链接；`bitable` 和 Base 是同一概念，`bitable` 是内部代号、Base 是产品名，裸 token 推荐传 `bitable`，`base` 仅作为兼容别名兜底。Base 不支持全局评论，所有评论都挂在记录上；定位信息必须是 file token（base token）+ `--block-id <table-id>!<record-id>!<view-id>`，其中 table/record/view ID 通常分别以 `tbl`/`rec`/`vew` 开头。view_id 只决定被提及时点击通知打开哪个视图，不影响评论挂载点，但必须传；ID 可通过 [`lark-base`](../../skills/lark-base/SKILL.md) 获取。
+- 如果 wiki 解析后不是 `doc`/`docx`/`file`/`sheet`/`slides`/`bitable`/`base`，不要用 `+add-comment`。
+- 如果需要更底层地直接调用评论 V2 协议，再走原生 API：先执行 `lark-cli schema drive.file.comments.create_v2`，再执行 `lark-cli drive file.comments create_v2 ...`。全文评论省略 `anchor`；docx/sheet/slides 局部评论传 `anchor.block_id`，Base 记录局部评论传 `anchor.block_id`（table_id）、`anchor.base_record_id`、`anchor.base_view_id`。
 
 ### 评论查询与统计口径（关键！）
 
@@ -189,7 +193,7 @@ lark-cli drive file.comments list --params '{"file_token": "xxx", "file_type": "
 |----------|------|----------|
 | `not exist` | 使用了错误的 token | 检查 token 类型，wiki 链接必须先查询获取 `obj_token` |
 | `permission denied` | 没有相关操作权限 | 引导用户检查当前身份对文档/文件是否有相应操作权限；如果需要，可以授予相应权限 |
-| `invalid file_type` | file_type 参数错误 | 根据 `obj_type` 传入正确的 file_type（docx/doc/sheet） |
+| `invalid file_type` | file_type 参数错误 | 根据 `obj_type` 传入正确的 file_type（docx/doc/sheet/slides/bitable） |
 
 ### 授权当前应用访问文档
 

skills/lark-drive/SKILL.md

@@ -25,7 +25,7 @@ metadata:
 - 用户给出 doubao.com 的云空间资源 URL/token，或明确提到豆包里的 file/folder/docx/sheet/bitable/wiki 资源时，仍按资源类型、URL 路径和 token 路由到本 skill；不要因为域名不是飞书而回退到 WebFetch。
 - 用户要把本地 `.xlsx` / `.csv` / `.base` 导入成 Base / 多维表格 / bitable，第一步必须使用 `lark-cli drive +import --type bitable`。
 - 用户要把本地 `.md` / `.docx` / `.doc` / `.txt` / `.html` 导入成在线文档，使用 `lark-cli drive +import --type docx`。
-- 用户要把本地 `.pptx` / `.pdf` 导入成飞书幻灯片，使用 `lark-cli drive +import --type slides`；当前 PPTX/PDF 导入上限是 500MB。PDF 导入 Slides 通常可按每页约 10 秒预估处理时间，轮询或续查时不要过早放弃。
+- 用户要把本地 `.pptx` 导入成飞书幻灯片，使用 `lark-cli drive +import --type slides`；当前 PPTX 导入上限是 500MB。
 - 用户要在 Drive 里上传、创建、读取、局部 patch 或覆盖更新**原生 `.md` 文件**（不是导入成 docx），切到 [`lark-markdown`](../lark-markdown/SKILL.md)。
 - 用户要比较原生 `.md` 文件的**历史版本差异**，或比较远端 Markdown 与本地草稿，切到 [`lark-markdown`](../lark-markdown/SKILL.md) 的 `lark-cli markdown +diff`；需要版本号时先用 `drive +version-history`。
 - 用户要查看、下载、回滚或删除文件的**历史版本**，使用 `drive +version-history`、`drive +version-get`、`drive +version-revert`、`drive +version-delete`；这组命令同时支持 `--as user` 和 `--as bot`，自动化场景优先 `--as bot`。
@@ -70,7 +70,7 @@ lark-cli drive +inspect --url 'https://xxx.feishu.cn/wiki/wikcnXXX'
 | 操作 | 需要的 Token | 说明 |
 |------|-------------|------|
 | 读取文档内容 | `file_token` / 通过 `docs +fetch --api-version v2` 自动处理 | `docs +fetch --api-version v2` 支持直接传入 URL |
-| 添加局部评论（划词评论） | `file_token` | 传 `--block-id` 时，`drive +add-comment` 会创建局部评论；`docx` 支持文本定位或 block_id，`sheet` 使用 `<sheetId>!<cell>`，`slides` 使用 `<slide-block-type>!<xml-id>`，且都支持最终解析到对应类型的 wiki URL；Drive file 不支持局部评论 |
+| 添加局部评论（划词评论） | `file_token` | 传 `--block-id` 时，`drive +add-comment` 会创建局部评论；`docx` 支持文本定位或 block_id，`sheet` 使用 `<sheetId>!<cell>`，`slides` 使用 `<slide-block-type>!<xml-id>`；Base 只有记录局部评论，定位为 file_token(base_token) + `--block-id <table-id>!<record-id>!<view-id>` |
 | 添加全文评论 | `file_token` | 不传 `--block-id` 时，`drive +add-comment` 默认创建全文评论；支持 `docx`、旧版 `doc` URL、白名单扩展名的 Drive file，以及最终解析为 `doc`/`docx`/`file` 的 wiki URL |
 | 下载文件 | `file_token` | 从文件 URL 中直接提取 |
 | 上传文件 | `folder_token` / `wiki_node_token` | 目标位置的 token |
@@ -82,6 +82,15 @@ lark-cli drive +inspect --url 'https://xxx.feishu.cn/wiki/wikcnXXX'
 - 评论查询、统计、排序、回复限制，先读 [`lark-drive-comments-guide.md`](references/lark-drive-comments-guide.md)。
 - 需要根据评论定位正文位置时，先确认目标是 `file_type=docx`，再读 [`lark-drive-comment-location.md`](references/lark-drive-comment-location.md)；其他文档类型暂不支持返回定位字段。
 - reaction / 表情相关操作先读 [`lark-drive-reactions.md`](references/lark-drive-reactions.md)；只有用户明确需要 reaction 信息时才带 `need_reaction=true`。
+- `drive +add-comment` 的 `--content` 需要传 `reply_elements` JSON 数组字符串，例如 `--content '[{"type":"text","text":"正文"}]'`。
+- `slides` 评论要求显式传 `--block-id <slide-block-type>!<xml-id>`；CLI 会将其拆分后写入 `anchor.block_id` 和 `anchor.slide_block_type`。其中 `<xml-id>` 是 PPT XML 协议中的元素 `id`；不支持 `--selection-with-ellipsis` 和 `--full-comment`。
+- 评论写入内容（添加评论、回复评论、编辑回复）里的文本不能直接出现 `<`、`>`；提交前必须先转义：`<` -> `&lt;`，`>` -> `&gt;`。
+- 使用 `drive +add-comment` 时，shortcut 会对 `type=text` 的文本元素自动做上述转义兜底；如果直接调用 `drive file.comments create_v2`、`drive file.comment.replys create`、`drive file.comment.replys update`，则需要在请求里自行传入已转义的内容。
+- Base 记录局部评论使用 `--type bitable` / `--type base` 或 `/base/`、`/bitable/`、wiki Base 链接；`bitable` 和 Base 是同一概念，`bitable` 是内部代号、Base 是产品名，裸 token 推荐传 `bitable`，`base` 仅作为兼容别名兜底。
+- Base 不支持全局评论，所有评论都挂在记录上；定位信息必须是 file token（base token）+ `--block-id <table-id>!<record-id>!<view-id>`，其中 table/record/view ID 通常分别以 `tbl`/`rec`/`vew` 开头。view_id 只决定被提及时点击通知打开哪个视图，不影响评论挂载点；只要在同一记录上都能看到评论，但必须传，否则通知无法确定跳转视图。ID 可通过 [`lark-base`](../lark-base/SKILL.md) 获取。
+- 如果 wiki 解析后不是 `doc`/`docx`/`file`/`sheet`/`slides`/`bitable`/`base`，不要用 `+add-comment`。
+- 如果需要更底层地直接调用评论 V2 协议，再走原生 API：先执行 `lark-cli schema drive.file.comments.create_v2`，再执行 `lark-cli drive file.comments create_v2 ...`。全文评论省略 `anchor`；docx/sheet/slides 局部评论传 `anchor.block_id`，Base 记录局部评论传 `anchor.block_id`（table_id）、`anchor.base_record_id`、`anchor.base_view_id`。
+- 直接调用原生 `drive.file.comments.*` / `drive.file.comment.replys.*` 评论 Base 文档时，`file_type` 填 `bitable`，不要填 `base`。
 
 ### 典型错误与解决方案
 
@@ -89,7 +98,7 @@ lark-cli drive +inspect --url 'https://xxx.feishu.cn/wiki/wikcnXXX'
 |----------|------|----------|
 | `not exist` | 使用了错误的 token | 检查 token 类型，wiki 链接必须先查询获取 `obj_token` |
 | `permission denied` | 没有相关操作权限 | 引导用户检查当前身份对文档/文件是否有相应操作权限；如果需要，可以授予相应权限 |
-| `invalid file_type` | file_type 参数错误 | 根据 `obj_type` 传入正确的 file_type（docx/doc/sheet/slides） |
+| `invalid file_type` | file_type 参数错误 | 根据 `obj_type` 传入正确的 file_type（docx/doc/sheet/slides/bitable） |
 
 ### 权限能力入口
 
@@ -122,7 +131,7 @@ Shortcut 是对常用操作的高级封装（`lark-cli drive +<verb> [flags]`）
 | `+sync` | 双向同步本地目录与 Drive 文件夹：拉取 `new_remote`、推送 `new_local`，`modified` 按 `--on-conflict=remote-wins\|local-wins\|keep-both\|ask` 处理；`--quick` 用修改时间近似比较；`--on-duplicate-remote` 支持 `fail` / `newest` / `oldest`；只同步 `type=file`，跳过在线文档和 shortcut，且不会删除两端多余文件。 |
 | [`+push`](references/lark-drive-push.md) | 将本地目录推送到 Drive 文件夹，支持 skip / smart / overwrite 与确认后删除远端。 |
 | [`+create-shortcut`](references/lark-drive-create-shortcut.md) | 在另一个文件夹里创建现有 Drive 文件的快捷方式。 |
-| [`+add-comment`](references/lark-drive-add-comment.md) | 给 doc/docx/file/sheet/slides 添加评论；评论统计、回复和 reaction 细则见 [`lark-drive-comments-guide.md`](references/lark-drive-comments-guide.md)。 |
+| [`+add-comment`](references/lark-drive-add-comment.md) | 给 doc/docx/file/sheet/slides/base(bitable) 添加评论，也支持解析到这些类型的 wiki URL；评论统计、回复和 reaction 细则见 [`lark-drive-comments-guide.md`](references/lark-drive-comments-guide.md)。 |
 | [`+export`](references/lark-drive-export.md) | 将 doc/docx/sheet/bitable/slides 导出为本地文件。 |
 | [`+export-download`](references/lark-drive-export-download.md) | 根据导出产物的 file_token 下载文件。 |
 | [`+import`](references/lark-drive-import.md) | 将本地文件导入为飞书在线文档、表格、多维表格或幻灯片。 |

skills/lark-drive/references/lark-drive-add-comment.md

@@ -3,7 +3,7 @@
 
 > **前置条件：** 先阅读 [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) 了解认证、全局参数和安全规则。
 
-给文档、受支持的 Drive 普通文件、电子表格或飞书幻灯片添加评论。底层统一走 `/open-apis/drive/v1/files/:file_token/new_comments`（`create_v2`）接口；未指定位置时省略 `anchor` 创建全文评论，指定 `--block-id` 时传入 `anchor.block_id` 创建局部评论。支持直接传 docx URL/token、旧版 doc URL（仅全文评论）、Drive file URL/token（**仅支持白名单扩展名，且只支持全文评论**）、sheet URL、slides URL，也支持传最终可解析为 doc/docx/file/sheet/slides 的 wiki URL。
+给文档、受支持的 Drive 普通文件、电子表格、飞书幻灯片或 Base 添加评论。未指定位置时创建全文评论，但仅适用于 doc/docx、白名单 Drive file，以及解析为这些类型的 wiki；sheet、slides、Base(bitable) 必须指定 `--block-id`。不同类型的 `--block-id` 格式见下文。支持直接传 docx URL/token、旧版 doc URL（仅全文评论）、Drive file URL/token（**仅支持白名单扩展名，且只支持全文评论**）、sheet URL、slides URL、base/bitable URL，也支持传最终可解析为 doc/docx/file/sheet/slides/base(bitable) 的 wiki URL。
 
 ## 命令
 
@@ -127,6 +127,18 @@ lark-cli drive file.comments create_v2 \
   --params '{"file_token":"<DOC_TOKEN>"}' \
   --data '{"file_type":"docx","reply_elements":[{"type":"text","text":"全文评论内容"}]}'
 
+# Base 记录局部评论；原生 file_type 传 bitable。
+lark-cli drive +add-comment \
+  --doc "<BASE_TOKEN>" --type bitable \
+  --block-id "<TABLE_ID>!<RECORD_ID>!<VIEW_ID>" \
+  --content '[{"type":"text","text":"Base record-local comment"}]'
+
+# `base` 也可作为裸 token 类型别名；/base/ 与 /bitable/ URL 都会自动识别为 Base。
+lark-cli drive +add-comment \
+  --doc "<BASE_TOKEN>" --type base \
+  --block-id "<TABLE_ID>!<RECORD_ID>!<VIEW_ID>" \
+  --content '[{"type":"text","text":"Base alias comment"}]'
+
 # 预览底层调用链
 lark-cli drive +add-comment \
   --doc "https://example.larksuite.com/docx/<DOC_ID>" \
@@ -139,11 +151,11 @@ lark-cli drive +add-comment \
 
 | 参数 | 必填 | 说明 |
 |------|------|------|
-| `--doc` | 是 | 文档 URL / token、file / sheet / slides URL，或可解析到 `doc`/`docx`/`file`/`sheet`/`slides` 的 wiki URL |
-| `--type` | 裸 token 时必填 | 文档类型：`doc`、`docx`、`file`、`sheet`、`slides`。URL 输入时自动识别，无需传 |
+| `--doc` | 是 | 文档 URL / token、file / sheet / slides / base / bitable URL，或可解析到 `doc`/`docx`/`file`/`sheet`/`slides`/`base(bitable)` 的 wiki URL |
+| `--type` | 裸 token 时必填 | 文档类型：`doc`、`docx`、`file`、`sheet`、`slides`、`bitable`、`base`；评论 Base 文档推荐传 `bitable`，`base` 仅作为兼容别名兜底。URL 输入时自动识别，无需传 |
 | `--content` | 是 | `reply_elements` JSON 数组字符串。示例：`'[{"type":"text","text":"文本"},{"type":"mention_user","text":"ou_xxx"},{"type":"link","text":"https://example.com"}]'` |
-| `--full-comment` | 否 | 显式指定创建全文评论；未传 `--block-id` 时也会默认走全文评论（不适用于 sheet） |
-| `--block-id` | 局部评论时必填 | 目标块 ID，可通过 `docs +fetch --api-version v2 --detail with-ids` 获取。**Sheet 评论**：格式为 `<sheetId>!<cell>`（如 `a281f9!D6`） |
+| `--full-comment` | 否 | 显式指定创建全文评论；未传 `--block-id` 时也会默认走全文评论（仅适用于 doc/docx、白名单 Drive file，以及解析为这些类型的 wiki；不适用于 sheet、slides、Base / bitable） |
+| `--block-id` | 局部评论时必填 | 目标块 ID，可通过 `docs +fetch --api-version v2 --detail with-ids` 获取；sheet 用 `<sheetId>!<cell>`，slides 用 `<slide-block-type>!<xml-id>`，Base 用 `<table-id>!<record-id>!<view-id>` |
 
 ## 行为说明
 
@@ -152,10 +164,11 @@ lark-cli drive +add-comment \
 - 未传 `--block-id` 时，shortcut 默认创建**全文评论**；也可以显式传 `--full-comment`。全文评论支持 `docx`、旧版 `doc` URL、白名单扩展名的 Drive file，以及最终可解析为 `doc`/`docx`/`file` 的 wiki URL。
 - **Drive file 评论**：仅支持白名单扩展名的普通文件。当前支持：`.md`、`.txt`、`.json`、`.csv`、`.go`、`.js`、`.py`、`.pptx`、`.png`、`.jpg`、`.jpeg`、`.zip`、`.mp3`、`.mp4`。
 - **Drive file 暂不支持**：`.pdf`、`.docx`、`.xlsx` 等未在白名单内的普通文件会被 CLI 拒绝，并提示“当前还不支持这种类型的评论”。这些类型虽然可能接受 OpenAPI 请求，但在页面评论展示上存在问题。
-- **Drive file 只支持全文评论**：file 目标不支持局部评论，不允许传 `--block-id` 或 `--selection-with-ellipsis`。由于当前 OpenAPI 要求 file 评论传入非空 `anchor.block_id`，CLI 会固定传占位值 `test`，UI 上仍表现为文件全文评论。
-- 传 `--block-id` 时，shortcut 创建**局部评论（划词评论）**；该模式支持 `docx`、`sheet`、`slides`，以及最终可解析为这些类型的 wiki URL。
+- **Drive file 只支持全文评论**：file 目标不支持局部评论，不允许传 `--block-id` 或 `--selection-with-ellipsis`。
+- 传 `--block-id` 时，shortcut 创建**局部评论（划词评论）**；该模式支持 `docx`、`sheet`、`slides`、Base / bitable，以及最终可解析为这些类型的 wiki URL。
 - **Sheet 评论**：当 `--doc` 为 sheet URL 或 wiki 解析为 sheet 时，使用 `--block-id "<sheetId>!<cell>"` 指定单元格（如 `a281f9!D6`）；sheet 没有全文评论，`--full-comment` 不可用。
-- **Slide 评论**：当 `--doc` 为 slides URL、`--type slides`，或 wiki 解析为 slides 时，必须传 `--block-id "<SLIDE_BLOCK_TYPE>!<XML_ELEMENT_ID>"`。CLI 会将其拆分映射到 `anchor.block_id` / `anchor.slide_block_type`。此时 `--full-comment` 和 `--selection-with-ellipsis` 不可用。
+- **Slide 评论**：当 `--doc` 为 slides URL、`--type slides`，或 wiki 解析为 slides 时，必须传 `--block-id "<SLIDE_BLOCK_TYPE>!<XML_ELEMENT_ID>"`。此时 `--full-comment` 和 `--selection-with-ellipsis` 不可用。
+- **Base 记录局部评论**：Base 不支持全局评论，所有评论都挂在记录上；裸 token 可传 `--type bitable` 或 `--type base`，推荐 `bitable`。定位信息必须是 file token（base token）+ `--block-id "<table-id>!<record-id>!<view-id>"`，其中 table/record/view ID 通常分别以 `tbl`/`rec`/`vew` 开头；view_id 只决定被提及时点击通知打开哪个视图，不影响评论挂载点，但必须传。ID 获取参考 [`lark-base`](../../lark-base/SKILL.md)。
 - **Slide 参数映射示例**：`--block-id` 由 PPT XML 元素类型和元素 `id` 组成。例如：
     - `<slide id="pkk">` 对应 `--block-id slide!pkk`，表示给整页评论。
     - `<img id="bPk" ... />` 对应 `--block-id img!bPk`，表示给图片元素评论。
@@ -165,13 +178,11 @@ lark-cli drive +add-comment \
 - `type=text` 的评论文本不能直接包含 `<`、`>`；应优先传 `&lt;`、`&gt;`。shortcut 在发送前也会自动将 `<`、`>` 转义为 `&lt;`、`&gt;` 作为兜底。
 - **所有 `type=text` 元素的字符总和 ≤ 10000**（按字符算，中英文 / 符号一视同仁）。超过会被 shortcut 在发送前拒绝，并指出累计超长的元素。**拆成多个 text element 不能绕过这个上限**——上限是总额，不是每元素。需要更长内容就缩短或拆成多条评论。
 - 长度限制只对 `type=text` 生效，`mention_user` / `link` 不计入。
-- 写入评论前会自动生成符合 OpenAPI 定义的请求体：
-  - 统一接口：`POST /new_comments`
-  - 统一字段：`file_type` + `reply_elements`
-  - 全文评论：省略 `anchor`
-  - 局部评论：传入 `anchor.block_id`
+- 写入评论前会自动生成符合 OpenAPI 定义的请求体；shortcut 用户只需要传 `--doc`、`--content`，局部评论再传对应格式的 `--block-id`。
 - `--dry-run` 仅预览调用链和请求体，不会实际写入。
 - 如果需要更底层的控制，仍可改用 `lark-cli schema drive.file.comments.create_v2` + `lark-cli drive file.comments create_v2`。
+- 直接调用原生 `drive.file.comments.create_v2` 时，全文评论省略 `anchor`；docx/sheet/slides 局部评论传 `anchor.block_id`，Base 记录局部评论传 `anchor.block_id`（table_id）、`anchor.base_record_id`、`anchor.base_view_id`。
+- 直接调用原生 `drive.file.comments.*` / `drive.file.comment.replys.*` 评论 Base 文档时，`file_type` 填 `bitable`，不要填 `base`。
 
 > [!CAUTION]
 > 这是**写入操作** —— 执行前必须确认用户意图。

skills/lark-drive/references/lark-drive-import.md

@@ -2,7 +2,7 @@
 
 > **前置条件：** 先阅读 [`../lark-shared/SKILL.md`](../../lark-shared/SKILL.md) 了解认证、全局参数和安全规则。
 
-将本地文件（如 Word、TXT、Markdown、Excel、PPTX、PDF 等）导入并转换为飞书在线云文档（docx、sheet、bitable、slides）。底层统一通过 `POST /open-apis/drive/v1/import_tasks` 接口创建导入任务，并在 shortcut 内做有限次数轮询 `GET /open-apis/drive/v1/import_tasks/:ticket`。
+将本地文件（如 Word、TXT、Markdown、Excel、PPTX 等）导入并转换为飞书在线云文档（docx、sheet、bitable、slides）。底层统一通过 `POST /open-apis/drive/v1/import_tasks` 接口创建导入任务，并在 shortcut 内做有限次数轮询 `GET /open-apis/drive/v1/import_tasks/:ticket`。
 
 > [!IMPORTANT]
 > 当用户说“把本地 Excel / CSV / `.base` 快照导入成 Base / 多维表格 / bitable 文档”时，第一步必须使用 `drive +import --type bitable`。
@@ -45,9 +45,8 @@ lark-cli drive +import --file ./crm.xlsx --type bitable --name "客户台账"
 # 导入 .base 快照为多维表格 / Base (bitable)（文件不能超过 20MB）
 lark-cli drive +import --file ./snapshot.base --type bitable --name "快照还原"
 
-# 导入 PPTX / PDF 为飞书幻灯片 (slides)（文件不能超过 500MB；PDF 可按每页约 10 秒预估导入处理时间）
+# 导入 PPTX 为飞书幻灯片 (slides)（文件不能超过 500MB）
 lark-cli drive +import --file ./deck.pptx --type slides --name "项目汇报"
-lark-cli drive +import --file ./deck.pdf --type slides --name "项目汇报"
 
 # 导入到指定文件夹，并指定导入后的文件名
 lark-cli drive +import --file ./data.csv --type bitable --folder-token <FOLDER_TOKEN> --name "导入数据表"
@@ -79,7 +78,6 @@ lark-cli drive +import --file ./README.md --type docx --dry-run
   3. 自动轮询查询导入任务状态；如果在内置轮询窗口内完成，则直接返回导入结果；如果仍未完成，则返回 `ticket`、当前状态和后续查询命令
 - **默认根目录行为**：不传 `--folder-token` 时，shortcut 会保留空的 `point.mount_key`，Lark Import API 会将其视为"导入到调用者根目录"。
 - **导入到已有 bitable**：当 `--type bitable` 且传了 `--target-token` 时，请求 body 中会增加一个 `token` 字段指向目标多维表格的 token，point 挂载点逻辑不变。数据会挂载到该已有多维表格中，而非创建新文档。
-- **PDF 导入 Slides 耗时预估**：PDF 导入 Slides 是长耗时异步任务，通常按每页约 10 秒估算总处理时间。`drive +import` 内置轮询超时只表示本地等待窗口结束，不代表任务失败。看到 `ready=false` / `timed_out=true` 时，必须继续执行返回的 `next_command` 查询同一个 `ticket`，不要重试导入、不要提前放弃；至少等待到 `页数 * 10 秒` 的预估处理时间后，再结合任务状态判断是否异常。
 
 ### 支持的文件类型转换
 
@@ -96,7 +94,6 @@ lark-cli drive +import --file ./README.md --type docx --dry-run
 | `.csv` | `sheet`, `bitable` | CSV 数据文件 |
 | `.base` | `bitable` | 多维表格快照文件 |
 | `.pptx` | `slides` | Microsoft PowerPoint 演示文稿 |
-| `.pdf` | `slides` | PDF 文档 |
 
 > [!IMPORTANT]
 > 用户口头说的 “Base” / “多维表格” / “bitable”，在命令里统一对应 `--type bitable`。
@@ -106,7 +103,7 @@ lark-cli drive +import --file ./README.md --type docx --dry-run
 > - `.xlsx` / `.csv` 文件**只能**导入为 `sheet` 或 `bitable`
 > - `.xls` 文件**只能**导入为 `sheet`
 > - `.base` 文件**只能**导入为 `bitable`
-> - `.pptx` / `.pdf` 文件**只能**导入为 `slides`
+> - `.pptx` 文件**只能**导入为 `slides`
 > - 例如：`.csv` 文件不能导入为 `docx`，`.md` 文件不能导入为 `sheet`
 
 > [!IMPORTANT]
@@ -140,7 +137,7 @@ lark-cli drive +import --file ./README.md --type docx --dry-run
 | `.csv` | `bitable` | 100MB |
 | `.xls` | `sheet` | 20MB |
 | `.base` | `bitable` | 20MB |
-| `.pptx`, `.pdf` | `slides` | 500MB |
+| `.pptx` | `slides` | 500MB |
 
 - 如果文件超出对应上限，shortcut 会在真正上传前直接返回验证错误。
 - “超过 20MB 自动切换分片上传”只表示上传链路会切到 multipart，不代表所有格式都允许导入超过 20MB 的文件。

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.

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