docs: remove write-then-readback constraint from lark-calendar skill

写操作后不再要求等待2秒回查验证，直接信任API返回结果，
减少不必要的耗时。

Change-Id: I2d7e1ecab4450839f8232e65d43bca39e851ae22

.golangci.yml

@@ -29,11 +29,11 @@ linters:
     - unused          # checks for unused constants, variables, functions and types
     - depguard        # blocks forbidden package imports
     - forbidigo       # forbids specific function calls
-    - errorlint       # enforces error wrapping (%w) and errors.Is/As over == and type asserts
 
     # To enable later after fixing existing issues:
     # - errcheck      # checks for unchecked errors
     # - errname       # checks that error types are named XxxError
+    # - errorlint     # checks error wrapping best practices
     # - gosec         # security-oriented linter
     # - misspell      # finds commonly misspelled English words
     # - staticcheck   # comprehensive static analysis
@@ -49,16 +49,9 @@ linters:
           - gocritic
           - depguard
           - forbidigo
-          - errorlint # tests legitimately do identity (==) and concrete type-assert checks
-      # forbidigo runs repo-wide (minus the boundaries below) so errs-no-bare-wrap
-      # has no gap. The framework bans (os/vfs, raw HTTP, fmt.Print, filepath,
-      # log) stay scoped to shortcuts/ + internal/ + config/auth/service via the
-      # next rule; elsewhere only errs-no-bare-wrap fires.
-      - path-except: (shortcuts/|internal/|cmd/|events/)
-        linters:
-          - forbidigo
+      # Paths that run forbidigo. Add an entry when a path joins one of
+      # the rules below.
       - path-except: (shortcuts/|internal/|cmd/auth/|cmd/config/|cmd/service/)
-        text: (vfs|IOStreams|ctx\.Out|shortcuts-no-raw-http|filepath functions|os\.Exit|structured error return)
         linters:
           - forbidigo
       - path: internal/vfs/
@@ -84,14 +77,25 @@ linters:
         text: shortcuts-no-raw-http
         linters:
           - forbidigo
-      # errs-no-bare-wrap enforced across every command/wire boundary by
-      # structural prefix, so any future business domain or command is covered
-      # without editing an allowlist. Genuine intermediate wraps inside these
-      # paths use //nolint:forbidigo with a reason.
-      - path-except: (cmd/|shortcuts/|events/)
+      # errs-typed-only enforced on paths already migrated to errs.NewXxxError.
+      # Add a path when its migration is complete.
+      - path-except: (internal/auth/|internal/errcompat/|internal/errclass/|internal/client/|internal/cmdutil/factory\.go|cmd/auth/|cmd/config/|cmd/service/|shortcuts/common/mcp_client\.go|shortcuts/apps/|shortcuts/base/|shortcuts/calendar/|shortcuts/contact/|shortcuts/doc/|shortcuts/drive/|shortcuts/im/|shortcuts/mail/|shortcuts/markdown/|shortcuts/minutes/|shortcuts/okr/|shortcuts/sheets/|shortcuts/slides/|shortcuts/task/|shortcuts/vc/|shortcuts/whiteboard/|shortcuts/wiki/|internal/event/consume/|cmd/event/|events/|shortcuts/event/)
+        text: errs-typed-only
+        linters:
+          - forbidigo
+      # errs-no-bare-wrap enforced on paths fully migrated to typed final
+      # errors. Scoped separately from errs-typed-only because cmd/auth/,
+      # cmd/config/ still have residual fmt.Errorf and must not be caught.
+      - path-except: (shortcuts/apps/|shortcuts/base/|shortcuts/calendar/|shortcuts/contact/|shortcuts/doc/|shortcuts/drive/|shortcuts/im/|shortcuts/mail/|shortcuts/markdown/|shortcuts/minutes/|shortcuts/okr/|shortcuts/sheets/|shortcuts/slides/|shortcuts/task/|shortcuts/vc/|shortcuts/whiteboard/|shortcuts/wiki/|shortcuts/common/mcp_client\.go|cmd/event/|events/|shortcuts/event/)
         text: errs-no-bare-wrap
         linters:
           - forbidigo
+      # errs-no-legacy-helper enforced on domains whose shared validation/save
+      # helpers have migrated to typed final errors.
+      - path-except: (shortcuts/apps/|shortcuts/base/|shortcuts/calendar/|shortcuts/contact/|shortcuts/doc/|shortcuts/drive/|shortcuts/im/|shortcuts/mail/|shortcuts/markdown/|shortcuts/minutes/|shortcuts/okr/|shortcuts/sheets/|shortcuts/slides/|shortcuts/task/|shortcuts/vc/|shortcuts/whiteboard/|shortcuts/wiki/|cmd/event/|events/|shortcuts/event/)
+        text: errs-no-legacy-helper
+        linters:
+          - forbidigo
 
   settings:
     depguard:
@@ -110,6 +114,22 @@ linters:
                 Use runtime.FileIO() for file operations or runtime.ValidatePath() for path validation.
     forbidigo:
       forbid:
+        # ── legacy output.Err* helpers banned on migrated paths ──
+        # output.ErrBare is intentionally not listed — it is the predicate-
+        # command silent-exit signal, outside the typed envelope contract.
+        - pattern: output\.(ErrValidation|ErrAuth|ErrNetwork|ErrAPI|ErrWithHint|Errorf)\b
+          msg: >-
+            [errs-typed-only] use errs.NewXxxError(...) builder
+            (see errs/types.go).
+        # ── legacy shared error helpers banned on migrated domains ──
+        # These helpers emit legacy output.Err* / bare error shapes or drop
+        # typed metadata such as Param/Cause. Migrated domains must use typed
+        # common replacements or local typed helpers instead.
+        - pattern: (common\.FlagErrorf|common\.RejectDangerousChars|common\.WrapInputStatError|common\.WrapSaveErrorByCategory)\b
+          msg: >-
+            [errs-no-legacy-helper] these shared helpers emit legacy or
+            metadata-poor error shapes. Use typed common replacements, typed
+            errs.NewXxxError builders, or domain-local typed helpers.
         # ── bare error wraps banned on fully-typed paths ──
         - pattern: (fmt\.Errorf|errors\.New)\b
           msg: >-

CHANGELOG.md

@@ -2,29 +2,6 @@
 
 All notable changes to this project will be documented in this file.
 
-## [v1.0.56] - 2026-06-18
-
-### Features
-
-- **apps**: Add `+session-messages-list` for session turn reply messages (#1402)
-
-### Bug Fixes
-
-- **api**: Align API success envelopes (#1489)
-- **base**: Reject out-of-range pagination flags (#1495)
-
-### Refactor
-
-- Retire legacy error envelopes and enforce typed contract (#1449)
-
-### Documentation
-
-- **skills**: Soften lark-doc style guidance (#1463)
-
-### Build
-
-- Add CI quality gate with semantic review
-
 ## [v1.0.55] - 2026-06-16
 
 ### Features
@@ -1212,7 +1189,6 @@ Bundled AI agent skills for intelligent assistance:
 - Bilingual documentation (English & Chinese).
 - CI/CD pipelines: linting, testing, coverage reporting, and automated releases.
 
-[v1.0.56]: https://github.com/larksuite/cli/releases/tag/v1.0.56
 [v1.0.55]: https://github.com/larksuite/cli/releases/tag/v1.0.55
 [v1.0.54]: https://github.com/larksuite/cli/releases/tag/v1.0.54
 [v1.0.53]: https://github.com/larksuite/cli/releases/tag/v1.0.53

cmd/api/api.go

@@ -10,7 +10,6 @@ import (
 	"regexp"
 	"strings"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/client"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -124,13 +123,7 @@ func buildAPIRequest(opts *APIOptions) (client.RawApiRequest, *cmdutil.FileUploa
 
 	// stdin conflict: --params and --data cannot both read from stdin, regardless of --file.
 	if opts.Params == "-" && opts.Data == "-" {
-		return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
-			"--params and --data cannot both read from stdin (-)").
-			WithHint("pass at most one flag as '-'; give the other inline JSON or @file").
-			WithParams(
-				errs.InvalidParam{Name: "--params", Reason: "reads from stdin (-)"},
-				errs.InvalidParam{Name: "--data", Reason: "reads from stdin (-)"},
-			)
+		return client.RawApiRequest{}, nil, output.ErrValidation("--params and --data cannot both read from stdin (-)")
 	}
 
 	params, err := cmdutil.ParseJSONMap(opts.Params, "--params", stdin, fileIO)
@@ -160,10 +153,7 @@ func buildAPIRequest(opts *APIOptions) (client.RawApiRequest, *cmdutil.FileUploa
 				return client.RawApiRequest{}, nil, err
 			}
 			if _, ok := dataFields.(map[string]any); !ok {
-				return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
-					"--data must be a JSON object when used with --file").
-					WithHint(`with --file, --data carries multipart form fields, e.g. --data '{"image_type":"message"}'`).
-					WithParam("--data")
+				return client.RawApiRequest{}, nil, output.ErrValidation("--data must be a JSON object when used with --file")
 			}
 		}
 
@@ -206,13 +196,7 @@ func apiRun(opts *APIOptions) error {
 	}
 
 	if opts.PageAll && opts.Output != "" {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument,
-			"--output and --page-all are mutually exclusive").
-			WithHint("drop --page-all to save a binary response, or drop --output to paginate JSON").
-			WithParams(
-				errs.InvalidParam{Name: "--output", Reason: "conflicts with --page-all"},
-				errs.InvalidParam{Name: "--page-all", Reason: "conflicts with --output"},
-			)
+		return output.ErrValidation("--output and --page-all are mutually exclusive")
 	}
 	if err := output.ValidateJqFlags(opts.JqExpr, opts.Output, opts.Format); err != nil {
 		return err
@@ -259,7 +243,7 @@ func apiRun(opts *APIOptions) error {
 		// pass on *output.ExitError values. Typed *errs.* errors that flow
 		// through here keep their canonical message / hint from BuildAPIError;
 		// MarkRaw is a no-op on those (it only flips a flag on *ExitError).
-		return errs.MarkRaw(err)
+		return output.MarkRaw(err)
 	}
 	err = client.HandleResponse(resp, client.ResponseOptions{
 		OutputPath:  opts.Output,
@@ -279,7 +263,7 @@ func apiRun(opts *APIOptions) error {
 	// MarkRaw: see comment above on the DoAPI path. Skips legacy
 	// *ExitError enrichment; typed errors flow through unchanged.
 	if err != nil {
-		return errs.MarkRaw(err)
+		return output.MarkRaw(err)
 	}
 	return nil
 }
@@ -296,11 +280,11 @@ func apiPaginate(ctx context.Context, ac *client.APIClient, request client.RawAp
 	if jqExpr != "" {
 		result, err := ac.PaginateAll(ctx, request, pagOpts)
 		if err != nil {
-			return errs.MarkRaw(err)
+			return output.MarkRaw(err)
 		}
 		if apiErr := ac.CheckResponse(result, pagOpts.Identity); apiErr != nil {
 			output.FormatValue(out, result, output.FormatJSON)
-			return errs.MarkRaw(apiErr)
+			return output.MarkRaw(apiErr)
 		}
 		return output.WriteSuccessEnvelope(output.SuccessEnvelopeData(result), output.SuccessEnvelopeOptions{
 			CommandPath: commandPath,
@@ -329,10 +313,10 @@ func apiPaginate(ctx context.Context, ac *client.APIClient, request client.RawAp
 			return nil
 		}, pagOpts)
 		if err != nil {
-			return errs.MarkRaw(err)
+			return output.MarkRaw(err)
 		}
 		if apiErr := ac.CheckResponse(result, pagOpts.Identity); apiErr != nil {
-			return errs.MarkRaw(apiErr)
+			return output.MarkRaw(apiErr)
 		}
 		if !hasItems {
 			fmt.Fprintf(errOut, "warning: this API does not return a list, format %q is not supported, falling back to json\n", format)
@@ -347,11 +331,11 @@ func apiPaginate(ctx context.Context, ac *client.APIClient, request client.RawAp
 	default:
 		result, err := ac.PaginateAll(ctx, request, pagOpts)
 		if err != nil {
-			return errs.MarkRaw(err)
+			return output.MarkRaw(err)
 		}
 		if apiErr := ac.CheckResponse(result, pagOpts.Identity); apiErr != nil {
 			output.FormatValue(out, result, output.FormatJSON)
-			return errs.MarkRaw(apiErr)
+			return output.MarkRaw(apiErr)
 		}
 		return output.WriteSuccessEnvelope(output.SuccessEnvelopeData(result), output.SuccessEnvelopeOptions{
 			CommandPath: commandPath,

cmd/auth/check_test.go

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

cmd/auth/list.go

@@ -9,7 +9,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	larkauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -60,7 +59,7 @@ func authListRun(opts *ListOptions) error {
 		// keep the same contract here. We still want the hint to be
 		// workspace-aware, so we pull the message+hint out of
 		// NotConfiguredError() instead of hard-coding it.
-		var cfgErr *errs.ConfigError
+		var cfgErr *core.ConfigError
 		if errors.As(core.NotConfiguredError(), &cfgErr) {
 			fmt.Fprintln(f.IOStreams.ErrOut, cfgErr.Message)
 			if cfgErr.Hint != "" {

cmd/auth/login_test.go

@@ -878,7 +878,7 @@ func TestAuthLoginRun_DeviceCodeTokenNilCleansScopeCache(t *testing.T) {
 // contract that when --json is set and pollDeviceToken returns OK=false,
 // stdout carries the structured authorization_failed event and stderr is
 // NOT polluted with a typed envelope. The returned error is a bare
-// BareError with ExitAuth so the dispatcher only propagates the exit code
+// ExitError with ExitAuth so the dispatcher only propagates the exit code
 // without emitting a second envelope on top of the JSON event.
 func TestAuthLoginRun_JSONAbort_StdoutEventOnly_StderrEmpty(t *testing.T) {
 	keyring.MockInit()
@@ -945,13 +945,16 @@ func TestAuthLoginRun_JSONAbort_StdoutEventOnly_StderrEmpty(t *testing.T) {
 		t.Errorf("stderr should not contain JSON envelope fields, got: %s", stderrStr)
 	}
 
-	// Returned error must be the bare *output.BareError signal (no envelope).
-	var bareErr *output.BareError
-	if !errors.As(err, &bareErr) {
-		t.Fatalf("expected *output.BareError, got %T: %v", err, err)
+	// Returned error must be the bare *output.ExitError signal (no envelope).
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
 	}
-	if bareErr.Code != output.ExitAuth {
-		t.Fatalf("BareError.Code = %d, want %d", bareErr.Code, output.ExitAuth)
+	if exitErr.Code != output.ExitAuth {
+		t.Fatalf("ExitError.Code = %d, want %d", exitErr.Code, output.ExitAuth)
+	}
+	if exitErr.Detail != nil {
+		t.Errorf("ExitError.Detail should be nil for bare signal, got: %+v", exitErr.Detail)
 	}
 }
 

cmd/completion/completion.go

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

cmd/config/bind.go

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

cmd/config/bind_test.go

@@ -20,29 +20,35 @@ import (
 	"github.com/larksuite/cli/internal/output"
 )
 
-// wantErrDetail is the normalized comparison shape for a typed error's wire
-// fields: Type is the error's Category string ("validation", "config", ...),
-// alongside Message and Hint.
-type wantErrDetail struct {
-	Type    string
-	Message string
-	Hint    string
-}
-
-// assertExitError checks the full structured error in one assertion against a
-// typed error (ValidationError or ConfigError), normalizing its Category /
-// Message / Hint to wantDetail.
-func assertExitError(t *testing.T, err error, wantCode int, wantDetail wantErrDetail) {
+// assertExitError checks the full structured error in one assertion. It
+// accepts both *output.ExitError (used by output.ErrWithHint) and the
+// typed errors (ValidationError, ConfigError) — they normalize to the same
+// wantDetail fields. The wantDetail.Type is matched against the typed error's
+// Category string ("validation", "config", etc.).
+func assertExitError(t *testing.T, err error, wantCode int, wantDetail output.ErrDetail) {
 	t.Helper()
 	if err == nil {
 		t.Fatal("expected error, got nil")
 	}
+	var exitErr *output.ExitError
+	if errors.As(err, &exitErr) {
+		if exitErr.Code != wantCode {
+			t.Errorf("exit code = %d, want %d", exitErr.Code, wantCode)
+		}
+		if exitErr.Detail == nil {
+			t.Fatal("expected non-nil error detail")
+		}
+		if !reflect.DeepEqual(*exitErr.Detail, wantDetail) {
+			t.Errorf("error detail mismatch:\n  got:  %+v\n  want: %+v", *exitErr.Detail, wantDetail)
+		}
+		return
+	}
 	var ve *errs.ValidationError
 	if errors.As(err, &ve) {
 		if got := output.ExitCodeOf(err); got != wantCode {
 			t.Errorf("exit code = %d, want %d", got, wantCode)
 		}
-		gotDetail := wantErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
+		gotDetail := output.ErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
 		if !reflect.DeepEqual(gotDetail, wantDetail) {
 			t.Errorf("validation error mismatch:\n  got:  %+v\n  want: %+v", gotDetail, wantDetail)
 		}
@@ -53,13 +59,13 @@ func assertExitError(t *testing.T, err error, wantCode int, wantDetail wantErrDe
 		if got := output.ExitCodeOf(err); got != wantCode {
 			t.Errorf("exit code = %d, want %d", got, wantCode)
 		}
-		gotDetail := wantErrDetail{Type: string(ce.Category), Message: ce.Message, Hint: ce.Hint}
+		gotDetail := output.ErrDetail{Type: string(ce.Category), Message: ce.Message, Hint: ce.Hint}
 		if !reflect.DeepEqual(gotDetail, wantDetail) {
 			t.Errorf("config error mismatch:\n  got:  %+v\n  want: %+v", gotDetail, wantDetail)
 		}
 		return
 	}
-	t.Fatalf("error type = %T, want *errs.ValidationError / *errs.ConfigError; error = %v", err, err)
+	t.Fatalf("error type = %T, want *output.ExitError or *errs.ValidationError / *errs.ConfigError; error = %v", err, err)
 }
 
 // assertEnvelope decodes stdout and checks it matches want exactly — every key
@@ -173,21 +179,15 @@ func TestConfigBindRun_InvalidLang(t *testing.T) {
 			if err == nil {
 				t.Fatalf("expected validation error for --lang %q, got nil", tc.lang)
 			}
-			var valErr *errs.ValidationError
-			if !errors.As(err, &valErr) {
-				t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
+			exitErr, ok := err.(*output.ExitError)
+			if !ok {
+				t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
 			}
-			if valErr.Subtype != errs.SubtypeInvalidArgument {
-				t.Errorf("subtype = %q, want %q", valErr.Subtype, errs.SubtypeInvalidArgument)
+			if exitErr.Code != output.ExitValidation {
+				t.Errorf("exit code = %d, want %d (validation)", exitErr.Code, output.ExitValidation)
 			}
-			if valErr.Param != "--lang" {
-				t.Errorf("param = %q, want %q", valErr.Param, "--lang")
-			}
-			if got := output.ExitCodeOf(err); got != output.ExitValidation {
-				t.Errorf("exit code = %d, want %d (validation)", got, output.ExitValidation)
-			}
-			if !strings.Contains(err.Error(), "invalid --lang") {
-				t.Errorf("error message %q does not contain 'invalid --lang'", err.Error())
+			if !strings.Contains(exitErr.Error(), "invalid --lang") {
+				t.Errorf("error message %q does not contain 'invalid --lang'", exitErr.Error())
 			}
 		})
 	}
@@ -365,7 +365,7 @@ func TestConfigBindRun_InvalidSource(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "invalid"})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: `invalid --source "invalid"; valid values: openclaw, hermes, lark-channel`,
 	})
@@ -382,7 +382,7 @@ func TestConfigBindRun_MissingSourceNonTTY(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	// TestFactory has IsTerminal=false by default
 	err := configBindRun(&BindOptions{Factory: f, Source: ""})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: "cannot determine Agent source: no --source flag and no Agent environment detected",
 		Hint:    "pass --source openclaw|hermes|lark-channel, or run this command inside the corresponding Agent context",
@@ -421,7 +421,7 @@ func TestConfigBindRun_SourceEnvMismatch_OpenClawFlagInHermesEnv(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: `--source "openclaw" does not match detected Agent environment (hermes)`,
 		Hint:    "remove --source to auto-detect, or run this command in the correct Agent context",
@@ -437,7 +437,7 @@ func TestConfigBindRun_SourceEnvMismatch_HermesFlagInOpenClawEnv(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: `--source "hermes" does not match detected Agent environment (openclaw)`,
 		Hint:    "remove --source to auto-detect, or run this command in the correct Agent context",
@@ -566,7 +566,7 @@ func TestConfigBindRun_HermesMissingEnvFile(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
 	envPath := filepath.Join(hermesHome, ".env")
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "failed to read Hermes config: open " + envPath + ": no such file or directory",
 		Hint:    "verify Hermes is installed and configured at " + envPath,
@@ -584,7 +584,7 @@ func TestConfigBindRun_OpenClawMissingFile(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
 	configPath := filepath.Join(openclawHome, ".openclaw", "openclaw.json")
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "cannot read " + configPath + ": open " + configPath + ": no such file or directory",
 		Hint:    "verify OpenClaw is installed and configured",
@@ -731,7 +731,7 @@ func TestConfigBindRun_SourceEnvMismatch_LarkChannelFlagInOpenClawEnv(t *testing
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: `--source "lark-channel" does not match detected Agent environment (openclaw)`,
 		Hint:    "remove --source to auto-detect, or run this command in the correct Agent context",
@@ -750,7 +750,7 @@ func TestConfigBindRun_LarkChannelMissingFile(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
 	configPath := filepath.Join(fakeHome, ".lark-channel", "config.json")
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "cannot read " + configPath + ": open " + configPath + ": no such file or directory",
 		Hint:    "verify lark-channel-bridge is installed and configured",
@@ -770,7 +770,7 @@ func TestConfigBindRun_LarkChannelEmptyAppID(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "accounts.app.id missing in " + configPath,
 		Hint:    "run lark-channel-bridge's setup to populate the app credential",
@@ -789,7 +789,7 @@ func TestConfigBindRun_LarkChannelEmptySecret(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "lark-channel"})
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "accounts.app.secret is empty in " + configPath,
 		Hint:    "run lark-channel-bridge's setup to populate the app credential",
@@ -835,19 +835,17 @@ func TestConfigShowRun_AgentWorkspaceNotBound(t *testing.T) {
 		t.Fatal("expected error for unbound workspace")
 	}
 	// Should be a structured ConfigError suggesting config bind, not config init.
-	var cfgErr *errs.ConfigError
+	var cfgErr *core.ConfigError
 	if !errors.As(err, &cfgErr) {
-		t.Fatalf("error type = %T, want *errs.ConfigError", err)
+		t.Fatalf("error type = %T, want *core.ConfigError", err)
 	}
 	// Config errors share ExitAuth (3); the workspace is detected but no
 	// binding exists yet, which is a config error.
-	if got := output.ExitCodeOf(err); got != output.ExitAuth {
-		t.Errorf("exit code = %d, want %d (config category → ExitAuth)", got, output.ExitAuth)
+	if cfgErr.Code != output.ExitAuth {
+		t.Errorf("exit code = %d, want %d (config category → ExitAuth)", cfgErr.Code, output.ExitAuth)
 	}
-	// The workspace name stays out of the wire subtype; it only appears in
-	// the message.
-	if cfgErr.Subtype != errs.SubtypeNotConfigured {
-		t.Errorf("subtype = %q, want not_configured", cfgErr.Subtype)
+	if cfgErr.Type != "openclaw" {
+		t.Errorf("type = %q, want %q", cfgErr.Type, "openclaw")
 	}
 	if !strings.Contains(cfgErr.Message, "openclaw context detected") {
 		t.Errorf("message missing 'openclaw context detected': %q", cfgErr.Message)
@@ -1189,7 +1187,7 @@ func TestConfigBindRun_OpenClawMultiAccount_TTYFlagMode(t *testing.T) {
 	// iterates a map — ordering is non-deterministic. DeepEqual inline against
 	// each accepted variant so every ErrDetail field (Type, Code, Message,
 	// Hint, ConsoleURL, Detail, and any future addition) is still compared.
-	base := wantErrDetail{
+	base := output.ErrDetail{
 		Type:    "validation",
 		Message: "multiple accounts in openclaw.json; pass --app-id <id>",
 	}
@@ -1205,7 +1203,7 @@ func TestConfigBindRun_OpenClawMultiAccount_TTYFlagMode(t *testing.T) {
 	if !errors.As(err, &ve) {
 		t.Fatalf("error type = %T, want *errs.ValidationError; err = %v", err, err)
 	}
-	got := wantErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
+	got := output.ErrDetail{Type: string(ve.Category), Message: ve.Message, Hint: ve.Hint}
 	if !reflect.DeepEqual(got, wantWorkFirst) && !reflect.DeepEqual(got, wantPersonalFirst) {
 		t.Errorf("error detail did not match any accepted variant:\n  got:  %+v\n  want: %+v OR %+v",
 			got, wantWorkFirst, wantPersonalFirst)
@@ -1232,7 +1230,7 @@ func TestConfigBindRun_OpenClawMultiAccount_WrongAppID(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw", AppID: "nonexistent"})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: `--app-id "nonexistent" not found in openclaw.json`,
 		Hint:    "available app IDs:\n  cli_only_one",
@@ -1252,7 +1250,7 @@ func TestConfigBindRun_InvalidIdentity(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes", Identity: "invalid"})
-	assertExitError(t, err, output.ExitValidation, wantErrDetail{
+	assertExitError(t, err, output.ExitValidation, output.ErrDetail{
 		Type:    "validation",
 		Message: `invalid --identity "invalid"; valid values: bot-only, user-default`,
 	})
@@ -1538,7 +1536,7 @@ func TestConfigBindRun_HermesMissingAppID(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
 	envPath := filepath.Join(hermesHome, ".env")
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "FEISHU_APP_ID not found in " + envPath,
 		Hint:    "run 'hermes setup' to configure Feishu credentials",
@@ -1558,7 +1556,7 @@ func TestConfigBindRun_HermesMissingAppSecret(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "hermes"})
 	envPath := filepath.Join(hermesHome, ".env")
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "FEISHU_APP_SECRET not found in " + envPath,
 		Hint:    "run 'hermes setup' to configure Feishu credentials",
@@ -1584,7 +1582,7 @@ func TestConfigBindRun_OpenClawMissingFeishu(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "openclaw.json missing channels.feishu section",
 		Hint:    "configure Feishu in OpenClaw first",
@@ -1612,7 +1610,7 @@ func TestConfigBindRun_OpenClawEmptyAppSecret(t *testing.T) {
 	openclawPath := filepath.Join(openclawDir, "openclaw.json")
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "appSecret is empty for app cli_no_secret in " + openclawPath,
 		Hint:    "configure channels.feishu.appSecret in openclaw.json",
@@ -1674,7 +1672,7 @@ func TestConfigBindRun_OpenClawDisabledAccount(t *testing.T) {
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
 	err := configBindRun(&BindOptions{Factory: f, Source: "openclaw"})
-	assertExitError(t, err, output.ExitAuth, wantErrDetail{
+	assertExitError(t, err, output.ExitAuth, output.ErrDetail{
 		Type:    "config",
 		Message: "no Feishu app configured in openclaw.json",
 		Hint:    "configure channels.feishu.appId in openclaw.json",

cmd/config/binder_test.go

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

cmd/config/config_test.go

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

cmd/config/init.go

@@ -6,11 +6,13 @@ package config
 import (
 	"bufio"
 	"context"
+	"errors"
 	"fmt"
 	"io"
 	"os"
 	"strings"
 
+	"github.com/charmbracelet/huh"
 	"github.com/spf13/cobra"
 
 	"github.com/larksuite/cli/errs"
@@ -125,9 +127,12 @@ func guardAgentWorkspace(opts *ConfigInitOptions) error {
 	if ws.IsLocal() {
 		return nil
 	}
-	return errs.NewConfigError(errs.SubtypeNotConfigured,
-		"config init is refused inside %s context (would create a parallel app and shadow the existing %s binding)", ws.Display(), ws.Display()).
-		WithHint("see `lark-cli config bind --help` to bind lark-cli to the Agent's existing app instead. Pass --force-init only if the user explicitly wants a separate app in this workspace.")
+	return &core.ConfigError{
+		Code:    2,
+		Type:    ws.Display(),
+		Message: fmt.Sprintf("config init is refused inside %s context (would create a parallel app and shadow the existing %s binding)", ws.Display(), ws.Display()),
+		Hint:    "see `lark-cli config bind --help` to bind lark-cli to the Agent's existing app instead. Pass --force-init only if the user explicitly wants a separate app in this workspace.",
+	}
 }
 
 // hasAnyNonInteractiveFlag returns true if any non-interactive flag is set.
@@ -178,20 +183,6 @@ func saveInitConfig(profileName string, existing *core.MultiAppConfig, f *cmduti
 	return saveAsOnlyApp(appId, secret, brand, string(preferredLang(i18n.Lang(lang), prior)))
 }
 
-// wrapSaveConfigError passes an already-typed error (e.g. the --name conflict
-// validation error from saveAsProfile) through unchanged, and classifies any
-// other failure as an internal storage error. Without the passthrough a user
-// input error would surface to agents as a system storage failure.
-func wrapSaveConfigError(err error) error {
-	if err == nil {
-		return nil
-	}
-	if _, ok := errs.ProblemOf(err); ok {
-		return err
-	}
-	return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
-}
-
 // saveAsProfile appends or updates a named profile in the config.
 // If a profile with the same name exists, it updates it; otherwise appends.
 // When updating, cleans up old keychain secrets if AppId changed.
@@ -216,9 +207,7 @@ func saveAsProfile(existing *core.MultiAppConfig, kc keychain.KeychainAccess, pr
 		multi.Apps[idx].Lang = preferredLang(i18n.Lang(lang), multi.Apps[idx].Lang)
 	} else {
 		if findAppIndexByAppID(multi, profileName) >= 0 {
-			return errs.NewValidationError(errs.SubtypeInvalidArgument,
-				"profile name %q conflicts with existing appId", profileName).
-				WithParam("--name")
+			return fmt.Errorf("profile name %q conflicts with existing appId", profileName)
 		}
 		// Append new profile
 		multi.Apps = append(multi.Apps, core.AppConfig{
@@ -260,8 +249,8 @@ func findAppIndexByAppID(multi *core.MultiAppConfig, appID string) int {
 // wrapUpdateExistingProfileErr classifies the error returned by
 // updateExistingProfileWithoutSecret. Typed errors (e.g. *errs.ValidationError
 // for blank-input) pass through unchanged so their exit code semantics
-// survive; everything else (filesystem, keychain, etc.) is wrapped as
-// InternalError.
+// survive; legacy *output.ExitError also passes through; everything else
+// (filesystem, keychain, etc.) is wrapped as InternalError.
 func wrapUpdateExistingProfileErr(err error) error {
 	if err == nil {
 		return nil
@@ -269,6 +258,10 @@ func wrapUpdateExistingProfileErr(err error) error {
 	if errs.IsTyped(err) {
 		return err
 	}
+	var exitErr *output.ExitError
+	if errors.As(err, &exitErr) {
+		return err
+	}
 	return errs.NewInternalError(errs.SubtypeSDKError, "failed to save config: %v", err).WithCause(err)
 }
 
@@ -343,7 +336,7 @@ func configInitRun(opts *ConfigInitOptions) error {
 			return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 		}
 		if err := saveInitConfig(opts.ProfileName, existing, f, opts.AppID, secret, brand, opts.Lang); err != nil {
-			return wrapSaveConfigError(err)
+			return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
 		}
 		output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Configuration saved to %s", core.GetConfigPath()))
 		printLangPreferenceConfirmation(opts)
@@ -360,7 +353,10 @@ func configInitRun(opts *ConfigInitOptions) error {
 	if f.IOStreams.IsTerminal && !opts.langExplicit && !opts.hasAnyNonInteractiveFlag() {
 		lang, err := promptLangSelection()
 		if err != nil {
-			return langSelectionError(err)
+			if err == huh.ErrUserAborted {
+				return output.ErrBare(1)
+			}
+			return output.Errorf(output.ExitInternal, "internal", "language selection failed: %v", err)
 		}
 		opts.Lang = string(lang)
 		opts.UILang = lang
@@ -383,7 +379,7 @@ func configInitRun(opts *ConfigInitOptions) error {
 			return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 		}
 		if err := saveInitConfig(opts.ProfileName, existing, f, result.AppID, secret, result.Brand, opts.Lang); err != nil {
-			return wrapSaveConfigError(err)
+			return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
 		}
 		printLangPreferenceConfirmation(opts)
 		output.PrintJson(f.IOStreams.Out, map[string]interface{}{"appId": result.AppID, "appSecret": "****", "brand": result.Brand})
@@ -413,7 +409,7 @@ func configInitRun(opts *ConfigInitOptions) error {
 				return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 			}
 			if err := saveInitConfig(opts.ProfileName, existing, f, result.AppID, secret, result.Brand, opts.Lang); err != nil {
-				return wrapSaveConfigError(err)
+				return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
 			}
 		} else if result.Mode == "existing" && result.AppID != "" {
 			// Existing app with unchanged secret — update app ID and brand only
@@ -518,7 +514,7 @@ func configInitRun(opts *ConfigInitOptions) error {
 		return errs.NewInternalError(errs.SubtypeSDKError, "%v", err).WithCause(err)
 	}
 	if err := saveInitConfig(opts.ProfileName, existing, f, resolvedAppId, storedSecret, parseBrand(resolvedBrand), opts.Lang); err != nil {
-		return wrapSaveConfigError(err)
+		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
 	}
 	output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Configuration saved to %s", core.GetConfigPath()))
 	printLangPreferenceConfirmation(opts)

cmd/config/init_guard_test.go

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

cmd/config/init_messages.go

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

cmd/config/init_test.go

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

cmd/doctor/doctor.go

@@ -14,7 +14,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/build"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -95,7 +94,7 @@ func doctorRun(opts *DoctorOptions) error {
 		// underlying problem is still visible.
 		msg, hint := err.Error(), ""
 		if errors.Is(err, os.ErrNotExist) {
-			var cfgErr *errs.ConfigError
+			var cfgErr *core.ConfigError
 			if errors.As(core.NotConfiguredError(), &cfgErr) {
 				msg, hint = cfgErr.Message, cfgErr.Hint
 			}
@@ -109,7 +108,7 @@ func doctorRun(opts *DoctorOptions) error {
 	cfg, err := f.Config()
 	if err != nil {
 		hint := ""
-		var cfgErr *errs.ConfigError
+		var cfgErr *core.ConfigError
 		if errors.As(err, &cfgErr) {
 			hint = cfgErr.Hint
 		}

cmd/error_auth_hint.go

@@ -15,6 +15,7 @@ import (
 	internalauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
 	"github.com/larksuite/cli/internal/registry"
 	"github.com/larksuite/cli/shortcuts"
 	shortcutcommon "github.com/larksuite/cli/shortcuts/common"
@@ -48,6 +49,32 @@ func applyNeedAuthorizationHint(f *cmdutil.Factory, err error) {
 	authErr.Hint += "\n" + scopeHint
 }
 
+// enrichMissingScopeError appends a "current command requires scope(s): X"
+// hint to a legacy *output.ExitError when the underlying error carries the
+// need_user_authorization marker AND the current command declares scopes
+// locally.
+//
+// Deprecated: enrichment for the legacy envelope; the typed path is
+// applyNeedAuthorizationHint above.
+func enrichMissingScopeError(f *cmdutil.Factory, exitErr *output.ExitError) {
+	if exitErr == nil || exitErr.Detail == nil {
+		return
+	}
+	if !internalauth.IsNeedUserAuthorizationError(exitErr) {
+		return
+	}
+	scopes := resolveDeclaredScopesForCurrentCommand(f)
+	if len(scopes) == 0 {
+		return
+	}
+	scopeHint := fmt.Sprintf("current command requires scope(s): %s", strings.Join(scopes, ", "))
+	if exitErr.Detail.Hint == "" {
+		exitErr.Detail.Hint = scopeHint
+		return
+	}
+	exitErr.Detail.Hint += "\n" + scopeHint
+}
+
 // resolveDeclaredScopesForCurrentCommand returns the scopes declared by the
 // current command for the resolved identity, checking shortcuts first and then
 // service methods from local registry metadata.

cmd/event/consume.go

@@ -386,9 +386,9 @@ func resolveTenantToken(ctx context.Context, f *cmdutil.Factory, appID string) (
 
 // Sentinels for errors.Is checks; call sites wrap them as typed ValidationError causes.
 var (
-	errInvalidParamFormat = errors.New("invalid --param format")                    //nolint:forbidigo // sentinel, typed at call sites
-	errOutputDirTilde     = errors.New("--output-dir does not support ~ expansion") //nolint:forbidigo // sentinel, typed at call sites
-	errOutputDirUnsafe    = errors.New("unsafe --output-dir")                       //nolint:forbidigo // sentinel, typed at call sites
+	errInvalidParamFormat = errors.New("invalid --param format")
+	errOutputDirTilde     = errors.New("--output-dir does not support ~ expansion")
+	errOutputDirUnsafe    = errors.New("unsafe --output-dir")
 )
 
 func parseParams(raw []string) (map[string]string, error) {

cmd/event/format_helpers_test.go

@@ -270,15 +270,15 @@ func TestExitForOrphan(t *testing.T) {
 	if err == nil {
 		t.Fatal("flag on + orphan → expected error, got nil")
 	}
-	var exit *output.BareError
+	var exit *output.ExitError
 	if !errorAs(err, &exit) || exit.Code != output.ExitValidation {
 		t.Errorf("exit code = %v, want ExitValidation", err)
 	}
 }
 
 func errorAs(err error, target interface{}) bool {
-	if e, ok := err.(*output.BareError); ok {
-		if t, ok := target.(**output.BareError); ok {
+	if e, ok := err.(*output.ExitError); ok {
+		if t, ok := target.(**output.ExitError); ok {
 			*t = e
 			return true
 		}

cmd/event/status_fail_on_orphan_test.go

@@ -19,12 +19,12 @@ func TestExitForOrphan_Orphan(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected error when failOnOrphan=true and orphan present")
 	}
-	var bareErr *output.BareError
-	if !errors.As(err, &bareErr) {
-		t.Fatalf("expected *output.BareError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	if bareErr.Code != output.ExitValidation {
-		t.Errorf("Code = %d, want %d", bareErr.Code, output.ExitValidation)
+	if exitErr.Code != output.ExitValidation {
+		t.Errorf("Code = %d, want %d", exitErr.Code, output.ExitValidation)
 	}
 }
 

cmd/flag_suggest_test.go

@@ -5,10 +5,10 @@ package cmd
 
 import (
 	"errors"
+	"slices"
 	"strings"
 	"testing"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/output"
 	"github.com/spf13/cobra"
 )
@@ -40,65 +40,31 @@ func TestFlagDidYouMean_UnknownFlagSuggestsAndListsValid(t *testing.T) {
 	c.Flags().Bool("dry-run", false, "")
 
 	err := flagDidYouMean(c, errors.New("unknown flag: --rang")) // typo of --range
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	if verr.Subtype != errs.SubtypeInvalidArgument {
-		t.Errorf("subtype = %q, want invalid_argument", verr.Subtype)
+	if exitErr.Detail.Type != "unknown_flag" {
+		t.Errorf("type = %q, want unknown_flag", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	if !strings.Contains(exitErr.Detail.Hint, "--range") {
+		t.Errorf("hint should suggest --range, got %q", exitErr.Detail.Hint)
 	}
-	// The offending flag is carried structurally on Params (replaces the
-	// legacy detail map) and named in the message.
-	if len(verr.Params) != 1 || verr.Params[0].Name != "--rang" {
-		t.Errorf("Params = %v, want one entry named --rang", verr.Params)
-	}
-	if len(verr.Params) == 1 && verr.Params[0].Reason == "" {
-		t.Error("Params[0].Reason must explain the rejection")
-	}
-	if !strings.Contains(verr.Message, "--rang") {
-		t.Errorf("message should name the offending flag, got %q", verr.Message)
-	}
-	// The ranked candidate rides on the param as a machine-readable suggestion
-	// so an agent can retry without parsing prose.
-	if len(verr.Params) == 1 {
-		found := false
-		for _, s := range verr.Params[0].Suggestions {
-			if s == "--range" {
-				found = true
-			}
-		}
-		if !found {
-			t.Errorf("Params[0].Suggestions should include --range, got %v", verr.Params[0].Suggestions)
-		}
-	}
-	// The same candidate is also carried in the human-facing hint.
-	if !strings.Contains(verr.Hint, "--range") {
-		t.Errorf("hint should suggest --range, got %q", verr.Hint)
+	detail, _ := exitErr.Detail.Detail.(map[string]any)
+	valid, _ := detail["valid_flags"].([]string)
+	if !slices.Contains(valid, "find") || !slices.Contains(valid, "range") {
+		t.Errorf("valid_flags should list find & range, got %v", valid)
 	}
 }
 
 func TestFlagDidYouMean_OtherErrorStaysGeneric(t *testing.T) {
 	c := &cobra.Command{Use: "demo"}
 	err := flagDidYouMean(c, errors.New("flag needs an argument: --find"))
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	// Non-unknown-flag errors stay generic: invalid_argument subtype, no
-	// structured param, generic --help hint (no "did you mean" suggestion).
-	if verr.Subtype != errs.SubtypeInvalidArgument {
-		t.Errorf("subtype = %q, want invalid_argument (non-unknown-flag errors stay generic)", verr.Subtype)
-	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if verr.Param != "" || len(verr.Params) != 0 {
-		t.Errorf("Param=%q Params=%v, want both empty for generic flag error", verr.Param, verr.Params)
-	}
-	if strings.Contains(verr.Hint, "did you mean") {
-		t.Errorf("generic flag error must not produce a did-you-mean hint, got %q", verr.Hint)
+	if exitErr.Detail.Type != "flag_error" {
+		t.Errorf("type = %q, want flag_error (non-unknown-flag errors stay generic)", exitErr.Detail.Type)
 	}
 }

cmd/platform_bootstrap_test.go

@@ -9,12 +9,10 @@ import (
 	"errors"
 	"os"
 	"path/filepath"
-	"strings"
 	"testing"
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/extension/platform"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/cmdutil"
@@ -104,7 +102,7 @@ func findLeaf(t *testing.T, parent *cobra.Command, names ...string) *cobra.Comma
 }
 
 // Happy path: a valid policy.yml denies one specific command. The denied
-// command's RunE returns a typed error envelope; allowed commands are
+// command's RunE returns a typed ExitError envelope; allowed commands are
 // untouched.
 func TestApplyUserPolicyPruning_appliesValidPolicy(t *testing.T) {
 	cfgDir := tmpHome(t)
@@ -129,27 +127,13 @@ max_risk: write
 	if err == nil {
 		t.Fatalf("+delete-doc RunE should return an error")
 	}
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil || exitErr.Detail.Type != "command_denied" {
+		t.Fatalf("expected command_denied ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
-	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	// The denial taxonomy (reason_code, layer, rule) is preserved on the
-	// wrapped *platform.CommandDeniedError cause and folded into the hint.
-	var cd *platform.CommandDeniedError
-	if !errors.As(err, &cd) {
-		t.Fatalf("error chain should expose *platform.CommandDeniedError")
-	}
-	if cd.ReasonCode != "command_denylisted" {
-		t.Errorf("CommandDeniedError.ReasonCode = %q, want command_denylisted", cd.ReasonCode)
-	}
-	if !strings.Contains(verr.Hint, "command_denylisted") {
-		t.Errorf("hint should surface reason_code command_denylisted, got %q", verr.Hint)
+	detail, ok := exitErr.Detail.Detail.(map[string]any)
+	if !ok || detail["reason_code"] != "command_denylisted" {
+		t.Errorf("reason_code = %v, want command_denylisted", detail["reason_code"])
 	}
 
 	// im/+send must be denied (domain not in Allow).

cmd/platform_guards.go

@@ -8,9 +8,9 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/hook"
+	"github.com/larksuite/cli/internal/output"
 	internalplatform "github.com/larksuite/cli/internal/platform"
 )
 
@@ -34,8 +34,16 @@ import (
 // lands directly on their RunE, which now carries the guard.
 //
 // makeErr is called for every guarded dispatch; it must return a fresh
-// typed error each time.
-func installFatalGuard(rootCmd *cobra.Command, makeErr func() error) {
+// *output.ExitError each time (the envelope writer mutates a few fields
+// as it serialises).
+// Deprecated: installFatalGuard accepts a *output.ExitError-producing lambda,
+// which is part of the legacy error surface that predates the typed error
+// contract introduced by errs/. New code MUST NOT add new callers — the
+// platform-extension fatal-guard plumbing will switch to typed errs.* errors
+// when the platform-extension framework migrates. This wrapper is retained
+// only for the existing in-tree call sites; it will be removed once they
+// have moved to the typed surface.
+func installFatalGuard(rootCmd *cobra.Command, makeErr func() *output.ExitError) {
 	// Two cobra subcommands are injected lazily at Execute() time and
 	// would otherwise slip past walkGuard. We pre-register both so
 	// walkGuard catches them.
@@ -72,65 +80,120 @@ func installFatalGuard(rootCmd *cobra.Command, makeErr func() error) {
 }
 
 // installPluginInstallErrorGuard surfaces a FailClosed plugin install
-// failure as a typed validation error (failed_precondition) before any
-// command runs.
+// failure as a structured plugin_install envelope before any command
+// runs.
+// Deprecated: installPluginInstallErrorGuard produces a legacy
+// *output.ExitError via its internal makeErr lambda. New code MUST NOT add
+// such producers — plugin install failures should surface as a typed
+// *errs.XxxError once the platform-extension framework migrates. This
+// helper is retained only while existing call sites are migrated; it will
+// be removed once they have moved to the typed surface.
 func installPluginInstallErrorGuard(rootCmd *cobra.Command, installErr error) {
-	makeErr := func() error {
+	makeErr := func() *output.ExitError {
 		var pi *internalplatform.PluginInstallError
 		if errors.As(installErr, &pi) {
-			return errs.NewValidationError(errs.SubtypeFailedPrecondition, "%s", pi.Error()).
-				WithHint("plugin %q failed to install (reason_code %s); fix or remove the plugin before running commands", pi.PluginName, pi.ReasonCode).
-				WithCause(installErr)
+			return &output.ExitError{
+				Code: output.ExitValidation,
+				Detail: &output.ErrDetail{
+					Type:    "plugin_install",
+					Message: pi.Error(),
+					Detail: map[string]any{
+						"plugin":      pi.PluginName,
+						"reason_code": pi.ReasonCode,
+						"reason":      pi.Reason,
+					},
+				},
+				Err: installErr,
+			}
+		}
+		return &output.ExitError{
+			Code: output.ExitValidation,
+			Detail: &output.ErrDetail{
+				Type:    "plugin_install",
+				Message: installErr.Error(),
+				Detail: map[string]any{
+					"reason_code": internalplatform.ReasonInstallFailed,
+				},
+			},
+			Err: installErr,
 		}
-		return errs.NewValidationError(errs.SubtypeFailedPrecondition, "%s", installErr.Error()).
-			WithHint("a plugin failed to install (reason_code %s); fix or remove the plugin before running commands", internalplatform.ReasonInstallFailed).
-			WithCause(installErr)
 	}
 	installFatalGuard(rootCmd, makeErr)
 }
 
 // installPluginConflictGuard surfaces a Plugin.Restrict() configuration
 // error (single plugin invalid Rule or multiple plugins each contributing
-// Restrict). The hint separates the two failure modes by reason code:
+// Restrict). The design separates the envelope type:
 //
-//   - "invalid_rule"              - single bad rule
-//   - "multiple_restrict_plugins" - multiple Restrict plugins conflict
+//   - "plugin_install" with reason_code "invalid_rule"           - single bad rule
+//   - "plugin_conflict" with reason_code "multiple_restrict_plugins" - multi
 //
 // Either way the CLI must NOT silently continue with a broken policy.
+// Deprecated: installPluginConflictGuard produces a legacy *output.ExitError
+// via its internal makeErr lambda. New code MUST NOT add such producers —
+// plugin conflict failures should surface as a typed *errs.XxxError once the
+// platform-extension framework migrates. This helper is retained only while
+// existing call sites are migrated; it will be removed once they have moved
+// to the typed surface.
 func installPluginConflictGuard(rootCmd *cobra.Command, err error) {
-	makeErr := func() error {
+	makeErr := func() *output.ExitError {
+		envelopeType := "plugin_install"
 		reasonCode := internalplatform.ReasonInvalidRule
 		if errors.Is(err, cmdpolicy.ErrMultipleRestricts) {
+			envelopeType = "plugin_conflict"
 			reasonCode = internalplatform.ReasonMultipleRestricts
 		}
-		return errs.NewValidationError(errs.SubtypeFailedPrecondition, "%s", err.Error()).
-			WithHint("plugin policy configuration is broken (reason_code %s); fix the plugin's Restrict rule or remove the conflicting plugin", reasonCode).
-			WithCause(err)
+		return &output.ExitError{
+			Code: output.ExitValidation,
+			Detail: &output.ErrDetail{
+				Type:    envelopeType,
+				Message: err.Error(),
+				Detail: map[string]any{
+					"reason_code": reasonCode,
+				},
+			},
+			Err: err,
+		}
 	}
 	installFatalGuard(rootCmd, makeErr)
 }
 
 // installPluginLifecycleErrorGuard surfaces a Startup lifecycle handler
-// failure as a typed validation error (failed_precondition). The hint's
-// reason code splits returned-error vs panic so consumers (audit /
-// on-call) can tell the two failure modes apart.
+// failure as a plugin_lifecycle envelope. The reason_code splits
+// returned-error vs panic so consumers (audit / on-call) can tell the
+// two failure modes apart.
+// Deprecated: installPluginLifecycleErrorGuard produces a legacy
+// *output.ExitError via its internal makeErr lambda. New code MUST NOT add
+// such producers — plugin lifecycle failures should surface as a typed
+// *errs.XxxError once the platform-extension framework migrates. This
+// helper is retained only while existing call sites are migrated; it will
+// be removed once they have moved to the typed surface.
 func installPluginLifecycleErrorGuard(rootCmd *cobra.Command, err error) {
-	makeErr := func() error {
+	makeErr := func() *output.ExitError {
 		reasonCode := "lifecycle_failed"
-		hookName := ""
+		detail := map[string]any{
+			"reason_code": reasonCode,
+		}
 		var le *hook.LifecycleError
 		if errors.As(err, &le) {
 			if le.Panic {
 				reasonCode = "lifecycle_panic"
 			}
-			hookName = le.HookName
+			detail = map[string]any{
+				"reason_code": reasonCode,
+				"hook_name":   le.HookName,
+				"event":       "startup",
+			}
 		}
-		typed := errs.NewValidationError(errs.SubtypeFailedPrecondition, "%s", err.Error()).
-			WithCause(err)
-		if hookName != "" {
-			return typed.WithHint("plugin startup hook %q failed (reason_code %s); fix or remove the plugin before running commands", hookName, reasonCode)
+		return &output.ExitError{
+			Code: output.ExitValidation,
+			Detail: &output.ErrDetail{
+				Type:    "plugin_lifecycle",
+				Message: err.Error(),
+				Detail:  detail,
+			},
+			Err: err,
 		}
-		return typed.WithHint("a plugin startup hook failed (reason_code %s); fix or remove the plugin before running commands", reasonCode)
 	}
 	installFatalGuard(rootCmd, makeErr)
 }
@@ -156,7 +219,14 @@ func installPluginLifecycleErrorGuard(rootCmd *cobra.Command, err error) {
 //
 // This way the very first non-nil step in cobra's chain is always our
 // guard, regardless of which leaf the user invoked.
-func walkGuard(cmd *cobra.Command, makeErr func() error) {
+// Deprecated: walkGuard accepts a *output.ExitError-producing lambda, part
+// of the legacy error surface that predates the typed error contract
+// introduced by errs/. New code MUST NOT add new callers — the platform-
+// extension guard plumbing will switch to typed errs.* errors when the
+// platform-extension framework migrates. This wrapper is retained only for
+// the existing in-tree call sites; it will be removed once they have moved
+// to the typed surface.
+func walkGuard(cmd *cobra.Command, makeErr func() *output.ExitError) {
 	if cmd == nil {
 		return
 	}

cmd/platform_guards_test.go

@@ -6,14 +6,12 @@ package cmd
 import (
 	"context"
 	"errors"
-	"strings"
 	"sync"
 	"testing"
 	"time"
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/extension/platform"
 	"github.com/larksuite/cli/internal/hook"
 	"github.com/larksuite/cli/internal/output"
@@ -34,7 +32,7 @@ func (failClosedAbortingPlugin) Install(platform.Registrar) error {
 }
 
 // When a FailClosed plugin fails to install, buildInternal must
-// install a PersistentPreRunE that returns a typed *errs.ValidationError.
+// install a PersistentPreRunE that returns a structured *output.ExitError.
 // The user must NEVER see a silent partial-install state.
 //
 // This pins the build.go fix for codex's NEW ISSUE about
@@ -95,31 +93,26 @@ func TestBuildInternal_failClosedAbortsCLI(t *testing.T) {
 	checkGuardError(t, leaf.RunE(leaf, nil))
 }
 
-// checkGuardError asserts that err is the typed validation error the
-// install guard produces: a failed_precondition *errs.ValidationError
-// (exit 2) whose message + hint preserve the plugin name and the
-// install_failed reason code (the recovery info that lived in the legacy
-// detail map).
+// checkGuardError asserts that err is the structured plugin_install
+// ExitError the guard produces.
 func checkGuardError(t *testing.T, err error) {
 	t.Helper()
 	if err == nil {
 		t.Fatalf("PersistentPreRunE must surface the install error, got nil")
 	}
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "plugin_install" {
+		t.Errorf("envelope type = %q, want plugin_install", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	detail := exitErr.Detail.Detail.(map[string]any)
+	if detail["plugin"] != "policy" {
+		t.Errorf("detail.plugin = %v, want policy", detail["plugin"])
 	}
-	if !strings.Contains(verr.Hint, "policy") {
-		t.Errorf("hint should name the failing plugin %q, got %q", "policy", verr.Hint)
-	}
-	if !strings.Contains(verr.Hint, internalplatform.ReasonInstallFailed) {
-		t.Errorf("hint should surface reason_code %q, got %q", internalplatform.ReasonInstallFailed, verr.Hint)
+	if detail["reason_code"] != internalplatform.ReasonInstallFailed {
+		t.Errorf("detail.reason_code = %v, want install_failed", detail["reason_code"])
 	}
 }
 

cmd/plugin_integration_test.go

@@ -8,13 +8,11 @@ import (
 	"errors"
 	"os"
 	"path/filepath"
-	"strings"
 	"sync/atomic"
 	"testing"
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/extension/platform"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/cmdutil"
@@ -158,23 +156,19 @@ func TestPluginPipeline_wrapAbortReachesEnvelope(t *testing.T) {
 	}
 
 	err = leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "hook" {
+		t.Errorf("envelope type = %q, want hook", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	detail := exitErr.Detail.Detail.(map[string]any)
+	if detail["reason_code"] != "aborted" {
+		t.Errorf("detail.reason_code = %v, want aborted", detail["reason_code"])
 	}
-	// The namespaced hook name and the abort semantics are preserved in the
-	// message so a caller can identify which plugin hook rejected the call.
-	if !strings.Contains(verr.Message, "policy-plugin.policy") {
-		t.Errorf("message should name the aborting hook policy-plugin.policy, got %q", verr.Message)
-	}
-	if !strings.Contains(verr.Message, "aborted") {
-		t.Errorf("message should describe the abort, got %q", verr.Message)
+	if detail["hook_name"] != "policy-plugin.policy" {
+		t.Errorf("detail.hook_name = %v, want policy-plugin.policy", detail["hook_name"])
 	}
 
 	// errors.As must still reach the original AbortError so consumers
@@ -415,20 +409,15 @@ func TestPluginConflictGuard_MultipleRestrictAbortsCLI(t *testing.T) {
 		t.Fatalf("no runnable leaf in command tree")
 	}
 	err := leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "plugin_conflict" {
+		t.Errorf("envelope type = %q, want plugin_conflict", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	// reason_code multiple_restrict_plugins is folded into the hint so the
-	// operator can distinguish a multi-Restrict conflict from a bad rule.
-	if !strings.Contains(verr.Hint, "multiple_restrict_plugins") {
-		t.Errorf("hint should surface reason_code multiple_restrict_plugins, got %q", verr.Hint)
+	if rc := exitErr.Detail.Detail.(map[string]any)["reason_code"]; rc != "multiple_restrict_plugins" {
+		t.Errorf("reason_code = %v, want multiple_restrict_plugins", rc)
 	}
 }
 
@@ -458,20 +447,15 @@ func TestPluginConflictGuard_InvalidRuleAbortsCLI(t *testing.T) {
 		t.Fatalf("no runnable leaf in command tree")
 	}
 	err := leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "plugin_install" {
+		t.Errorf("envelope type = %q, want plugin_install", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	// reason_code invalid_rule is folded into the hint, distinct from the
-	// multiple_restrict_plugins conflict path.
-	if !strings.Contains(verr.Hint, "invalid_rule") {
-		t.Errorf("hint should surface reason_code invalid_rule, got %q", verr.Hint)
+	if rc := exitErr.Detail.Detail.(map[string]any)["reason_code"]; rc != "invalid_rule" {
+		t.Errorf("reason_code = %v, want invalid_rule", rc)
 	}
 }
 
@@ -500,24 +484,19 @@ func TestPluginLifecycleGuard_StartupErrorAbortsCLI(t *testing.T) {
 
 	leaf := findRunnableLeaf(root)
 	err := leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "plugin_lifecycle" {
+		t.Errorf("envelope type = %q, want plugin_lifecycle", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	d := exitErr.Detail.Detail.(map[string]any)
+	if d["reason_code"] != "lifecycle_failed" {
+		t.Errorf("reason_code = %v, want lifecycle_failed", d["reason_code"])
 	}
-	// reason_code lifecycle_failed (vs lifecycle_panic) and the failing
-	// hook name are folded into the hint so audit / on-call can tell the
-	// failure mode and which hook failed.
-	if !strings.Contains(verr.Hint, "lifecycle_failed") {
-		t.Errorf("hint should surface reason_code lifecycle_failed, got %q", verr.Hint)
-	}
-	if !strings.Contains(verr.Hint, "lc.start") {
-		t.Errorf("hint should name the failing hook lc.start, got %q", verr.Hint)
+	if d["hook_name"] != "lc.start" {
+		t.Errorf("hook_name = %v, want lc.start", d["hook_name"])
 	}
 }
 
@@ -541,20 +520,12 @@ func TestPluginLifecycleGuard_StartupPanicAbortsCLI(t *testing.T) {
 	}
 	leaf := findRunnableLeaf(root)
 	err := leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
-	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	// A panicking startup hook is distinguished from a returned error by
-	// reason_code lifecycle_panic in the hint.
-	if !strings.Contains(verr.Hint, "lifecycle_panic") {
-		t.Errorf("hint should surface reason_code lifecycle_panic, got %q", verr.Hint)
+	if rc := exitErr.Detail.Detail.(map[string]any)["reason_code"]; rc != "lifecycle_panic" {
+		t.Errorf("reason_code = %v, want lifecycle_panic", rc)
 	}
 }
 
@@ -608,24 +579,19 @@ func TestWrapperPanic_BecomesHookPanicEnvelope(t *testing.T) {
 	}()
 
 	err = leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "hook" {
+		t.Errorf("envelope type = %q, want hook", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	d := exitErr.Detail.Detail.(map[string]any)
+	if d["reason_code"] != "panic" {
+		t.Errorf("reason_code = %v, want panic", d["reason_code"])
 	}
-	// The recovered panic surfaces as a structured error naming the
-	// namespaced hook (p.boom) and describing the panic, so the process
-	// never crashes and the caller can attribute the failure.
-	if !strings.Contains(verr.Message, "p.boom") {
-		t.Errorf("message should name the namespaced hook p.boom, got %q", verr.Message)
-	}
-	if !strings.Contains(verr.Message, "panic") {
-		t.Errorf("message should describe the panic, got %q", verr.Message)
+	if d["hook_name"] != "p.boom" {
+		t.Errorf("hook_name = %v, want p.boom (namespaced)", d["hook_name"])
 	}
 }
 
@@ -687,24 +653,19 @@ func TestWrapperFactoryPanic_BecomesHookPanicEnvelope(t *testing.T) {
 	}()
 
 	err = leaf.RunE(leaf, nil)
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T %+v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError, got %T %+v", err, err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if exitErr.Detail.Type != "hook" {
+		t.Errorf("envelope type = %q, want hook", exitErr.Detail.Type)
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	d := exitErr.Detail.Detail.(map[string]any)
+	if d["reason_code"] != "panic" {
+		t.Errorf("reason_code = %v, want panic", d["reason_code"])
 	}
-	// A panic in the wrapper FACTORY (not just the inner handler) is
-	// recovered into the same structured panic error, naming the
-	// namespaced hook fac.bad-factory.
-	if !strings.Contains(verr.Message, "fac.bad-factory") {
-		t.Errorf("message should name the namespaced hook fac.bad-factory, got %q", verr.Message)
-	}
-	if !strings.Contains(verr.Message, "panic") {
-		t.Errorf("message should describe the panic, got %q", verr.Message)
+	if d["hook_name"] != "fac.bad-factory" {
+		t.Errorf("hook_name = %v, want fac.bad-factory (namespaced)", d["hook_name"])
 	}
 }
 

cmd/profile/add.go

@@ -12,7 +12,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/i18n"
@@ -54,9 +53,7 @@ func NewCmdProfileAdd(f *cmdutil.Factory) *cobra.Command {
 
 func profileAddRun(f *cmdutil.Factory, name, appID string, appSecretStdin bool, brand, lang string, useAfter bool) error {
 	if err := core.ValidateProfileName(name); err != nil {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "%v", err).
-			WithCause(err).
-			WithParam("--name")
+		return output.ErrValidation("%v", err)
 	}
 
 	langPref, err := cmdutil.ParseLangFlag(lang)
@@ -67,57 +64,46 @@ func profileAddRun(f *cmdutil.Factory, name, appID string, appSecretStdin bool,
 
 	// Read secret from stdin
 	if !appSecretStdin {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "app secret must be provided via stdin").
-			WithHint("use --app-secret-stdin and pipe the secret").
-			WithParam("--app-secret-stdin")
+		return output.ErrValidation("app secret must be provided via stdin: use --app-secret-stdin and pipe the secret")
 	}
 	scanner := bufio.NewScanner(f.IOStreams.In)
 	if !scanner.Scan() {
 		if err := scanner.Err(); err != nil {
-			return errs.NewValidationError(errs.SubtypeFailedPrecondition, "failed to read secret from stdin: %v", err).
-				WithCause(err).
-				WithParam("--app-secret-stdin")
+			return output.ErrValidation("failed to read secret from stdin: %v", err)
 		}
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "stdin is empty, expected app secret").
-			WithHint("pipe the app secret to stdin").
-			WithParam("--app-secret-stdin")
+		return output.ErrValidation("stdin is empty, expected app secret")
 	}
 	appSecret := strings.TrimSpace(scanner.Text())
 	if appSecret == "" {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "app secret read from stdin is empty").
-			WithHint("pipe a non-empty app secret to stdin").
-			WithParam("--app-secret-stdin")
+		return output.ErrValidation("app secret read from stdin is empty")
 	}
 
 	// Load or create config
 	multi, err := core.LoadMultiAppConfig()
 	if err != nil {
 		if !errors.Is(err, os.ErrNotExist) {
-			return errs.NewInternalError(errs.SubtypeFileIO, "failed to load config: %v", err).WithCause(err)
+			return output.Errorf(output.ExitInternal, "internal", "failed to load config: %v", err)
 		}
 		multi = &core.MultiAppConfig{}
 	}
 
 	// Check name uniqueness
 	if multi.FindApp(name) != nil {
-		return errs.NewValidationError(errs.SubtypeFailedPrecondition, "profile %q already exists", name).
-			WithHint("choose a different name, or remove the existing profile first").
-			WithParam("--name")
+		return output.ErrValidation("profile %q already exists", name)
 	}
 
 	// Check app-id uniqueness — keychain stores secrets by appId, so
 	// multiple profiles sharing the same appId would collide on credentials.
 	for _, a := range multi.Apps {
 		if a.AppId == appID {
-			return errs.NewValidationError(errs.SubtypeFailedPrecondition, "app-id %q is already used by profile %q; each profile must have a unique app-id", appID, a.ProfileName()).
-				WithParam("--app-id")
+			return output.ErrValidation("app-id %q is already used by profile %q; each profile must have a unique app-id", appID, a.ProfileName())
 		}
 	}
 
 	// Store secret securely
 	secret, err := core.ForStorage(appID, core.PlainSecret(appSecret), f.Keychain)
 	if err != nil {
-		return errs.NewInternalError(errs.SubtypeStorage, "%v", err).WithCause(err)
+		return output.Errorf(output.ExitInternal, "internal", "%v", err)
 	}
 
 	parsedBrand := core.ParseBrand(brand)
@@ -148,7 +134,7 @@ func profileAddRun(f *cmdutil.Factory, name, appID string, appSecretStdin bool,
 	}
 
 	if err := core.SaveMultiAppConfig(multi); err != nil {
-		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+		return output.Errorf(output.ExitInternal, "internal", "failed to save config: %v", err)
 	}
 
 	output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Profile %q added (%s, %s)", name, appID, parsedBrand))

cmd/profile/list.go

@@ -9,7 +9,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	larkauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -46,7 +45,7 @@ func profileListRun(f *cmdutil.Factory) error {
 			output.PrintJson(f.IOStreams.Out, []profileListItem{})
 			return nil
 		}
-		return errs.NewValidationError(errs.SubtypeFailedPrecondition, "failed to load config: %v", err).WithCause(err)
+		return output.Errorf(output.ExitValidation, "config", "failed to load config: %v", err)
 	}
 	if multi == nil || len(multi.Apps) == 0 {
 		output.PrintJson(f.IOStreams.Out, []profileListItem{})

cmd/profile/profile_test.go

@@ -11,7 +11,6 @@ import (
 	"strings"
 	"testing"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/i18n"
@@ -51,16 +50,6 @@ func TestProfileAddRun_InvalidExistingConfigReturnsError(t *testing.T) {
 	if !strings.Contains(err.Error(), "failed to load config") {
 		t.Fatalf("error = %v, want failed to load config", err)
 	}
-	var internalErr *errs.InternalError
-	if !errors.As(err, &internalErr) {
-		t.Fatalf("error type = %T, want *errs.InternalError; err=%v", err, err)
-	}
-	if internalErr.Subtype != errs.SubtypeFileIO {
-		t.Fatalf("subtype = %q, want %q", internalErr.Subtype, errs.SubtypeFileIO)
-	}
-	if code := output.ExitCodeOf(err); code != output.ExitInternal {
-		t.Fatalf("exit code = %d, want %d (ExitInternal)", code, output.ExitInternal)
-	}
 }
 
 // TestProfileAddRun_Lang covers the unified --lang contract on profile add:
@@ -106,9 +95,9 @@ func TestProfileAddRun_Lang(t *testing.T) {
 		if err == nil {
 			t.Fatal("expected validation error for --lang ZH, got nil")
 		}
-		var valErr *errs.ValidationError
-		if !errors.As(err, &valErr) || output.ExitCodeOf(err) != output.ExitValidation {
-			t.Fatalf("expected typed validation error with ExitValidation, got %T: %v", err, err)
+		exitErr, ok := err.(*output.ExitError)
+		if !ok || exitErr.Code != output.ExitValidation {
+			t.Fatalf("expected ExitValidation, got %T: %v", err, err)
 		}
 	})
 }
@@ -417,226 +406,17 @@ func TestProfileUseRun_SaveFailureReturnsStructuredError(t *testing.T) {
 func assertInternalExitError(t *testing.T, err error, wantMsg string) {
 	t.Helper()
 
-	var internalErr *errs.InternalError
-	if !errors.As(err, &internalErr) {
-		t.Fatalf("error type = %T, want *errs.InternalError; err=%v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("error type = %T, want *output.ExitError; err=%v", err, err)
 	}
-	if internalErr.Subtype != errs.SubtypeStorage {
-		t.Fatalf("subtype = %q, want %q", internalErr.Subtype, errs.SubtypeStorage)
+	if exitErr.Code != output.ExitInternal {
+		t.Fatalf("exit code = %d, want %d", exitErr.Code, output.ExitInternal)
 	}
-	if internalErr.Cause == nil {
-		t.Fatalf("cause = nil, want wrapped underlying error")
+	if exitErr.Detail == nil || exitErr.Detail.Type != "internal" {
+		t.Fatalf("detail = %#v, want internal detail", exitErr.Detail)
 	}
-	if !strings.Contains(internalErr.Message, wantMsg) {
-		t.Fatalf("message = %q, want contains %q", internalErr.Message, wantMsg)
-	}
-	if code := output.ExitCodeOf(err); code != output.ExitInternal {
-		t.Fatalf("exit code = %d, want %d (ExitInternal)", code, output.ExitInternal)
-	}
-}
-
-// assertValidationError asserts err is a typed *errs.ValidationError with the
-// given subtype, message fragment, and exit code 2.
-func assertValidationError(t *testing.T, err error, wantSubtype errs.Subtype, wantMsg string) *errs.ValidationError {
-	t.Helper()
-
-	if err == nil {
-		t.Fatal("expected error, got nil")
-	}
-	var valErr *errs.ValidationError
-	if !errors.As(err, &valErr) {
-		t.Fatalf("error type = %T, want *errs.ValidationError; err=%v", err, err)
-	}
-	if valErr.Subtype != wantSubtype {
-		t.Fatalf("subtype = %q, want %q", valErr.Subtype, wantSubtype)
-	}
-	if !strings.Contains(valErr.Message, wantMsg) {
-		t.Fatalf("message = %q, want contains %q", valErr.Message, wantMsg)
-	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Fatalf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	return valErr
-}
-
-func saveTwoProfiles(t *testing.T) {
-	t.Helper()
-	multi := &core.MultiAppConfig{
-		CurrentApp: "default",
-		Apps: []core.AppConfig{
-			{Name: "default", AppId: "app-default", AppSecret: core.PlainSecret("secret-default"), Brand: core.BrandFeishu},
-			{Name: "target", AppId: "app-target", AppSecret: core.PlainSecret("secret-target"), Brand: core.BrandLark},
-		},
-	}
-	if err := core.SaveMultiAppConfig(multi); err != nil {
-		t.Fatalf("SaveMultiAppConfig() error = %v", err)
-	}
-}
-
-func TestProfileAddRun_ValidationErrors(t *testing.T) {
-	t.Run("invalid profile name", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		f.IOStreams.In = strings.NewReader("secret\n")
-		err := profileAddRun(f, "bad name!", "app-x", true, "feishu", "", false)
-		valErr := assertValidationError(t, err, errs.SubtypeInvalidArgument, "")
-		if valErr.Param != "--name" {
-			t.Fatalf("param = %q, want %q", valErr.Param, "--name")
-		}
-		if valErr.Cause == nil {
-			t.Fatal("cause = nil, want wrapped validation error")
-		}
-	})
-
-	t.Run("missing app-secret-stdin flag", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileAddRun(f, "p", "app-x", false, "feishu", "", false)
-		valErr := assertValidationError(t, err, errs.SubtypeInvalidArgument, "app secret must be provided via stdin")
-		if valErr.Param != "--app-secret-stdin" {
-			t.Fatalf("param = %q, want %q", valErr.Param, "--app-secret-stdin")
-		}
-		if valErr.Hint == "" {
-			t.Fatal("hint is empty, want actionable hint")
-		}
-	})
-
-	t.Run("empty stdin", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		f.IOStreams.In = strings.NewReader("")
-		err := profileAddRun(f, "p", "app-x", true, "feishu", "", false)
-		valErr := assertValidationError(t, err, errs.SubtypeInvalidArgument, "stdin is empty")
-		if valErr.Param != "--app-secret-stdin" {
-			t.Fatalf("param = %q, want %q", valErr.Param, "--app-secret-stdin")
-		}
-	})
-
-	t.Run("blank secret on stdin", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		f.IOStreams.In = strings.NewReader("   \n")
-		err := profileAddRun(f, "p", "app-x", true, "feishu", "", false)
-		assertValidationError(t, err, errs.SubtypeInvalidArgument, "app secret read from stdin is empty")
-	})
-
-	t.Run("duplicate profile name", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		f.IOStreams.In = strings.NewReader("secret\n")
-		err := profileAddRun(f, "default", "app-new", true, "feishu", "", false)
-		valErr := assertValidationError(t, err, errs.SubtypeFailedPrecondition, `profile "default" already exists`)
-		if valErr.Param != "--name" {
-			t.Fatalf("param = %q, want %q", valErr.Param, "--name")
-		}
-	})
-
-	t.Run("duplicate app-id", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		f.IOStreams.In = strings.NewReader("secret\n")
-		err := profileAddRun(f, "fresh", "app-default", true, "feishu", "", false)
-		valErr := assertValidationError(t, err, errs.SubtypeFailedPrecondition, "already used by profile")
-		if valErr.Param != "--app-id" {
-			t.Fatalf("param = %q, want %q", valErr.Param, "--app-id")
-		}
-	})
-}
-
-func TestProfileUseRun_ValidationErrors(t *testing.T) {
-	t.Run("no previous profile for toggle", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileUseRun(f, "-")
-		valErr := assertValidationError(t, err, errs.SubtypeFailedPrecondition, "no previous profile to switch back to")
-		if valErr.Hint == "" {
-			t.Fatal("hint is empty, want actionable hint")
-		}
-	})
-
-	t.Run("profile not found", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileUseRun(f, "ghost")
-		assertValidationError(t, err, errs.SubtypeInvalidArgument, `profile "ghost" not found`)
-	})
-}
-
-func TestProfileRenameRun_ValidationErrors(t *testing.T) {
-	t.Run("invalid new name", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileRenameRun(f, "default", "bad name!")
-		valErr := assertValidationError(t, err, errs.SubtypeInvalidArgument, "")
-		if valErr.Cause == nil {
-			t.Fatal("cause = nil, want wrapped validation error")
-		}
-	})
-
-	t.Run("old profile not found", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileRenameRun(f, "ghost", "fresh")
-		assertValidationError(t, err, errs.SubtypeInvalidArgument, `profile "ghost" not found`)
-	})
-
-	t.Run("new name already exists", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileRenameRun(f, "default", "target")
-		valErr := assertValidationError(t, err, errs.SubtypeFailedPrecondition, `profile "target" already exists`)
-		if valErr.Hint == "" {
-			t.Fatal("hint is empty, want actionable hint")
-		}
-	})
-}
-
-func TestProfileRemoveRun_ValidationErrors(t *testing.T) {
-	t.Run("profile not found", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		saveTwoProfiles(t)
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileRemoveRun(f, "ghost")
-		assertValidationError(t, err, errs.SubtypeInvalidArgument, `profile "ghost" not found`)
-	})
-
-	t.Run("cannot remove the only profile", func(t *testing.T) {
-		setupProfileConfigDir(t)
-		multi := &core.MultiAppConfig{
-			CurrentApp: "solo",
-			Apps: []core.AppConfig{
-				{Name: "solo", AppId: "app-solo", AppSecret: core.PlainSecret("secret-solo"), Brand: core.BrandFeishu},
-			},
-		}
-		if err := core.SaveMultiAppConfig(multi); err != nil {
-			t.Fatalf("SaveMultiAppConfig() error = %v", err)
-		}
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		err := profileRemoveRun(f, "solo")
-		valErr := assertValidationError(t, err, errs.SubtypeFailedPrecondition, "cannot remove the only profile")
-		if valErr.Hint == "" {
-			t.Fatal("hint is empty, want actionable hint")
-		}
-	})
-}
-
-func TestProfileListRun_InvalidConfigReturnsValidationError(t *testing.T) {
-	dir := setupProfileConfigDir(t)
-	if err := os.WriteFile(filepath.Join(dir, "config.json"), []byte("{invalid json"), 0600); err != nil {
-		t.Fatalf("WriteFile() error = %v", err)
-	}
-
-	f, _, _, _ := cmdutil.TestFactory(t, nil)
-	err := profileListRun(f)
-	valErr := assertValidationError(t, err, errs.SubtypeFailedPrecondition, "failed to load config")
-	if valErr.Cause == nil {
-		t.Fatal("cause = nil, want wrapped load error")
+	if !strings.Contains(exitErr.Detail.Message, wantMsg) {
+		t.Fatalf("message = %q, want contains %q", exitErr.Detail.Message, wantMsg)
 	}
 }

cmd/profile/remove.go

@@ -9,7 +9,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	larkauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -41,12 +40,11 @@ func profileRemoveRun(f *cmdutil.Factory, name string) error {
 
 	idx := multi.FindAppIndex(name)
 	if idx < 0 {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "profile %q not found, available profiles: %s", name, strings.Join(multi.ProfileNames(), ", "))
+		return output.ErrValidation("profile %q not found, available profiles: %s", name, strings.Join(multi.ProfileNames(), ", "))
 	}
 
 	if len(multi.Apps) == 1 {
-		return errs.NewValidationError(errs.SubtypeFailedPrecondition, "cannot remove the only profile").
-			WithHint("add another profile first: lark-cli profile add")
+		return output.ErrValidation("cannot remove the only profile")
 	}
 
 	app := &multi.Apps[idx]
@@ -67,7 +65,7 @@ func profileRemoveRun(f *cmdutil.Factory, name string) error {
 	}
 
 	if err := core.SaveMultiAppConfig(multi); err != nil {
-		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+		return output.Errorf(output.ExitInternal, "internal", "failed to save config: %v", err)
 	}
 
 	// Best-effort credential cleanup after config commit

cmd/profile/rename.go

@@ -9,7 +9,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/output"
@@ -31,7 +30,7 @@ func NewCmdProfileRename(f *cmdutil.Factory) *cobra.Command {
 
 func profileRenameRun(f *cmdutil.Factory, oldName, newName string) error {
 	if err := core.ValidateProfileName(newName); err != nil {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "%v", err).WithCause(err)
+		return output.ErrValidation("%v", err)
 	}
 
 	multi, err := core.LoadOrNotConfigured()
@@ -41,7 +40,7 @@ func profileRenameRun(f *cmdutil.Factory, oldName, newName string) error {
 
 	idx := multi.FindAppIndex(oldName)
 	if idx < 0 {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "profile %q not found, available profiles: %s", oldName, strings.Join(multi.ProfileNames(), ", "))
+		return output.ErrValidation("profile %q not found, available profiles: %s", oldName, strings.Join(multi.ProfileNames(), ", "))
 	}
 
 	// Check new name uniqueness across other profiles, allowing renames to this
@@ -51,8 +50,7 @@ func profileRenameRun(f *cmdutil.Factory, oldName, newName string) error {
 			continue
 		}
 		if multi.Apps[i].Name == newName || multi.Apps[i].AppId == newName {
-			return errs.NewValidationError(errs.SubtypeFailedPrecondition, "profile %q already exists", newName).
-				WithHint("choose a different name")
+			return output.ErrValidation("profile %q already exists", newName)
 		}
 	}
 
@@ -68,7 +66,7 @@ func profileRenameRun(f *cmdutil.Factory, oldName, newName string) error {
 	}
 
 	if err := core.SaveMultiAppConfig(multi); err != nil {
-		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+		return output.Errorf(output.ExitInternal, "internal", "failed to save config: %v", err)
 	}
 
 	output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Profile renamed: %q -> %q", oldProfileName, newName))

cmd/profile/use.go

@@ -9,7 +9,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/output"
@@ -41,15 +40,14 @@ func profileUseRun(f *cmdutil.Factory, name string) error {
 	// Handle "-" for toggle-back
 	if name == "-" {
 		if multi.PreviousApp == "" {
-			return errs.NewValidationError(errs.SubtypeFailedPrecondition, "no previous profile to switch back to").
-				WithHint("switch to a profile by name first: lark-cli profile use <name>")
+			return output.ErrValidation("no previous profile to switch back to")
 		}
 		name = multi.PreviousApp
 	}
 
 	app := multi.FindApp(name)
 	if app == nil {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "profile %q not found, available profiles: %s", name, strings.Join(multi.ProfileNames(), ", "))
+		return output.ErrValidation("profile %q not found, available profiles: %s", name, strings.Join(multi.ProfileNames(), ", "))
 	}
 
 	targetName := app.ProfileName()
@@ -68,7 +66,7 @@ func profileUseRun(f *cmdutil.Factory, name string) error {
 	multi.CurrentApp = targetName
 
 	if err := core.SaveMultiAppConfig(multi); err != nil {
-		return errs.NewInternalError(errs.SubtypeStorage, "failed to save config: %v", err).WithCause(err)
+		return output.Errorf(output.ExitInternal, "internal", "failed to save config: %v", err)
 	}
 
 	output.PrintSuccess(f.IOStreams.ErrOut, fmt.Sprintf("Switched to profile %q (%s, %s)", targetName, app.AppId, app.Brand))

cmd/prune.go

@@ -9,10 +9,10 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/output"
 )
 
 // pruneForStrictMode removes commands incompatible with the active strict mode.
@@ -65,10 +65,10 @@ func strictModeStubFrom(child *cobra.Command, mode core.StrictMode) *cobra.Comma
 	//     pick auth's instead of our denial. A leaf-level no-op makes
 	//     cobra stop here and proceed to the wrapped RunE.
 	//
-	// strict-mode keeps its short Message + independent Hint and wraps
-	// the CommandDeniedError as the Cause by hand; BuildDenialError
-	// would override Message with the CommandDeniedError.Error() long
-	// form.
+	// strict-mode keeps its short Message + independent Hint and
+	// composes the shared detail.* / wrapped-CommandDeniedError shape
+	// by hand; BuildDenialError would override Message with the
+	// CommandDeniedError.Error() long form.
 	stubMessage := fmt.Sprintf(
 		"strict mode is %q, only %s-identity commands are available",
 		mode, mode.ForcedIdentity())
@@ -105,9 +105,20 @@ func strictModeStubFrom(child *cobra.Command, mode core.StrictMode) *cobra.Comma
 		},
 		RunE: func(c *cobra.Command, _ []string) error {
 			cd := cmdpolicy.CommandDeniedFromDenial(cmdpolicy.CanonicalPath(c), denial)
-			return errs.NewValidationError(errs.SubtypeFailedPrecondition, "%s", stubMessage).
-				WithHint("denied by %s policy (reason_code %s); %s", cd.Layer, cd.ReasonCode, stubHint).
-				WithCause(cd)
+			// Legacy *output.ExitError producer: this literal predates the
+			// typed error contract introduced by errs/. New denial sites MUST
+			// NOT construct *output.ExitError directly — they should return a
+			// typed *errs.XxxError once the cmdpolicy framework migrates.
+			return &output.ExitError{
+				Code: output.ExitValidation,
+				Detail: &output.ErrDetail{
+					Type:    "command_denied",
+					Message: stubMessage,
+					Hint:    stubHint,
+					Detail:  cmdpolicy.DenialDetailMap(cd),
+				},
+				Err: cd,
+			}
 		},
 	}
 }

cmd/prune_test.go

@@ -8,7 +8,6 @@ import (
 	"strings"
 	"testing"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/extension/platform"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/cmdutil"
@@ -248,12 +247,9 @@ func TestStrictModeStub_BypassesArgsValidator(t *testing.T) {
 	}
 }
 
-// Pins the strict-mode typed envelope: a failed_precondition
-// *errs.ValidationError (exit 2) carrying the short historical Message,
-// a Hint that still surfaces the policy layer + reason code (the
-// safety-critical recovery info that lived in the legacy detail map),
-// and the wrapped *platform.CommandDeniedError so external agents can
-// still inspect the structured denial taxonomy via errors.As.
+// Pins the strict-mode envelope shape: structured detail.* / wrapped
+// CommandDeniedError for external agents, AND the historical short
+// Message + independent Hint for existing consumers.
 func TestStrictModeStub_StructuredEnvelope(t *testing.T) {
 	root := newTestTree()
 	pruneForStrictMode(root, core.StrictModeBot)
@@ -266,33 +262,30 @@ func TestStrictModeStub_StructuredEnvelope(t *testing.T) {
 		t.Fatalf("strict-mode stub RunE should return error")
 	}
 
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("err is not *errs.ValidationError: %T", err)
+	var ee *output.ExitError
+	if !errors.As(err, &ee) {
+		t.Fatalf("err is not *output.ExitError: %T", err)
 	}
-	if verr.Subtype != errs.SubtypeFailedPrecondition {
-		t.Errorf("subtype = %q, want failed_precondition", verr.Subtype)
+	if ee.Detail == nil {
+		t.Fatalf("ExitError.Detail is nil; envelope writer cannot emit JSON")
 	}
-	if code := output.ExitCodeOf(err); code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
+	if ee.Detail.Type != "command_denied" {
+		t.Errorf("Detail.Type = %q, want command_denied", ee.Detail.Type)
 	}
-	// Short historical Message is preserved verbatim.
-	if verr.Message != `strict mode is "bot", only bot-identity commands are available` {
-		t.Errorf("Message = %q, want short historical form", verr.Message)
+	dm, ok := ee.Detail.Detail.(map[string]any)
+	if !ok {
+		t.Fatalf("Detail.Detail = %T, want map[string]any", ee.Detail.Detail)
 	}
-	// The denial layer + reason code remain user-readable in the hint, and
-	// the historical switch-policy guidance is still appended.
-	if !strings.Contains(verr.Hint, cmdpolicy.LayerStrictMode) {
-		t.Errorf("Hint = %q, want substring %q (policy layer)", verr.Hint, cmdpolicy.LayerStrictMode)
+	if got, _ := dm["layer"].(string); got != cmdpolicy.LayerStrictMode {
+		t.Errorf("Detail.Detail[layer] = %q, want %q", got, cmdpolicy.LayerStrictMode)
 	}
-	if !strings.Contains(verr.Hint, "identity_not_supported") {
-		t.Errorf("Hint = %q, want substring identity_not_supported (reason code)", verr.Hint)
+	if got, _ := dm["reason_code"].(string); got != "identity_not_supported" {
+		t.Errorf("Detail.Detail[reason_code] = %q, want identity_not_supported", got)
 	}
-	if !strings.Contains(verr.Hint, "if the user explicitly wants to switch policy") {
-		t.Errorf("Hint = %q, want historical switch-policy guidance", verr.Hint)
+	if got, _ := dm["policy_source"].(string); got != "strict-mode" {
+		t.Errorf("Detail.Detail[policy_source] = %q, want strict-mode", got)
 	}
 
-	// The structured denial taxonomy survives on the wrapped cause.
 	var cd *platform.CommandDeniedError
 	if !errors.As(err, &cd) {
 		t.Fatalf("err does not unwrap to *platform.CommandDeniedError")
@@ -303,12 +296,15 @@ func TestStrictModeStub_StructuredEnvelope(t *testing.T) {
 	if cd.ReasonCode != "identity_not_supported" {
 		t.Errorf("CommandDeniedError.ReasonCode = %q, want identity_not_supported", cd.ReasonCode)
 	}
-	if cd.PolicySource != "strict-mode" {
-		t.Errorf("CommandDeniedError.PolicySource = %q, want strict-mode", cd.PolicySource)
-	}
 	if !strings.Contains(cd.Reason, `strict mode is "bot"`) {
 		t.Errorf("CommandDeniedError.Reason = %q, want substring 'strict mode is \"bot\"'", cd.Reason)
 	}
+	if ee.Detail.Message != `strict mode is "bot", only bot-identity commands are available` {
+		t.Errorf("Detail.Message = %q, want short historical form", ee.Detail.Message)
+	}
+	if !strings.HasPrefix(ee.Detail.Hint, "if the user explicitly wants to switch policy") {
+		t.Errorf("Detail.Hint = %q, want historical hint", ee.Detail.Hint)
+	}
 }
 
 // strictModeStubFrom must write the denial annotations so the hook

cmd/root.go

@@ -13,12 +13,17 @@ import (
 
 	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/extension/platform"
+	internalauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/build"
 	"github.com/larksuite/cli/internal/cmdpolicy"
 	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/deprecation"
+	"github.com/larksuite/cli/internal/errclass"
+	"github.com/larksuite/cli/internal/errcompat"
 	"github.com/larksuite/cli/internal/hook"
 	"github.com/larksuite/cli/internal/output"
+	"github.com/larksuite/cli/internal/registry"
 	"github.com/larksuite/cli/internal/skillscheck"
 	"github.com/larksuite/cli/internal/suggest"
 	"github.com/larksuite/cli/internal/update"
@@ -212,37 +217,56 @@ func configureFlagCompletions(args []string) {
 // and returns the process exit code.
 //
 // Dispatch order:
-//  1. Typed errors from errs/ (e.g. *errs.PermissionError, *errs.APIError,
-//     *errs.SecurityPolicyError, *errs.AuthenticationError, *errs.ConfigError):
-//     render via the typed envelope writer, which lifts extension fields
-//     (missing_scopes, console_url, challenge_url, ...) to the top level.
-//     Routed by errs.CategoryOf via ExitCodeOf. Auth and config errors are
-//     constructed typed at their origin (internal/auth, internal/core), so the
-//     dispatcher no longer promotes any legacy shape here.
-//  2. PartialFailure / BareError signals: the result envelope is already on
-//     stdout; honor the exit code and write nothing to stderr.
-//  3. Residual cobra usage errors (missing required flag, unknown command,
-//     argument validation): typed as an invalid_argument envelope (exit 2),
-//     matching the explicit flag/subcommand guards. Flag parse errors are
-//     already typed upstream by the root FlagErrorFunc.
+//  1. Legacy shapes (*core.ConfigError, *internalauth.NeedAuthorizationError)
+//     are promoted via errcompat to their typed errs/ counterparts, with the
+//     original preserved in the Cause chain.
+//  2. Typed errors from errs/ (e.g. *errs.PermissionError, *errs.APIError,
+//     *errs.SecurityPolicyError, *errs.AuthenticationError): render via the
+//     typed envelope writer, which lifts extension fields (missing_scopes,
+//     console_url, challenge_url, ...) to the top level. Routed by
+//     errs.CategoryOf via ExitCodeOf.
+//  3. Legacy *output.ExitError: asExitError adapts it to the legacy
+//     envelope, written via WriteErrorEnvelope.
+//  4. Cobra errors (required flags, unknown commands, etc.): plain text.
 func handleRootError(f *cmdutil.Factory, err error) int {
 	errOut := f.IOStreams.ErrOut
 
+	// Promote legacy error shapes into typed errs/ before envelope marshal.
+	// NeedAuthorizationError check is first because it is the more specific
+	// shape; *core.ConfigError check follows. errors.As preserves the original
+	// in the Cause chain, so external errors.As(&core.ConfigError{}) consumers
+	// (cmd/auth/list.go, cmd/doctor/doctor.go, ...) still match.
+	//
+	// Outer-typed short-circuit: if err is already a typed *errs.* error,
+	// skip PromoteXxxError so the producer's Subtype / Hint / extension
+	// fields are not overwritten by a coarser promoted shape derived from a
+	// legacy error buried in its Cause chain. Promotion is only for legacy
+	// untyped entry points.
+	if !isOuterTypedError(err) {
+		var needAuthErr *internalauth.NeedAuthorizationError
+		if errors.As(err, &needAuthErr) {
+			err = errcompat.PromoteAuthError(needAuthErr)
+		} else {
+			var cfgErr *core.ConfigError
+			if errors.As(err, &cfgErr) {
+				err = errcompat.PromoteConfigError(cfgErr)
+			}
+		}
+	}
+
 	// When the typed error is a need_user_authorization signal, fold in the
 	// current command's declared scopes as a Hint so the user/AI sees the
 	// concrete scope(s) to re-auth with. The hint is computed on the fly from
 	// local shortcut/service metadata — it never depends on server state.
-	if !errs.IsRaw(err) {
-		applyNeedAuthorizationHint(f, err)
-	}
+	applyNeedAuthorizationHint(f, err)
 
 	// Staged dispatch: capture the typed exit code BEFORE attempting the
 	// envelope write. WriteTypedErrorEnvelope is best-effort on the wire
 	// (partial-write still returns true) so the exit code we read here is
 	// preserved even if stderr is torn — torn stderr must not downgrade
-	// typed exits 3/4/6/10 to the plain "Error:" path with exit 1.
+	// typed exits 3/4/6/10 to the legacy "Error:" path with exit 1.
 	// WriteTypedErrorEnvelope still returns false when err carries no
-	// Problem; in that case we fall through to the signal / plain-text paths.
+	// Problem; in that case we fall through to the legacy bridge below.
 	typedExit := output.ExitCodeOf(err)
 	if output.WriteTypedErrorEnvelope(errOut, err, string(f.ResolvedIdentity)) {
 		return typedExit
@@ -255,63 +279,58 @@ func handleRootError(f *cmdutil.Factory, err error) int {
 		return pfErr.Code
 	}
 
-	// Silent-exit signal (e.g. `auth check` predicate, or `update --json`):
-	// stdout already carries the result; honor the requested exit code and
-	// write nothing to stderr.
-	var bareErr *output.BareError
-	if errors.As(err, &bareErr) {
-		return bareErr.Code
-	}
-
-	// Errors reaching here are untyped: every RunE returns a typed errs.* error
-	// and flag-parse errors are typed by the root FlagErrorFunc. The remainder
-	// is either a cobra usage mistake (missing required flag, unknown command,
-	// wrong arg count), which cobra surfaces as a plain error identified by its
-	// stable text — the same external contract unknownFlagName relies on — or an
-	// untyped error that leaked past the typed boundary. Classify the former as
-	// invalid_argument (exit 2, like the explicit guards); treat the latter as an
-	// internal fault (exit 5) rather than blaming the user's input. The message
-	// is preserved either way, and the typed envelope still carries any pending
-	// deprecation notice.
-	var fallback error
-	if isCobraUsageError(err) {
-		fallback = errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", err.Error())
-	} else {
-		fallback = errs.NewInternalError(errs.SubtypeUnknown, "%s", err.Error()).WithCause(err)
-	}
-	output.WriteTypedErrorEnvelope(errOut, fallback, string(f.ResolvedIdentity))
-	return output.ExitCodeOf(fallback)
-}
-
-// cobraUsageErrorMarkers are the stable error-text fragments cobra / pflag
-// (pinned at v1.10.2) emit for usage mistakes — missing required flag, unknown
-// command / flag, wrong argument count. Cobra surfaces these as plain errors,
-// not a typed value we can match on, so the dispatcher recognizes them by text;
-// this is the same external contract unknownFlagName already depends on. A
-// residual error matching none of these has leaked the typed boundary and is
-// treated as an internal fault, not a user error.
-var cobraUsageErrorMarkers = []string{
-	"unknown command ",
-	"unknown flag: ",
-	"unknown shorthand",
-	"required flag(s) ",
-	"flag needs an argument",
-	"bad flag syntax:",
-	"no such flag ",
-	"invalid argument ",
-	"arg(s), ", // accepts / requires N arg(s), received / only received M
-}
-
-// isCobraUsageError reports whether err is a cobra / pflag usage mistake,
-// identified by the stable error text of the pinned cobra version.
-func isCobraUsageError(err error) bool {
-	msg := err.Error()
-	for _, m := range cobraUsageErrorMarkers {
-		if strings.Contains(msg, m) {
-			return true
+	if exitErr := asExitError(err); exitErr != nil {
+		if !exitErr.Raw {
+			// Raw errors (e.g. from `api` command via output.MarkRaw)
+			// preserve the original API error detail; skip enrichment
+			// which would clear it.
+			enrichMissingScopeError(f, exitErr)
+			enrichPermissionError(f, exitErr)
 		}
+		output.WriteErrorEnvelope(errOut, exitErr, string(f.ResolvedIdentity))
+		return exitErr.Code
 	}
-	return false
+
+	// A backward-compat alias records its deprecation notice in PreRunE, which
+	// runs before cobra's required-flag validation — but a missing required flag
+	// fails before RunE and lands here, where the bare "Error:" line would drop
+	// the notice. When a deprecation is pending, route through the structured
+	// envelope so the migration hint still reaches the caller; all other errors
+	// keep the existing plain output.
+	if deprecation.GetPending() != nil {
+		output.WriteErrorEnvelope(errOut, &output.ExitError{
+			Code:   1,
+			Detail: &output.ErrDetail{Type: "validation", Message: err.Error()},
+		}, string(f.ResolvedIdentity))
+		return 1
+	}
+	fmt.Fprintln(errOut, "Error:", err)
+	return 1
+}
+
+// isOuterTypedError returns true if err is a typed *errs.* error AT THE
+// TOP OF THE CHAIN (not buried inside Unwrap). Used by handleRootError
+// to gate PromoteXxxError so a producer's outer typed envelope is never
+// overwritten by a coarser shape derived from its legacy Cause.
+func isOuterTypedError(err error) bool {
+	_, ok := err.(errs.TypedError)
+	return ok
+}
+
+// asExitError converts known structured error types to *output.ExitError.
+// Returns nil for unrecognized errors (e.g. cobra flag errors).
+//
+// Deprecated: legacy *output.ExitError bridge.
+func asExitError(err error) *output.ExitError {
+	var cfgErr *core.ConfigError
+	if errors.As(err, &cfgErr) {
+		return output.ErrWithHint(cfgErr.Code, cfgErr.Type, cfgErr.Message, cfgErr.Hint)
+	}
+	var exitErr *output.ExitError
+	if errors.As(err, &exitErr) {
+		return exitErr
+	}
+	return nil
 }
 
 // installUnknownSubcommandGuard replaces cobra's silent help fallback on
@@ -342,10 +361,13 @@ func installUnknownSubcommandGuard(cmd *cobra.Command) {
 	}
 }
 
-// unknownSubcommandRunE replaces cobra's silent help fallback on group commands
-// with a typed *errs.ValidationError: a flag that belongs to a missing
-// subcommand, a misplaced subcommand-only flag, or an unknown subcommand name
-// each fail structured (exit 2) instead of degrading to help + exit 0.
+// Deprecated: unknownSubcommandRunE produces a legacy *output.ExitError that
+// predates the typed error contract introduced by errs/. New code MUST NOT
+// add producers of this shape — unknown-subcommand signals should move to
+// a typed *errs.ValidationError (or a dedicated typed error) carrying the
+// agent-protocol metadata as typed extension fields. This helper is retained
+// only while existing dispatch sites are migrated; it will be removed once
+// they have moved to the typed surface.
 func unknownSubcommandRunE(cmd *cobra.Command, args []string) error {
 	if len(args) == 0 {
 		// A bare group (e.g. `sheets`), or one carrying only group-valid flags
@@ -361,13 +383,28 @@ func unknownSubcommandRunE(cmd *cobra.Command, args []string) error {
 			return cmd.Help()
 		}
 		if unknown := unknownFlagTokens(cmd, rawInvocationArgs); len(unknown) > 0 {
-			verr := errs.NewValidationError(errs.SubtypeInvalidArgument,
-				"unknown flag %s before a subcommand for %q", strings.Join(unknown, ", "), cmd.CommandPath()).
-				WithHint("flags belong to a subcommand; run `%s --help` to list subcommands and their flags", cmd.CommandPath())
-			for _, flag := range unknown {
-				verr.WithParams(errs.InvalidParam{Name: flag, Reason: "unknown flag before a subcommand"})
+			return &output.ExitError{
+				Code: output.ExitValidation,
+				Detail: &output.ErrDetail{
+					Type:    "unknown_flag",
+					Message: fmt.Sprintf("unknown flag %s before a subcommand for %q", strings.Join(unknown, ", "), cmd.CommandPath()),
+					Hint:    fmt.Sprintf("flags belong to a subcommand; run `%s --help` to list subcommands and their flags", cmd.CommandPath()),
+					Detail: map[string]any{
+						// Keep the same detail keys as flagDidYouMean's unknown_flag
+						// so a consumer keyed on Type can read a stable shape. The
+						// subcommand isn't resolved here, so suggestions/valid_flags
+						// have no meaningful universe to draw from — emit empty
+						// rather than the group's own (misleading) flags. unknown is
+						// the back-compat singular field; unknown_flags carries the
+						// full list when more than one flag was supplied.
+						"unknown":       strings.Join(unknown, ", "),
+						"unknown_flags": unknown,
+						"command_path":  cmd.CommandPath(),
+						"suggestions":   []string{},
+						"valid_flags":   []string{},
+					},
+				},
 			}
-			return verr
 		}
 		// The remaining flags are all defined somewhere in the tree. Those valid
 		// on the group itself or inherited (e.g. the global --profile) do not
@@ -379,13 +416,19 @@ func unknownSubcommandRunE(cmd *cobra.Command, args []string) error {
 		if len(misplaced) == 0 {
 			return cmd.Help()
 		}
-		verr := errs.NewValidationError(errs.SubtypeInvalidArgument,
-			"missing subcommand for %q; flag %s belongs to a subcommand, not the group", cmd.CommandPath(), strings.Join(misplaced, ", ")).
-			WithHint("run `%s --help` to list subcommands and their flags", cmd.CommandPath())
-		for _, flag := range misplaced {
-			verr.WithParams(errs.InvalidParam{Name: flag, Reason: "flag belongs to a subcommand, not the group"})
+		return &output.ExitError{
+			Code: output.ExitValidation,
+			Detail: &output.ErrDetail{
+				Type:    "missing_subcommand",
+				Message: fmt.Sprintf("missing subcommand for %q; flag %s belongs to a subcommand, not the group", cmd.CommandPath(), strings.Join(misplaced, ", ")),
+				Hint:    fmt.Sprintf("run `%s --help` to list subcommands and their flags", cmd.CommandPath()),
+				Detail: map[string]any{
+					"command_path": cmd.CommandPath(),
+					"flags":        misplaced,
+					"suggestions":  []string{},
+				},
+			},
 		}
-		return verr
 	}
 	unknown := args[0]
 	available, deprecated := availableSubcommandNames(cmd)
@@ -399,12 +442,27 @@ func unknownSubcommandRunE(cmd *cobra.Command, args []string) error {
 		hint = fmt.Sprintf("did you mean one of: %s? (run `%s --help` for the full list)",
 			strings.Join(suggestions, ", "), cmd.CommandPath())
 	}
-	// Record the offending subcommand and its ranked candidates as a param with
-	// machine-readable Suggestions so an agent can retry without parsing the
-	// hint; the hint carries the same candidates as prose.
-	return errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", msg).
-		WithParams(errs.InvalidParam{Name: unknown, Reason: "unknown subcommand", Suggestions: suggestions}).
-		WithHint("%s", hint)
+	detail := map[string]any{
+		"unknown":      unknown,
+		"command_path": cmd.CommandPath(),
+		"suggestions":  suggestions,
+		"available":    available,
+	}
+	// Only services with backward-compat aliases (currently sheets) carry a
+	// deprecated bucket; omit the key elsewhere so every other service's
+	// envelope is unchanged.
+	if len(deprecated) > 0 {
+		detail["deprecated"] = deprecated
+	}
+	return &output.ExitError{
+		Code: output.ExitValidation,
+		Detail: &output.ErrDetail{
+			Type:    "unknown_subcommand",
+			Message: msg,
+			Hint:    hint,
+			Detail:  detail,
+		},
+	}
 }
 
 // flagTokensInArgs returns the flag-like tokens (-x, --foo, --foo=bar) in
@@ -530,34 +588,47 @@ func availableSubcommandNames(cmd *cobra.Command) (available, deprecated []strin
 }
 
 // flagDidYouMean is the root FlagErrorFunc (inherited by all subcommands). It
-// converts cobra's flag-parse errors into a typed validation envelope: an
-// unknown flag gets a focused "did you mean" hint (so agents recover even when
-// the typo is semantic, e.g. --query vs --find, where edit distance alone finds
-// nothing) and the offending flag in `params`. Other flag errors stay typed
-// but generic.
+// converts cobra's flag-parse errors into the structured ErrorEnvelope: an
+// unknown flag gets a focused "did you mean" hint plus the full valid-flag list
+// in detail (so agents recover even when the typo is semantic, e.g. --query vs
+// --find, where edit distance alone finds nothing). Other flag errors stay
+// structured but generic.
 func flagDidYouMean(c *cobra.Command, ferr error) error {
 	name, isUnknown := unknownFlagName(ferr)
 	if !isUnknown {
-		return errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", ferr.Error()).
-			WithHint("run `%s --help` for valid flags", c.CommandPath())
+		return &output.ExitError{
+			Code: output.ExitValidation,
+			Detail: &output.ErrDetail{
+				Type:    "flag_error",
+				Message: ferr.Error(),
+				Hint:    fmt.Sprintf("run `%s --help` for valid flags", c.CommandPath()),
+			},
+		}
 	}
 	valid := visibleFlagNames(c)
 	suggestions := suggest.Closest(name, valid, 3)
-	for i := range suggestions {
-		suggestions[i] = "--" + suggestions[i]
-	}
 	hint := fmt.Sprintf("run `%s --help` to see valid flags", c.CommandPath())
 	if len(suggestions) > 0 {
+		for i := range suggestions {
+			suggestions[i] = "--" + suggestions[i]
+		}
 		hint = fmt.Sprintf("did you mean %s? (run `%s --help` for all flags)",
 			strings.Join(suggestions, ", "), c.CommandPath())
 	}
-	// The ranked candidates ride on the param as machine-readable Suggestions so
-	// an agent can retry without parsing the hint; the hint carries the same
-	// candidates as prose. The full valid-flag list stays recoverable via --help.
-	return errs.NewValidationError(errs.SubtypeInvalidArgument,
-		"unknown flag %q for %q", "--"+name, c.CommandPath()).
-		WithParams(errs.InvalidParam{Name: "--" + name, Reason: "unknown flag", Suggestions: suggestions}).
-		WithHint("%s", hint)
+	return &output.ExitError{
+		Code: output.ExitValidation,
+		Detail: &output.ErrDetail{
+			Type:    "unknown_flag",
+			Message: fmt.Sprintf("unknown flag %q for %q", "--"+name, c.CommandPath()),
+			Hint:    hint,
+			Detail: map[string]any{
+				"unknown":      "--" + name,
+				"command_path": c.CommandPath(),
+				"suggestions":  suggestions,
+				"valid_flags":  valid,
+			},
+		},
+	}
 }
 
 // unknownFlagName extracts the offending long-flag name from cobra's flag-parse
@@ -627,3 +698,56 @@ func installTipsHelpFunc(root *cobra.Command) {
 		}
 	})
 }
+
+// enrichPermissionError rewrites the legacy *output.ExitError envelope so its
+// Message + Hint match the per-subtype canonical text produced by the typed
+// dispatcher path (errclass.CanonicalPermissionMessage / errclass.PermissionHint).
+// This guarantees a caller observing the wire envelope cannot tell whether
+// the error reached the dispatcher via the legacy *ExitError bridge or via
+// the typed *errs.PermissionError fast path.
+//
+// Deprecated: legacy *output.ExitError enrichment; typed PermissionError
+// values produced by errclass.BuildAPIError already carry MissingScopes +
+// ConsoleURL directly.
+func enrichPermissionError(f *cmdutil.Factory, exitErr *output.ExitError) {
+	if exitErr.Detail == nil {
+		return
+	}
+	// Only the legacy permission-class envelope types route here. "app_status"
+	// covers 99991662 (app_disabled) / 99991673 (app_unavailable); "permission"
+	// covers the four scope-class codes (99991672 / 99991676 / 99991679 / 230027).
+	if exitErr.Detail.Type != "permission" && exitErr.Detail.Type != "app_status" {
+		return
+	}
+
+	larkCode := exitErr.Detail.Code
+	meta, ok := errclass.LookupCodeMeta(larkCode)
+	if !ok || meta.Category != errs.CategoryAuthorization {
+		return
+	}
+
+	// Extract required scopes from API error detail (shared helper). May be
+	// empty for app-status codes — canonical message + hint still apply.
+	missing := registry.ExtractRequiredScopes(exitErr.Detail.Detail)
+
+	cfg, err := f.Config()
+	if err != nil {
+		return
+	}
+
+	// Reuse the same console URL builder as the typed path so both wire
+	// envelopes carry identical console_url values for the same input.
+	consoleURL := errclass.ConsoleURL(string(cfg.Brand), cfg.AppID, missing)
+
+	// Clear raw API detail — useful info is now in message/hint/console_url.
+	exitErr.Detail.Detail = nil
+
+	identity := string(f.ResolvedIdentity)
+	if identity == "" {
+		identity = "user"
+	}
+
+	exitErr.Detail.Message = errclass.CanonicalPermissionMessage(meta.Subtype, cfg.AppID, missing, exitErr.Detail.Message)
+	exitErr.Detail.Hint = errclass.PermissionHint(missing, identity, meta.Subtype, consoleURL)
+	exitErr.Detail.ConsoleURL = consoleURL
+}

cmd/root_integration_test.go

@@ -8,6 +8,7 @@ import (
 	"context"
 	"encoding/json"
 	"os"
+	"reflect"
 	"strings"
 	"testing"
 
@@ -26,12 +27,12 @@ import (
 	"github.com/spf13/cobra"
 )
 
-// Canonical strict-mode envelope messages shared across fixtures. The
-// switch-policy hint text is asserted by substring in
-// assertStrictModeDenialEnvelope.
+// Canonical strict-mode envelope strings shared across fixtures
+// (reflect.DeepEqual pins them; keep in sync with strictModeStubFrom).
 const (
 	strictModeBotMessage  = `strict mode is "bot", only bot-identity commands are available`
 	strictModeUserMessage = `strict mode is "user", only user-identity commands are available`
+	strictModeHint        = "if the user explicitly wants to switch policy, see `lark-cli config strict-mode --help` (confirm with the user before switching; switching does NOT require re-bind)"
 )
 
 // buildIntegrationRootCmd creates a root command with api, service, and shortcut
@@ -62,46 +63,37 @@ func executeRootIntegration(t *testing.T, f *cmdutil.Factory, rootCmd *cobra.Com
 	return 0
 }
 
-// typedErrorEnvelope mirrors the typed wire shape produced by
-// WriteTypedErrorEnvelope: the inner error marshals an errs.Problem
-// directly, so "type" is the category, "subtype" is top-level, and there
-// is no nested "detail" object. Recovery info (policy source, reason
-// code, suggestions) is folded into "hint".
-type typedErrorEnvelope struct {
-	OK       bool   `json:"ok"`
-	Identity string `json:"identity,omitempty"`
-	Error    struct {
-		Type    string `json:"type"`
-		Subtype string `json:"subtype"`
-		Message string `json:"message"`
-		Hint    string `json:"hint"`
-		Param   string `json:"param,omitempty"`
-	} `json:"error"`
-}
-
-// parseTypedEnvelope decodes stderr as the typed envelope and fails if the
-// legacy nested "detail" object is present (the migration removed it).
-func parseTypedEnvelope(t *testing.T, stderr *bytes.Buffer) typedErrorEnvelope {
+// parseEnvelope parses stderr bytes into an ErrorEnvelope.
+func parseEnvelope(t *testing.T, stderr *bytes.Buffer) output.ErrorEnvelope {
 	t.Helper()
 	if stderr.Len() == 0 {
 		t.Fatal("expected non-empty stderr, got empty")
 	}
-	var raw map[string]any
-	if err := json.Unmarshal(stderr.Bytes(), &raw); err != nil {
-		t.Fatalf("failed to parse stderr as JSON: %v\nstderr: %s", err, stderr.String())
-	}
-	if errObj, ok := raw["error"].(map[string]any); ok {
-		if _, hasDetail := errObj["detail"]; hasDetail {
-			t.Errorf("typed envelope must not carry a nested 'detail' object, got: %s", stderr.String())
-		}
-	}
-	var env typedErrorEnvelope
+	var env output.ErrorEnvelope
 	if err := json.Unmarshal(stderr.Bytes(), &env); err != nil {
-		t.Fatalf("failed to parse stderr as typed envelope: %v\nstderr: %s", err, stderr.String())
+		t.Fatalf("failed to parse stderr as ErrorEnvelope: %v\nstderr: %s", err, stderr.String())
 	}
 	return env
 }
 
+// assertEnvelope verifies exit code, stdout is empty, and stderr matches the
+// expected ErrorEnvelope exactly via reflect.DeepEqual.
+func assertEnvelope(t *testing.T, code int, wantCode int, stdout *bytes.Buffer, stderr *bytes.Buffer, want output.ErrorEnvelope) {
+	t.Helper()
+	if code != wantCode {
+		t.Errorf("exit code: got %d, want %d", code, wantCode)
+	}
+	if stdout.Len() != 0 {
+		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
+	}
+	got := parseEnvelope(t, stderr)
+	if !reflect.DeepEqual(got, want) {
+		gotJSON, _ := json.MarshalIndent(got, "", "  ")
+		wantJSON, _ := json.MarshalIndent(want, "", "  ")
+		t.Errorf("stderr envelope mismatch:\ngot:\n%s\nwant:\n%s", gotJSON, wantJSON)
+	}
+}
+
 func buildStrictModeIntegrationRootCmd(t *testing.T, f *cmdutil.Factory) *cobra.Command {
 	t.Helper()
 	rootCmd := &cobra.Command{Use: "lark-cli"}
@@ -213,71 +205,23 @@ func TestIntegration_StrictModeBot_ProfileOverride_DirectAuthLoginReturnsEnvelop
 
 	// auth login is user-only, so it gets pruned in strict-mode-bot and the
 	// stub error fires (not login.go's inline check, which is shadowed by
-	// pruning). The typed envelope is a failed_precondition validation
-	// error (exit 2); the strict-mode layer + reason code are folded into
-	// the hint.
-	if code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	env := parseTypedEnvelope(t, stderr)
-	assertStrictModeDenialEnvelope(t, env, strictModeBotMessage)
-}
-
-// assertStrictModeDenialEnvelope pins the shared strict-mode denial shape:
-// a validation/failed_precondition envelope whose message is the short
-// historical strict-mode line and whose hint still names the strict_mode
-// layer + identity_not_supported reason code (the safety-critical recovery
-// info), plus the historical switch-policy guidance.
-func assertStrictModeDenialEnvelope(t *testing.T, env typedErrorEnvelope, wantMessage string) {
-	t.Helper()
-	if env.OK {
-		t.Errorf("envelope ok = true, want false")
-	}
-	if env.Error.Type != "validation" {
-		t.Errorf("error.type = %q, want validation", env.Error.Type)
-	}
-	if env.Error.Subtype != "failed_precondition" {
-		t.Errorf("error.subtype = %q, want failed_precondition", env.Error.Subtype)
-	}
-	if env.Error.Message != wantMessage {
-		t.Errorf("error.message = %q, want %q", env.Error.Message, wantMessage)
-	}
-	if !strings.Contains(env.Error.Hint, "strict_mode") {
-		t.Errorf("error.hint = %q, want substring strict_mode (policy layer)", env.Error.Hint)
-	}
-	if !strings.Contains(env.Error.Hint, "identity_not_supported") {
-		t.Errorf("error.hint = %q, want substring identity_not_supported (reason code)", env.Error.Hint)
-	}
-	if !strings.Contains(env.Error.Hint, "config strict-mode --help") {
-		t.Errorf("error.hint = %q, want historical switch-policy guidance", env.Error.Hint)
-	}
-}
-
-// assertCheckStrictModeEnvelope pins the typed envelope produced by
-// cmdutil.Factory.CheckStrictMode (the identity-guard path for explicit
-// --as on shortcuts / service methods / api): a *errs.ValidationError with
-// subtype invalid_argument, the canonical strict-mode message, and the
-// switch-policy hint.
-func assertCheckStrictModeEnvelope(t *testing.T, env typedErrorEnvelope, wantMessage string) {
-	t.Helper()
-	if env.OK {
-		t.Errorf("envelope ok = true, want false")
-	}
-	if env.Error.Type != "validation" {
-		t.Errorf("error.type = %q, want validation", env.Error.Type)
-	}
-	if env.Error.Subtype != "invalid_argument" {
-		t.Errorf("error.subtype = %q, want invalid_argument", env.Error.Subtype)
-	}
-	if env.Error.Message != wantMessage {
-		t.Errorf("error.message = %q, want %q", env.Error.Message, wantMessage)
-	}
-	if !strings.Contains(env.Error.Hint, "config strict-mode --help") {
-		t.Errorf("error.hint = %q, want switch-policy guidance", env.Error.Hint)
-	}
+	// pruning).
+	assertEnvelope(t, code, output.ExitValidation, stdout, stderr, output.ErrorEnvelope{
+		OK: false,
+		Error: &output.ErrDetail{
+			Type:    "command_denied",
+			Message: strictModeBotMessage,
+			Hint:    strictModeHint,
+			Detail: map[string]any{
+				"path":          "auth/login",
+				"layer":         "strict_mode",
+				"policy_source": "strict-mode",
+				"rule_name":     "",
+				"reason_code":   "identity_not_supported",
+				"reason":        strictModeBotMessage,
+			},
+		},
+	})
 }
 
 func TestIntegration_StrictModeBot_ProfileOverride_DirectUserShortcutReturnsEnvelope(t *testing.T) {
@@ -288,14 +232,22 @@ func TestIntegration_StrictModeBot_ProfileOverride_DirectUserShortcutReturnsEnve
 		"im", "+messages-search", "--chat-id", "oc_xxx", "--query", "hello",
 	})
 
-	if code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	env := parseTypedEnvelope(t, stderr)
-	assertStrictModeDenialEnvelope(t, env, strictModeBotMessage)
+	assertEnvelope(t, code, output.ExitValidation, stdout, stderr, output.ErrorEnvelope{
+		OK: false,
+		Error: &output.ErrDetail{
+			Type:    "command_denied",
+			Message: strictModeBotMessage,
+			Hint:    strictModeHint,
+			Detail: map[string]any{
+				"path":          "im/+messages-search",
+				"layer":         "strict_mode",
+				"policy_source": "strict-mode",
+				"rule_name":     "",
+				"reason_code":   "identity_not_supported",
+				"reason":        strictModeBotMessage,
+			},
+		},
+	})
 }
 
 func TestIntegration_StrictModeUser_ProfileOverride_ChatCreateDryRunSucceeds(t *testing.T) {
@@ -325,14 +277,15 @@ func TestIntegration_StrictModeUser_ProfileOverride_ShortcutExplicitBotReturnsEn
 		"im", "+chat-create", "--name", "probe", "--as", "bot", "--dry-run",
 	})
 
-	if code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	env := parseTypedEnvelope(t, stderr)
-	assertCheckStrictModeEnvelope(t, env, strictModeUserMessage)
+	assertEnvelope(t, code, output.ExitValidation, stdout, stderr, output.ErrorEnvelope{
+		OK:       false,
+		Identity: "bot",
+		Error: &output.ErrDetail{
+			Type:    "validation",
+			Message: `strict mode is "user", only user-identity commands are available`,
+			Hint:    "if the user explicitly wants to switch policy, see `lark-cli config strict-mode --help` (confirm with the user before switching; switching does NOT require re-bind)",
+		},
+	})
 }
 
 func TestIntegration_StrictModeBot_ProfileOverride_ServiceExplicitUserReturnsEnvelope(t *testing.T) {
@@ -343,14 +296,15 @@ func TestIntegration_StrictModeBot_ProfileOverride_ServiceExplicitUserReturnsEnv
 		"im", "chats", "get", "--params", `{"chat_id":"oc_test"}`, "--as", "user", "--dry-run",
 	})
 
-	if code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	env := parseTypedEnvelope(t, stderr)
-	assertCheckStrictModeEnvelope(t, env, strictModeBotMessage)
+	assertEnvelope(t, code, output.ExitValidation, stdout, stderr, output.ErrorEnvelope{
+		OK:       false,
+		Identity: "user",
+		Error: &output.ErrDetail{
+			Type:    "validation",
+			Message: `strict mode is "bot", only bot-identity commands are available`,
+			Hint:    "if the user explicitly wants to switch policy, see `lark-cli config strict-mode --help` (confirm with the user before switching; switching does NOT require re-bind)",
+		},
+	})
 }
 
 func TestIntegration_StrictModeUser_ProfileOverride_ServiceBotOnlyMethodReturnsEnvelope(t *testing.T) {
@@ -361,14 +315,22 @@ func TestIntegration_StrictModeUser_ProfileOverride_ServiceBotOnlyMethodReturnsE
 		"im", "images", "create", "--data", `{"image_type":"message","image":"x"}`, "--dry-run",
 	})
 
-	if code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	env := parseTypedEnvelope(t, stderr)
-	assertStrictModeDenialEnvelope(t, env, strictModeUserMessage)
+	assertEnvelope(t, code, output.ExitValidation, stdout, stderr, output.ErrorEnvelope{
+		OK: false,
+		Error: &output.ErrDetail{
+			Type:    "command_denied",
+			Message: strictModeUserMessage,
+			Hint:    strictModeHint,
+			Detail: map[string]any{
+				"path":          "im/images/create",
+				"layer":         "strict_mode",
+				"policy_source": "strict-mode",
+				"rule_name":     "",
+				"reason_code":   "identity_not_supported",
+				"reason":        strictModeUserMessage,
+			},
+		},
+	})
 }
 
 func TestIntegration_StrictModeBot_ProfileOverride_APIExplicitUserReturnsEnvelope(t *testing.T) {
@@ -379,14 +341,15 @@ func TestIntegration_StrictModeBot_ProfileOverride_APIExplicitUserReturnsEnvelop
 		"api", "--as", "user", "GET", "/open-apis/im/v1/chats/oc_test", "--dry-run",
 	})
 
-	if code != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d (ExitValidation)", code, output.ExitValidation)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	env := parseTypedEnvelope(t, stderr)
-	assertCheckStrictModeEnvelope(t, env, strictModeBotMessage)
+	assertEnvelope(t, code, output.ExitValidation, stdout, stderr, output.ErrorEnvelope{
+		OK:       false,
+		Identity: "user",
+		Error: &output.ErrDetail{
+			Type:    "validation",
+			Message: `strict mode is "bot", only bot-identity commands are available`,
+			Hint:    "if the user explicitly wants to switch policy, see `lark-cli config strict-mode --help` (confirm with the user before switching; switching does NOT require re-bind)",
+		},
+	})
 }
 
 // --- shortcut command ---
@@ -409,43 +372,16 @@ func TestIntegration_Shortcut_BusinessError_OutputsEnvelope(t *testing.T) {
 		"im", "+messages-send", "--as", "bot", "--chat-id", "oc_xxx", "--text", "test",
 	})
 
-	// shortcut: typed errs.APIError via the CallAPITyped → BuildAPIError path.
-	if code != output.ExitAPI {
-		t.Errorf("exit code = %d, want %d (ExitAPI)", code, output.ExitAPI)
-	}
-	if stdout.Len() != 0 {
-		t.Errorf("expected empty stdout, got:\n%s", stdout.String())
-	}
-	if stderr.Len() == 0 {
-		t.Fatal("expected non-empty stderr, got empty")
-	}
-	var raw struct {
-		OK       bool   `json:"ok"`
-		Identity string `json:"identity"`
-		Error    struct {
-			Type    string `json:"type"`
-			Code    int    `json:"code"`
-			Message string `json:"message"`
-		} `json:"error"`
-	}
-	if err := json.Unmarshal(stderr.Bytes(), &raw); err != nil {
-		t.Fatalf("failed to parse typed envelope: %v\nstderr: %s", err, stderr.String())
-	}
-	if raw.OK {
-		t.Errorf("envelope ok = true, want false")
-	}
-	if raw.Identity != "bot" {
-		t.Errorf("identity = %q, want bot", raw.Identity)
-	}
-	if raw.Error.Type != "api" {
-		t.Errorf("error.type = %q, want api", raw.Error.Type)
-	}
-	if raw.Error.Code != 230002 {
-		t.Errorf("error.code = %d, want 230002", raw.Error.Code)
-	}
-	if raw.Error.Message != "Bot/User can NOT be out of the chat." {
-		t.Errorf("error.message = %q, want %q", raw.Error.Message, "Bot/User can NOT be out of the chat.")
-	}
+	// shortcut: typed error via DoAPIJSON path
+	assertEnvelope(t, code, output.ExitAPI, stdout, stderr, output.ErrorEnvelope{
+		OK:       false,
+		Identity: "bot",
+		Error: &output.ErrDetail{
+			Type:    "api",
+			Code:    230002,
+			Message: "Bot/User can NOT be out of the chat.",
+		},
+	})
 }
 
 // TestSetupNotices_ColdStart_NoNotice verifies that missing state

cmd/root_test.go

@@ -137,6 +137,9 @@ func TestIsCompletionCommand(t *testing.T) {
 	}
 }
 
+// TestPromoteConfigError_* lives with the implementation in
+// internal/errcompat/promote_test.go.
+
 // TestHandleRootError_SecurityPolicyCanonicalEnvelope verifies that
 // *errs.SecurityPolicyError flows through the canonical typed envelope
 // (output.WriteTypedErrorEnvelope) — type=policy, numeric code, subtype,
@@ -266,11 +269,12 @@ func (f *failingWriter) Write(p []byte) (int, error) {
 	return len(p), nil
 }
 
-// TestHandleRootError_DeprecatedAliasMissingFlagStructured pins that a
-// backward-compat alias failing on a cobra-level required flag (which
-// short-circuits before RunE) routes through the structured envelope, so the
-// deprecation notice OnInvoke records in PreRunE is carried on the wire instead
-// of being dropped on a plain "Error:" line.
+// TestHandleRootError_DeprecatedAliasMissingFlagStructured pins issue #4: a
+// backward-compat alias that fails on a cobra-level required flag (which
+// short-circuits before RunE) still routes through the structured envelope,
+// because OnInvoke records the deprecation in PreRunE and the legacy fallback
+// switches to WriteErrorEnvelope when a deprecation is pending — so the
+// migration notice is no longer dropped on the plain "Error:" line.
 func TestHandleRootError_DeprecatedAliasMissingFlagStructured(t *testing.T) {
 	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
 	t.Cleanup(func() { deprecation.SetPending(nil) })
@@ -282,9 +286,9 @@ func TestHandleRootError_DeprecatedAliasMissingFlagStructured(t *testing.T) {
 	deprecation.SetPending(&deprecation.Notice{
 		Command: "+write", Replacement: "+cells-set", Skill: "lark-sheets",
 	})
-	// The bare error shape cobra's ValidateRequiredFlags produces: not a typed
-	// errs.* error, so it reaches the deprecation fallback.
-	exit := handleRootError(f, fmt.Errorf(`required flag(s) %q not set`, "values"))
+	// The bare error shape cobra's ValidateRequiredFlags produces: neither typed
+	// nor an *output.ExitError, so it reaches the legacy fallback.
+	handleRootError(f, fmt.Errorf(`required flag(s) %q not set`, "values"))
 
 	out := errOut.String()
 	if strings.HasPrefix(strings.TrimSpace(out), "Error:") {
@@ -293,96 +297,12 @@ func TestHandleRootError_DeprecatedAliasMissingFlagStructured(t *testing.T) {
 	if !strings.Contains(out, `"message"`) || !strings.Contains(out, "values") {
 		t.Errorf("expected a JSON error envelope carrying the failure message; got:\n%s", out)
 	}
-	// The envelope is typed validation, so the exit code must derive from that
-	// category (2) — the wire type and the exit code must not disagree.
-	if exit != int(output.ExitValidation) {
-		t.Errorf("exit = %d, want %d (validation envelope → category-derived exit)", exit, int(output.ExitValidation))
-	}
 }
 
-// TestHandleRootError_AuthConfigWireGolden is the wire-consistency regression
-// baseline for auth/config errors: it pins the typed envelope and exit code the
-// dispatcher produces for the two source-of-truth shapes, which are constructed
-// typed at their origin in internal/auth and internal/core.
-func TestHandleRootError_AuthConfigWireGolden(t *testing.T) {
-	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
-
-	t.Run("token missing exits 3 with token_missing authentication envelope", func(t *testing.T) {
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		errOut := &bytes.Buffer{}
-		f.IOStreams.ErrOut = errOut
-
-		exit := handleRootError(f, internalauth.NewNeedUserAuthorizationError("u_golden"))
-		if exit != int(output.ExitAuth) {
-			t.Errorf("exit = %d, want %d (ExitAuth)", exit, int(output.ExitAuth))
-		}
-
-		errObj := decodeErrorEnvelope(t, errOut.Bytes())
-		if got := errObj["type"]; got != "authentication" {
-			t.Errorf("error.type = %v, want %q", got, "authentication")
-		}
-		if got := errObj["subtype"]; got != "token_missing" {
-			t.Errorf("error.subtype = %v, want %q", got, "token_missing")
-		}
-		if got, _ := errObj["message"].(string); !strings.Contains(got, "need_user_authorization") {
-			t.Errorf("error.message = %q, must keep the need_user_authorization marker", got)
-		}
-		if got, _ := errObj["message"].(string); !strings.Contains(got, "u_golden") {
-			t.Errorf("error.message = %q, must carry the user open id", got)
-		}
-		if got, _ := errObj["hint"].(string); !strings.Contains(got, "auth login") {
-			t.Errorf("error.hint = %q, must point at auth login", got)
-		}
-		if got := errObj["user_open_id"]; got != "u_golden" {
-			t.Errorf("error.user_open_id = %v, want %q", got, "u_golden")
-		}
-	})
-
-	t.Run("not configured exits 3 with not_configured config envelope", func(t *testing.T) {
-		f, _, _, _ := cmdutil.TestFactory(t, nil)
-		errOut := &bytes.Buffer{}
-		f.IOStreams.ErrOut = errOut
-
-		exit := handleRootError(f, core.NotConfiguredError())
-		if exit != int(output.ExitAuth) {
-			t.Errorf("exit = %d, want %d (config shares ExitAuth)", exit, int(output.ExitAuth))
-		}
-
-		errObj := decodeErrorEnvelope(t, errOut.Bytes())
-		if got := errObj["type"]; got != "config" {
-			t.Errorf("error.type = %v, want %q", got, "config")
-		}
-		if got := errObj["subtype"]; got != "not_configured" {
-			t.Errorf("error.subtype = %v, want %q", got, "not_configured")
-		}
-		if got, _ := errObj["message"].(string); !strings.Contains(got, "not configured") {
-			t.Errorf("error.message = %q, want the not-configured message", got)
-		}
-		if got, _ := errObj["hint"].(string); !strings.Contains(got, "config init") {
-			t.Errorf("error.hint = %q, must point at config init", got)
-		}
-	})
-}
-
-// decodeErrorEnvelope unmarshals a typed error envelope and returns its
-// top-level "error" object, failing the test if the shape is unexpected.
-func decodeErrorEnvelope(t *testing.T, raw []byte) map[string]any {
-	t.Helper()
-	var env map[string]any
-	if err := json.Unmarshal(raw, &env); err != nil {
-		t.Fatalf("envelope is not valid JSON: %v\n%s", err, raw)
-	}
-	errObj, ok := env["error"].(map[string]any)
-	if !ok {
-		t.Fatalf("envelope missing top-level error object: %s", raw)
-	}
-	return errObj
-}
-
-// TestHandleRootError_NoDeprecationTypesUsageError pins that a residual cobra
-// usage error (missing required flag) is typed as invalid_argument with exit 2
-// even with no deprecation pending — never cobra's plain "Error:" line.
-func TestHandleRootError_NoDeprecationTypesUsageError(t *testing.T) {
+// TestHandleRootError_NoDeprecationKeepsPlainError pins the other half: with no
+// deprecation pending, the legacy fallback stays a plain "Error:" line, so the
+// fix does not reshape every unrecognized cobra error.
+func TestHandleRootError_NoDeprecationKeepsPlainError(t *testing.T) {
 	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
 	t.Cleanup(func() { deprecation.SetPending(nil) })
 	deprecation.SetPending(nil)
@@ -391,45 +311,9 @@ func TestHandleRootError_NoDeprecationTypesUsageError(t *testing.T) {
 	errOut := &bytes.Buffer{}
 	f.IOStreams.ErrOut = errOut
 
-	exit := handleRootError(f, fmt.Errorf(`required flag(s) %q not set`, "values"))
-
-	out := errOut.String()
-	if strings.HasPrefix(strings.TrimSpace(out), "Error:") {
-		t.Fatalf("want a structured envelope, got a plain Error: line:\n%s", out)
-	}
-	errObj := decodeErrorEnvelope(t, errOut.Bytes())
-	if got := errObj["type"]; got != "validation" {
-		t.Errorf("error.type = %v, want %q", got, "validation")
-	}
-	if got, _ := errObj["message"].(string); !strings.Contains(got, "values") {
-		t.Errorf("error.message = %q, must carry the failing flag name", got)
-	}
-	if exit != int(output.ExitValidation) {
-		t.Errorf("exit = %d, want %d (validation envelope → category-derived exit)", exit, int(output.ExitValidation))
-	}
-}
-
-// TestHandleRootError_LeakedUntypedErrorBecomesInternal pins that an untyped
-// error that does NOT match a cobra usage shape (i.e. one that leaked past the
-// typed boundary from a helper) is classified as an internal fault (exit 5),
-// not blamed on the user's input as a validation error.
-func TestHandleRootError_LeakedUntypedErrorBecomesInternal(t *testing.T) {
-	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
-	t.Cleanup(func() { deprecation.SetPending(nil) })
-	deprecation.SetPending(nil)
-
-	f, _, _, _ := cmdutil.TestFactory(t, nil)
-	errOut := &bytes.Buffer{}
-	f.IOStreams.ErrOut = errOut
-
-	exit := handleRootError(f, fmt.Errorf("upstream helper exploded: %w", io.ErrUnexpectedEOF))
-
-	errObj := decodeErrorEnvelope(t, errOut.Bytes())
-	if got := errObj["type"]; got != "internal" {
-		t.Errorf("error.type = %v, want %q (leaked untyped error must not be mislabeled validation)", got, "internal")
-	}
-	if exit != int(output.ExitInternal) {
-		t.Errorf("exit = %d, want %d (internal envelope → category-derived exit)", exit, int(output.ExitInternal))
+	handleRootError(f, fmt.Errorf(`required flag(s) %q not set`, "values"))
+	if !strings.HasPrefix(errOut.String(), "Error:") {
+		t.Errorf("no deprecation pending: want a plain 'Error:' line, got:\n%s", errOut.String())
 	}
 }
 
@@ -453,32 +337,12 @@ func TestHandleRootError_PartialWritePreservesExitCode(t *testing.T) {
 	}
 }
 
-// TestHandleRootError_BareErrorExitCodeNoStderr pins the silent-exit
-// contract: a *output.BareError is honored for its exit code while stderr stays
-// empty (stdout already carries the result, so the dispatcher must not layer a
-// second envelope on top).
-func TestHandleRootError_BareErrorExitCodeNoStderr(t *testing.T) {
-	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
-
-	f, _, _, _ := cmdutil.TestFactory(t, nil)
-	errOut := &bytes.Buffer{}
-	f.IOStreams.ErrOut = errOut
-
-	exit := handleRootError(f, output.ErrBare(output.ExitAuth))
-	if exit != int(output.ExitAuth) {
-		t.Errorf("exit = %d, want %d (BareError code propagated)", exit, int(output.ExitAuth))
-	}
-	if errOut.Len() != 0 {
-		t.Errorf("stderr must stay empty for a bare predicate signal, got:\n%s", errOut.String())
-	}
-}
-
-// TestHandleRootError_TypedAuthErrorWithLegacyCausePreserved pins that a typed
-// *errs.AuthenticationError carrying a legacy *NeedAuthorizationError in its
-// Cause chain renders the producer's TokenExpired subtype + custom hint
-// verbatim — the legacy sentinel in the Cause chain never coarsens the wire
-// shape.
-func TestHandleRootError_TypedAuthErrorWithLegacyCausePreserved(t *testing.T) {
+// TestHandleRootError_TypedOuterShortCircuitsPromote pins that when a typed
+// *errs.AuthenticationError carries a legacy *NeedAuthorizationError in its
+// Cause chain, the dispatcher does NOT run PromoteAuthError — doing so
+// would replace the producer's TokenExpired subtype + custom hint with the
+// promoted shape's TokenMissing.
+func TestHandleRootError_TypedOuterShortCircuitsPromote(t *testing.T) {
 	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
 
 	f, _, _, _ := cmdutil.TestFactory(t, nil)
@@ -630,3 +494,136 @@ func TestApplyNeedAuthorizationHint_AppendsExistingHint(t *testing.T) {
 		t.Errorf("expected appended hint %q, got %q", want, authErr.Hint)
 	}
 }
+
+// TestEnrichPermissionError_CanonicalConvergence pins that the legacy
+// *output.ExitError dispatch path produces the same canonical Message + Hint
+// + ConsoleURL as the typed *errs.PermissionError dispatch path. Both paths
+// share errclass.CanonicalPermissionMessage / errclass.PermissionHint /
+// errclass.ConsoleURL — so a wire consumer cannot tell which path produced
+// the envelope.
+func TestEnrichPermissionError_CanonicalConvergence(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+	cases := []struct {
+		name            string
+		larkCode        int
+		legacyErrType   string
+		wantMsgSubstrs  []string
+		wantHintSubstrs []string
+		wantConsoleURL  bool
+		wantNoAuthLogin bool // hint must not suggest `auth login`
+	}{
+		{
+			name:            "99991672 app_scope_not_applied",
+			larkCode:        99991672,
+			legacyErrType:   "permission",
+			wantMsgSubstrs:  []string{"access denied", "app cli_test", "drive:drive:read"},
+			wantHintSubstrs: []string{"developer console", "open.feishu.cn"},
+			wantConsoleURL:  true,
+			wantNoAuthLogin: true,
+		},
+		{
+			name:            "99991679 missing_scope",
+			larkCode:        99991679,
+			legacyErrType:   "permission",
+			wantMsgSubstrs:  []string{"unauthorized", "user authorization"},
+			wantHintSubstrs: []string{"lark-cli auth login"},
+		},
+		{
+			name:            "99991673 app_unavailable",
+			larkCode:        99991673,
+			legacyErrType:   "app_status",
+			wantMsgSubstrs:  []string{"unauthorized app", "app cli_test", "not properly installed"},
+			wantHintSubstrs: []string{"tenant admin", "install status"},
+		},
+		{
+			name:            "99991662 app_disabled",
+			larkCode:        99991662,
+			legacyErrType:   "app_status",
+			wantMsgSubstrs:  []string{"app cli_test", "not in use", "currently disabled"},
+			wantHintSubstrs: []string{"tenant admin", "re-enable"},
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+				AppID: "cli_test", AppSecret: "s", Brand: core.BrandFeishu,
+			})
+			f.ResolvedIdentity = core.AsUser
+
+			// Mimic the wire shape ErrAPI produces: legacy *ExitError with
+			// Detail.Type populated by ClassifyLarkError, Detail.Detail
+			// carrying the permission_violations block so ExtractRequiredScopes
+			// can recover the missing scope.
+			scopeForDetail := "drive:drive:read"
+			exitErr := &output.ExitError{
+				Code: output.ExitAPI,
+				Detail: &output.ErrDetail{
+					Type:    tc.legacyErrType,
+					Code:    tc.larkCode,
+					Message: "upstream raw message — must be replaced",
+					Detail: map[string]interface{}{
+						"permission_violations": []interface{}{
+							map[string]interface{}{"subject": scopeForDetail},
+						},
+					},
+				},
+			}
+			enrichPermissionError(f, exitErr)
+
+			for _, sub := range tc.wantMsgSubstrs {
+				if !strings.Contains(exitErr.Detail.Message, sub) {
+					t.Errorf("Message %q missing substring %q", exitErr.Detail.Message, sub)
+				}
+			}
+			if exitErr.Detail.Message == "upstream raw message — must be replaced" {
+				t.Errorf("Message must be rewritten to canonical text; got upstream verbatim")
+			}
+			for _, sub := range tc.wantHintSubstrs {
+				if !strings.Contains(exitErr.Detail.Hint, sub) {
+					t.Errorf("Hint %q missing substring %q", exitErr.Detail.Hint, sub)
+				}
+			}
+			if tc.wantNoAuthLogin && strings.Contains(exitErr.Detail.Hint, "auth login") {
+				t.Errorf("Hint must not suggest `auth login` for this subtype; got %q", exitErr.Detail.Hint)
+			}
+			if tc.wantConsoleURL && exitErr.Detail.ConsoleURL == "" {
+				t.Error("ConsoleURL should be populated when missing scopes are present")
+			}
+		})
+	}
+}
+
+// TestEnrichPermissionError_SkipsUnrelatedTypes pins that an ExitError whose
+// Detail.Type is neither "permission" nor "app_status" is left untouched —
+// no Message rewrite, no Hint rewrite, no ConsoleURL injection.
+func TestEnrichPermissionError_SkipsUnrelatedTypes(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "cli_test", AppSecret: "s", Brand: core.BrandFeishu,
+	})
+	f.ResolvedIdentity = core.AsUser
+
+	for _, ty := range []string{"api_error", "validation", "rate_limit", "auth"} {
+		exitErr := &output.ExitError{
+			Code: output.ExitAPI,
+			Detail: &output.ErrDetail{
+				Type:    ty,
+				Code:    99991400,
+				Message: "untouched",
+				Hint:    "original hint",
+			},
+		}
+		enrichPermissionError(f, exitErr)
+		if exitErr.Detail.Message != "untouched" {
+			t.Errorf("type=%q: Message was rewritten unexpectedly: %q", ty, exitErr.Detail.Message)
+		}
+		if exitErr.Detail.Hint != "original hint" {
+			t.Errorf("type=%q: Hint was rewritten unexpectedly: %q", ty, exitErr.Detail.Hint)
+		}
+		if exitErr.Detail.ConsoleURL != "" {
+			t.Errorf("type=%q: ConsoleURL should not be injected; got %q", ty, exitErr.Detail.ConsoleURL)
+		}
+	}
+}

cmd/schema/schema_test.go

@@ -5,11 +5,9 @@ package schema
 
 import (
 	"encoding/json"
-	"errors"
 	"strings"
 	"testing"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 )
@@ -211,45 +209,6 @@ func TestSchemaCmd_UnknownService(t *testing.T) {
 	if !strings.Contains(err.Error(), "Unknown service") {
 		t.Errorf("expected 'Unknown service' error, got: %v", err)
 	}
-	var ve *errs.ValidationError
-	if !errors.As(err, &ve) {
-		t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
-	}
-	if ve.Subtype != errs.SubtypeInvalidArgument {
-		t.Errorf("Subtype = %q, want %q", ve.Subtype, errs.SubtypeInvalidArgument)
-	}
-	if !strings.Contains(ve.Hint, "Available:") {
-		t.Errorf("expected hint listing available services, got: %q", ve.Hint)
-	}
-}
-
-// TestSchemaCmd_UnknownMethod_TypedValidation pins the typed envelope for the
-// JSON-mode unknown-method path: *errs.ValidationError with
-// subtype invalid_argument and a hint listing the available methods.
-func TestSchemaCmd_UnknownMethod_TypedValidation(t *testing.T) {
-	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
-		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
-	})
-
-	cmd := NewCmdSchema(f, nil)
-	cmd.SetArgs([]string{"calendar.events.nonexistent_method"})
-	err := cmd.Execute()
-	if err == nil {
-		t.Fatal("expected error for unknown method")
-	}
-	var ve *errs.ValidationError
-	if !errors.As(err, &ve) {
-		t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
-	}
-	if ve.Subtype != errs.SubtypeInvalidArgument {
-		t.Errorf("Subtype = %q, want %q", ve.Subtype, errs.SubtypeInvalidArgument)
-	}
-	if !strings.Contains(err.Error(), "Unknown method") {
-		t.Errorf("expected 'Unknown method' error, got: %v", err)
-	}
-	if !strings.Contains(ve.Hint, "Available:") {
-		t.Errorf("expected hint listing available methods, got: %q", ve.Hint)
-	}
 }
 
 // Completion candidate generation (dotted + space forms, strict-mode filtering,

cmd/unknown_subcommand_test.go

@@ -11,7 +11,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/output"
 )
@@ -127,20 +126,29 @@ func TestUnknownSubcommandRunE_FlagBeforeSubcommandIsStructured(t *testing.T) {
 		t.Errorf("error = %q, want it to mention an unknown flag", err.Error())
 	}
 
-	// Typed surface: a validation error (exit 2) whose Params carries the
-	// offending flag so an agent can recover the token without parsing prose.
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	// The detail must stay schema-compatible with flagDidYouMean's unknown_flag
+	// (same Type → same keys), so a consumer keyed on Type reads a stable shape.
+	exitErr, ok := err.(*output.ExitError)
+	if !ok || exitErr.Detail == nil {
+		t.Fatalf("expected *output.ExitError with Detail, got %T", err)
 	}
-	if verr.Subtype != errs.SubtypeInvalidArgument {
-		t.Errorf("subtype = %q, want invalid_argument", verr.Subtype)
+	if exitErr.Detail.Type != "unknown_flag" {
+		t.Errorf("detail.Type = %q, want unknown_flag", exitErr.Detail.Type)
 	}
-	if output.ExitCodeOf(err) != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d", output.ExitCodeOf(err), output.ExitValidation)
+	detail, ok := exitErr.Detail.Detail.(map[string]any)
+	if !ok {
+		t.Fatalf("expected detail to be map[string]any, got %T", exitErr.Detail.Detail)
 	}
-	if len(verr.Params) != 1 || verr.Params[0].Name != "--badflag" {
-		t.Errorf("params = %v, want one entry named --badflag", verr.Params)
+	if detail["unknown"] != "--badflag" {
+		t.Errorf("detail.unknown = %v, want --badflag", detail["unknown"])
+	}
+	if got, _ := detail["unknown_flags"].([]string); len(got) != 1 || got[0] != "--badflag" {
+		t.Errorf("detail.unknown_flags = %v, want [--badflag]", detail["unknown_flags"])
+	}
+	for _, key := range []string{"suggestions", "valid_flags"} {
+		if _, present := detail[key]; !present {
+			t.Errorf("detail.%s missing; must be present (empty) to match the unknown_flag schema", key)
+		}
 	}
 }
 
@@ -164,21 +172,25 @@ func TestUnknownSubcommandRunE_ValidFlagWithoutSubcommandIsStructured(t *testing
 	if err == nil {
 		t.Fatal("expected a structured missing_subcommand error, got nil (help fallthrough)")
 	}
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	if output.ExitCodeOf(err) != output.ExitValidation {
-		t.Errorf("exit code = %d, want %d", output.ExitCodeOf(err), output.ExitValidation)
+	if exitErr.Code != output.ExitValidation {
+		t.Errorf("exit code = %d, want %d", exitErr.Code, output.ExitValidation)
 	}
-	if !strings.Contains(verr.Message, "missing subcommand") {
-		t.Errorf("message = %q, want it to mention a missing subcommand", verr.Message)
+	if exitErr.Detail == nil || exitErr.Detail.Type != "missing_subcommand" {
+		t.Fatalf("detail.Type = %v, want missing_subcommand", exitErr.Detail)
 	}
-	if len(verr.Params) != 1 || verr.Params[0].Name != "--query" {
-		t.Errorf("params = %v, want one entry named --query", verr.Params)
+	detail, ok := exitErr.Detail.Detail.(map[string]any)
+	if !ok {
+		t.Fatalf("detail is not a map: %#v", exitErr.Detail.Detail)
 	}
-	if !strings.Contains(verr.Message, "lark-cli drive") {
-		t.Errorf("message = %q, want it to name the group path", verr.Message)
+	if flags, _ := detail["flags"].([]string); len(flags) != 1 || flags[0] != "--query" {
+		t.Errorf("detail.flags = %v, want [--query]", detail["flags"])
+	}
+	if detail["command_path"] != "lark-cli drive" {
+		t.Errorf("detail.command_path = %v, want lark-cli drive", detail["command_path"])
 	}
 }
 
@@ -229,23 +241,45 @@ func TestUnknownSubcommandRunE_UnknownReturnsStructuredError(t *testing.T) {
 		t.Fatal("expected error for unknown subcommand")
 	}
 
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	if output.ExitCodeOf(err) != output.ExitValidation {
-		t.Errorf("expected exit code %d, got %d", output.ExitValidation, output.ExitCodeOf(err))
+	if exitErr.Code != output.ExitValidation {
+		t.Errorf("expected exit code %d, got %d", output.ExitValidation, exitErr.Code)
 	}
-	if !strings.Contains(verr.Message, `"+bogus"`) {
-		t.Errorf("message should echo the unknown token, got %q", verr.Message)
+	if exitErr.Detail == nil {
+		t.Fatal("expected ExitError to carry Detail")
 	}
-	if !strings.Contains(verr.Message, "lark-cli drive") {
-		t.Errorf("message should name the group path, got %q", verr.Message)
+	if exitErr.Detail.Type != "unknown_subcommand" {
+		t.Errorf("expected Detail.Type=unknown_subcommand, got %q", exitErr.Detail.Type)
+	}
+	if !strings.Contains(exitErr.Detail.Message, `"+bogus"`) {
+		t.Errorf("message should echo the unknown token, got %q", exitErr.Detail.Message)
 	}
 	// "+bogus" has no close neighbor among drive's subcommands, so the hint falls
-	// back to pointing at --help (suggestions, when present, are folded into hint).
-	if !strings.Contains(verr.Hint, "--help") {
-		t.Errorf("hint should guide to --help when there is no suggestion, got %q", verr.Hint)
+	// back to pointing at --help; the full machine-readable list lives in
+	// detail.available below (which also excludes hidden commands).
+	if !strings.Contains(exitErr.Detail.Hint, "--help") {
+		t.Errorf("hint should guide to --help when there is no suggestion, got %q", exitErr.Detail.Hint)
+	}
+
+	detail, ok := exitErr.Detail.Detail.(map[string]any)
+	if !ok {
+		t.Fatalf("expected Detail.Detail to be map[string]any, got %T", exitErr.Detail.Detail)
+	}
+	if detail["unknown"] != "+bogus" {
+		t.Errorf("detail.unknown should be +bogus, got %v", detail["unknown"])
+	}
+	if detail["command_path"] != "lark-cli drive" {
+		t.Errorf("detail.command_path should be %q, got %v", "lark-cli drive", detail["command_path"])
+	}
+	available, ok := detail["available"].([]string)
+	if !ok {
+		t.Fatalf("detail.available should be []string, got %T", detail["available"])
+	}
+	if len(available) != 3 {
+		t.Errorf("expected 3 available entries (hidden excluded), got %d: %v", len(available), available)
 	}
 }
 
@@ -254,12 +288,13 @@ func TestUnknownSubcommandRunE_NestedResourceGroup(t *testing.T) {
 	installUnknownSubcommandGuard(root)
 
 	err := files.RunE(files, []string{"bogus"})
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError on nested group, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError on nested group, got %T", err)
 	}
-	if !strings.Contains(verr.Message, "lark-cli drive files") {
-		t.Errorf("message should reflect the nested resource path, got %q", verr.Message)
+	if exitErr.Detail.Detail.(map[string]any)["command_path"] != "lark-cli drive files" {
+		t.Errorf("command_path should reflect the nested resource, got %v",
+			exitErr.Detail.Detail.(map[string]any)["command_path"])
 	}
 }
 
@@ -302,10 +337,10 @@ func TestAvailableSubcommandNames_SplitsDeprecatedGroup(t *testing.T) {
 	}
 }
 
-// unknownSubcommandRunE ranks suggestions across both current and deprecated
-// subcommands so a mistyped legacy alias resolves; the closest match is folded
-// into the hint.
-func TestUnknownSubcommandRunE_SuggestsAcrossDeprecatedBucket(t *testing.T) {
+// unknownSubcommandRunE must split current vs deprecated subcommands into
+// separate detail buckets, while suggestions still rank across both so a
+// mistyped legacy alias resolves.
+func TestUnknownSubcommandRunE_SplitsDeprecatedBucket(t *testing.T) {
 	svc := &cobra.Command{Use: "sheets"}
 	svc.AddGroup(&cobra.Group{ID: cmdutil.DeprecatedGroupID, Title: "Deprecated"})
 	svc.AddCommand(
@@ -314,26 +349,31 @@ func TestUnknownSubcommandRunE_SuggestsAcrossDeprecatedBucket(t *testing.T) {
 	)
 
 	err := unknownSubcommandRunE(svc, []string{"+reat"})
-	var verr *errs.ValidationError
-	if !errors.As(err, &verr) {
-		t.Fatalf("expected *errs.ValidationError, got %T", err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T", err)
 	}
-	// "+reat" is closest to the deprecated +read: the candidate must surface
-	// both as a machine-readable param suggestion (for agent retry) and in the
-	// hint, proving ranking spans the deprecated bucket.
-	if len(verr.Params) != 1 || verr.Params[0].Name != "+reat" {
-		t.Fatalf("params = %v, want one entry named +reat (the offending subcommand)", verr.Params)
+	detail, ok := exitErr.Detail.Detail.(map[string]any)
+	if !ok {
+		t.Fatalf("detail is not a map: %#v", exitErr.Detail.Detail)
 	}
-	foundSuggestion := false
-	for _, s := range verr.Params[0].Suggestions {
+
+	if available, _ := detail["available"].([]string); len(available) != 1 || available[0] != "+cells-get" {
+		t.Errorf("available = %v, want [+cells-get]", available)
+	}
+	deprecated, ok := detail["deprecated"].([]string)
+	if !ok || len(deprecated) != 1 || deprecated[0] != "+read" {
+		t.Errorf("deprecated = %v, want [+read]", deprecated)
+	}
+	// suggestions rank across both buckets: "+reat" is closest to +read.
+	suggestions, _ := detail["suggestions"].([]string)
+	found := false
+	for _, s := range suggestions {
 		if s == "+read" {
-			foundSuggestion = true
+			found = true
 		}
 	}
-	if !foundSuggestion {
-		t.Errorf("Params[0].Suggestions should include +read, got %v", verr.Params[0].Suggestions)
-	}
-	if !strings.Contains(verr.Hint, "+read") {
-		t.Errorf("hint %q should suggest +read (typo target across deprecated bucket)", verr.Hint)
+	if !found {
+		t.Errorf("suggestions %v should include +read (typo target)", suggestions)
 	}
 }

cmd/update/update.go

@@ -10,7 +10,6 @@ import (
 
 	"github.com/spf13/cobra"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/build"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/output"
@@ -133,14 +132,12 @@ func updateRun(opts *UpdateOptions) error {
 	// 1. Fetch latest version
 	latest, err := fetchLatest()
 	if err != nil {
-		return reportError(opts, io, "network",
-			errs.NewNetworkError(errs.SubtypeNetworkTransport, "failed to check latest version: %s", err).WithCause(err))
+		return reportError(opts, io, output.ExitNetwork, "network", "failed to check latest version: %s", err)
 	}
 
 	// 2. Validate version format
 	if update.ParseVersion(latest) == nil {
-		return reportError(opts, io, "update_error",
-			errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid version from registry: %s", latest))
+		return reportError(opts, io, output.ExitInternal, "update_error", "invalid version from registry: %s", latest)
 	}
 
 	// 3. Compare versions
@@ -169,18 +166,15 @@ func updateRun(opts *UpdateOptions) error {
 
 // --- Output helpers ---
 
-// reportError emits the failure on the requested surface: JSON mode prints the
-// {ok:false, error:{type, message}} envelope to stdout and signals the typed
-// error's exit code bare; human mode returns the typed error for the
-// dispatcher to render.
-func reportError(opts *UpdateOptions, io *cmdutil.IOStreams, errType string, typedErr errs.TypedError) error {
+func reportError(opts *UpdateOptions, io *cmdutil.IOStreams, exitCode int, errType, format string, args ...interface{}) error {
+	msg := fmt.Sprintf(format, args...)
 	if opts.JSON {
 		output.PrintJson(io.Out, map[string]interface{}{
-			"ok": false, "error": map[string]interface{}{"type": errType, "message": typedErr.ProblemDetail().Message},
+			"ok": false, "error": map[string]interface{}{"type": errType, "message": msg},
 		})
-		return output.ErrBare(output.ExitCodeOf(typedErr))
+		return output.ErrBare(exitCode)
 	}
-	return typedErr
+	return output.Errorf(exitCode, errType, "%s", msg)
 }
 
 func reportCheckResult(opts *UpdateOptions, io *cmdutil.IOStreams, cur, latest string, canAutoUpdate bool) error {
@@ -234,8 +228,7 @@ func doManualUpdate(opts *UpdateOptions, io *cmdutil.IOStreams, cur, latest stri
 func doNpmUpdate(opts *UpdateOptions, io *cmdutil.IOStreams, cur, latest string, updater *selfupdate.Updater) error {
 	restore, err := updater.PrepareSelfReplace()
 	if err != nil {
-		return reportError(opts, io, "update_error",
-			errs.NewAPIError(errs.SubtypeUnknown, "failed to prepare update: %s", err).WithCause(err))
+		return reportError(opts, io, output.ExitAPI, "update_error", "failed to prepare update: %s", err)
 	}
 
 	if !opts.JSON {

cmd/update/update_test.go

@@ -14,7 +14,6 @@ import (
 	"testing"
 	"time"
 
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/output"
@@ -335,88 +334,13 @@ func TestUpdateFetchError_Human(t *testing.T) {
 	if err == nil {
 		t.Fatal("expected non-nil error, got nil")
 	}
-	var netErr *errs.NetworkError
-	if !errors.As(err, &netErr) {
-		t.Fatalf("expected *errs.NetworkError, got %T: %v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
 	}
-	if netErr.Subtype != errs.SubtypeNetworkTransport {
-		t.Errorf("subtype = %q, want %q", netErr.Subtype, errs.SubtypeNetworkTransport)
+	if exitErr.Code != output.ExitNetwork {
+		t.Errorf("expected ExitNetwork (%d), got %d", output.ExitNetwork, exitErr.Code)
 	}
-	if got := output.ExitCodeOf(err); got != output.ExitNetwork {
-		t.Errorf("expected ExitNetwork (%d), got %d", output.ExitNetwork, got)
-	}
-}
-
-// TestUpdateInvalidVersion_Human verifies a malformed registry version surfaces
-// as a typed internal error in human mode, keeping the legacy exit code 5.
-func TestUpdateInvalidVersion_Human(t *testing.T) {
-	f, _, _ := newTestFactory(t)
-	cmd := NewCmdUpdate(f)
-	cmd.SetArgs([]string{})
-
-	origFetch := fetchLatest
-	fetchLatest = func() (string, error) { return "not-a-version", nil }
-	defer func() { fetchLatest = origFetch }()
-
-	cmd.SilenceErrors = true
-	cmd.SilenceUsage = true
-
-	err := cmd.Execute()
-	if err == nil {
-		t.Fatal("expected non-nil error, got nil")
-	}
-	var intErr *errs.InternalError
-	if !errors.As(err, &intErr) {
-		t.Fatalf("expected *errs.InternalError, got %T: %v", err, err)
-	}
-	if intErr.Subtype != errs.SubtypeInvalidResponse {
-		t.Errorf("subtype = %q, want %q", intErr.Subtype, errs.SubtypeInvalidResponse)
-	}
-	if got := output.ExitCodeOf(err); got != output.ExitInternal {
-		t.Errorf("expected ExitInternal (%d), got %d", output.ExitInternal, got)
-	}
-}
-
-// TestReportError pins reportError's two surfaces after the typed migration:
-// human mode returns the typed error unchanged; JSON mode prints the legacy
-// {ok:false, error:{type, message}} envelope and exits bare with the typed
-// error's exit code (parity with the legacy explicit exit-code argument).
-func TestReportError(t *testing.T) {
-	t.Run("human mode returns the typed error", func(t *testing.T) {
-		f, _, _ := newTestFactory(t)
-		typed := errs.NewAPIError(errs.SubtypeUnknown, "failed to prepare update: disk full")
-		err := reportError(&UpdateOptions{JSON: false}, f.IOStreams, "update_error", typed)
-		var apiErr *errs.APIError
-		if !errors.As(err, &apiErr) {
-			t.Fatalf("expected *errs.APIError, got %T: %v", err, err)
-		}
-		if apiErr != typed {
-			t.Errorf("reportError must return the typed error unchanged")
-		}
-		if got := output.ExitCodeOf(err); got != output.ExitAPI {
-			t.Errorf("exit code = %d, want %d (ExitAPI, legacy parity)", got, output.ExitAPI)
-		}
-	})
-
-	t.Run("json mode prints envelope and exits bare with typed code", func(t *testing.T) {
-		f, stdout, _ := newTestFactory(t)
-		typed := errs.NewNetworkError(errs.SubtypeNetworkTransport, "failed to check latest version: timeout")
-		err := reportError(&UpdateOptions{JSON: true}, f.IOStreams, "network", typed)
-		var bareErr *output.BareError
-		if !errors.As(err, &bareErr) {
-			t.Fatalf("expected bare *output.BareError, got %T: %v", err, err)
-		}
-		if bareErr.Code != output.ExitNetwork {
-			t.Errorf("bare exit code = %d, want %d", bareErr.Code, output.ExitNetwork)
-		}
-		out := stdout.String()
-		if !strings.Contains(out, `"type": "network"`) && !strings.Contains(out, `"type":"network"`) {
-			t.Errorf("JSON envelope missing type, got: %s", out)
-		}
-		if !strings.Contains(out, "failed to check latest version: timeout") {
-			t.Errorf("JSON envelope missing message, got: %s", out)
-		}
-	})
 }
 
 func TestUpdateInvalidVersion_JSON(t *testing.T) {
@@ -579,12 +503,12 @@ func TestUpdateNpmVerifyFail_JSON_NoRestoreHintWhenBackupUnavailable(t *testing.
 	if err == nil {
 		t.Fatal("expected verification failure")
 	}
-	var bareErr *output.BareError
-	if !errors.As(err, &bareErr) {
-		t.Fatalf("expected *output.BareError, got %T: %v", err, err)
+	var exitErr *output.ExitError
+	if !errors.As(err, &exitErr) {
+		t.Fatalf("expected *output.ExitError, got %T: %v", err, err)
 	}
-	if bareErr.Code != output.ExitAPI {
-		t.Fatalf("expected ExitAPI (%d), got %d", output.ExitAPI, bareErr.Code)
+	if exitErr.Code != output.ExitAPI {
+		t.Fatalf("expected ExitAPI (%d), got %d", output.ExitAPI, exitErr.Code)
 	}
 
 	out := stdout.String()

errs/ERROR_CONTRACT.md

@@ -6,16 +6,25 @@ envelope on stderr; **protocol adapters** mapping CLI errors into MCP /
 OAuth shapes; and **framework + business code** producing errors. This file
 is the single source of truth for all three.
 
-Something off in production? See **Troubleshooting**.
+This document describes the **typed authoring target**. The refactor lands
+in stages; some boundaries (e.g. `client.WrapDoAPIError`) still operate on
+legacy shapes today — see **Migration** for what is live in each stage.
+
+Migrating an `*output.ExitError` call site? See **Migration**. Something off
+in production? See **Troubleshooting**.
 
 ## Invariants
 
 1. Every error belongs to exactly one **Category**. The set is closed
    (`errs/category.go`); adding a member requires deliberate review.
-2. Every typed error has a **Subtype** — a stable
+2. Every **newly constructed** typed error has a **Subtype** — a stable
    lowercase-with-underscores identifier declared in `errs/subtypes*.go`.
-   Undeclared subtypes fail CI. Every error path constructs a typed
-   `*errs.*` error at its origin, so the constraint applies uniformly.
+   Undeclared subtypes fail CI. The constraint applies only to typed
+   `*errs.*` literals; stage-1 legacy `*core.ConfigError` flows via the
+   dispatcher's `asExitError` → legacy envelope path (not the typed
+   taxonomy) and is unaffected. `errcompat.PromoteConfigError` is a
+   stage-1 passthrough; its stage-2+ typed migration will subject the
+   promoted typed error to this Subtype constraint at that time.
 3. **`Category` + `Subtype`** are wire-stable identifiers consumers may
    branch on. Renaming either is a breaking change.
 4. `Code` is the upstream numeric code when known (e.g. Lark API code).
@@ -26,10 +35,11 @@ Something off in production? See **Troubleshooting**.
    unchanged across the `errors.As` / `errors.Unwrap` chain.
 7. For the typed-envelope path, exit codes derive from `Category` only
    via `output.ExitCodeForCategory` — including `SecurityPolicyError`,
-   which exits `6` via `CategoryPolicy`. `output.ErrBare(code)` is the
-   exception: it constructs an `*output.BareError`, a deliberate
-   silent-exit signal (stdout already carries the answer) that bypasses
-   the envelope (see **Predicate commands** below).
+   which exits `6` via `CategoryPolicy`. Unmigrated `*output.ExitError`
+   producers still carry a hand-set `Code` until they finish migrating.
+   `output.ErrBare(code)` is the lone exception: a deliberate
+   predicate-command signal that bypasses the envelope (see
+   **Predicate commands** below).
 
 ## Wire format
 
@@ -63,14 +73,13 @@ Typed errors render to **stderr** as one JSON object per process exit:
 | `error.hint` | informational | actionable recovery guidance |
 | `error.log_id` | informational | upstream request id (server-side trace) |
 | `error.retryable` | wire-stable | `true` when present; omitted when `false` |
-| `error.param` | per-Subtype-stable | single offending parameter (`ValidationError`); see **Validation parameters** |
-| `error.params` | per-Subtype-stable | per-parameter validation detail array (`ValidationError`); see **Validation parameters** |
 | per-Subtype extension fields | per-Subtype-stable | e.g. `missing_scopes`, `console_url`, `challenge_url` |
 
 `SecurityPolicyError` renders through the same typed envelope as every
 other category. `error.type` is `"policy"`, `error.subtype` is one of
 `challenge_required` / `access_denied`, and process exit is `6` via
-`CategoryPolicy`.
+`CategoryPolicy`. The legacy `auth_error` envelope at exit `1` has been
+retired.
 
 ## Categories
 
@@ -110,21 +119,20 @@ Canonical mapping: `internal/output/exitcode.go` `ExitCodeForCategory`.
      │
      ▼
   cmd/root.go handleRootError dispatches:
+     ├─ output.ErrBare(code)      → no envelope (stdout already written); exit = code
      ├─ typed (errs.ProblemOf)    → typed JSON envelope; exit = ExitCodeOf(err)
-     │     (includes *errs.SecurityPolicyError → policy envelope, exit 6;
-     │      *errs.ConfigError, constructed typed at origin)
-     ├─ *output.PartialFailureError → no stderr envelope (ok:false result already on stdout); exit = code
-     ├─ *output.BareError         → no envelope (stdout already written); exit = code
-     └─ Cobra usage error         → typed validation envelope (invalid_argument); exit 2
+     │     (includes *errs.SecurityPolicyError → policy envelope, exit 6)
+     ├─ *core.ConfigError         → promoted to typed via errcompat ↑
+     ├─ *output.ExitError         → legacy JSON envelope; exit = exitErr.Code
+     └─ untyped / Cobra error     → plain "Error: <msg>" (no envelope); exit 1
 ```
 
-The dispatcher emits a JSON envelope on stderr for both the typed branch and
-residual Cobra usage errors (missing required flag, unknown command,
-argument validation): the latter are classified into a typed validation
-envelope (`invalid_argument`) and exit `2`, matching the explicit flag and
-subcommand guards.
+Only the typed and `*output.ExitError` branches emit a JSON envelope on
+stderr. Untyped errors (including Cobra's "required flag missing" / unknown
+subcommand messages) print plain text and exit `1` — consumers must
+tolerate that fallback.
 
-### Predicate commands (`output.BareError`)
+### Predicate commands (`output.ErrBare`)
 
 A small class of commands is **predicates**: they answer a yes/no
 question and signal the answer through the shell exit code so callers
@@ -134,27 +142,19 @@ example — its `README` contract is `exit 0 = ok, 1 = missing`.
 These commands deliberately:
 
 1. write a structured JSON answer to **stdout** themselves, and
-2. return `output.ErrBare(exitCode)` — an `*output.BareError` — to
-   communicate the exit code to the dispatcher without producing a
-   `stderr` envelope.
+2. return `output.ErrBare(exitCode)` to communicate the exit code to
+   the dispatcher without producing a `stderr` envelope.
 
-`*output.BareError` is **not** an error in the typed-envelope sense — it
-carries no category, subtype, or message, only an exit code. It is a
-one-bit output-control signal that lives outside the contract for the
-same reason `grep -q` / `diff` / `systemctl is-active` set non-zero exit
-codes without printing anything to stderr: pollution of stderr by a
+`output.ErrBare` is **not** an error in the typed-envelope sense — it
+carries no category, subtype, or message. It is a one-bit output-
+control signal that lives outside the contract for the same reason
+`grep -q` / `diff` / `systemctl is-active` set non-zero exit codes
+without printing anything to stderr: pollution of stderr by a
 predicate's negative answer would break `2>/dev/null` log hygiene in
 caller scripts.
 
-A second class also uses `ErrBare`: a command that emits its own complete
-structured result envelope on **stdout** under `--json` (e.g. `update`, whose
-`{ok:false, error:{type, message}}` is its established output shape) and needs
-only the exit code conveyed, with no `stderr` envelope. Like a predicate, its
-answer is already on stdout; `ErrBare` carries the exit code alone.
-
-New code should not reach for `ErrBare` unless the command's full answer is
-already on stdout — a predicate's yes/no, or a self-contained result envelope
-as above. Anything whose error content must reach the caller on `stderr`
+New code should not reach for `ErrBare` unless the command is
+genuinely a predicate. Anything carrying recoverable error content
 belongs in a typed `*errs.XxxError` — or, for a batch result, in the
 partial-failure outcome below.
 
@@ -214,7 +214,7 @@ exitCode := output.ExitCodeOf(err) // ExitInternal for non-typed errors
 out=$(lark-cli ... 2>&1)
 code=$?
 
-# Defensive guard: tolerate any non-JSON output before parsing with jq.
+# Untyped / Cobra errors print plain text — guard before jq.
 if ! jq -e . >/dev/null 2>&1 <<<"$out"; then
     printf '%s\n' "$out" >&2
     exit "$code"
@@ -303,10 +303,9 @@ Do not pick exit codes by hand in new typed producers — `ExitCodeForCategory`
 maps `Category` to the shell code. A new exit-code requirement means a
 new `Category`, not a one-off override at the call site.
 
-(The only exits not derived from `Category` are the
-`*output.BareError` and the `*output.PartialFailureError` signals, which
-carry their own code by design and sit outside the typed-envelope contract —
-see **Predicate commands**.)
+(Legacy `*output.ExitError` retains hand-set codes until removal;
+`SecurityPolicyError` retains a hand-set code on main until the framework
+migration PR retires the carve-out — see **Migration**.)
 
 #### Split `Message`, `Hint`, and `Cause`
 
@@ -341,54 +340,15 @@ Message: fmt.Sprintf("request failed: %v — retry later", ioErr)
 // conflates what + what-to-do + cause into one string
 ```
 
-#### Validation parameters: `Param` and `Params`
+#### `ValidationError.Param` uses the `--flag` form
 
-`ValidationError` carries two additive parameter fields. Both are
-optional; a producer sets whichever fits the failure.
+When a `*ValidationError` originates from a flag value, `Param` holds the
+flag name with leading dashes (`"--priority"`, not `"priority"`). AI
+agents grep this field literally to surface "the bad flag was `--X`".
 
-**`Param string` (wire `param`)** — the single offending parameter. When a
-`*ValidationError` originates from a flag value, `Param` holds the flag
-name with leading dashes (`"--priority"`, not `"priority"`). AI agents
-grep this field literally to surface "the bad flag was `--X`". For
-positional arguments, use the canonical name without dashes
+For positional arguments, use the canonical name without dashes
 (`"target_user_id"`).
 
-**`Params []InvalidParam` (wire `params`)** — per-parameter validation
-detail, for failures that need to report *which* parameters failed and
-*why*, one entry each. Each `errs.InvalidParam` is
-`{Name, Reason string, Suggestions []string}`: `Name` identifies the
-parameter, `Reason` states why it failed, and the optional `Suggestions`
-(wire `suggestions`, omitted when empty) carries ranked candidate
-corrections an agent can retry with — the did-you-mean candidates for an
-unknown flag or subcommand — without parsing the human-facing `hint`. This
-is the CLI's rendering of the RFC 7807 `invalid-params` extension member
-(RFC 7807 §3.1). The wire key is `params`, not `invalid_params`: the
-enclosing envelope already carries `type:"validation"`, so the `invalid_`
-qualifier would be redundant on the wire.
-
-`Param` and `Params` are independent additive fields, not alternates of a
-single representation. Use `Param` for the common single-parameter error;
-use `Params` when one failure spans several parameters or needs a
-per-parameter reason. Set with `.WithParam("--flag")` / `.WithParams(...)`.
-
-A `params` wire example (multiple parameters each carrying a reason):
-
-```json
-{
-  "ok": false,
-  "identity": "user",
-  "error": {
-    "type": "validation",
-    "subtype": "invalid_argument",
-    "message": "2 parameters failed validation",
-    "params": [
-      { "name": "--start", "reason": "expected RFC3339, got \"yesterday\"" },
-      { "name": "--end", "reason": "must be after --start" }
-    ]
-  }
-}
-```
-
 ### Constructing typed errors
 
 Prefer the **builder API**. The constructor pins `Category` + `Subtype` +
@@ -418,11 +378,44 @@ them on the dynamic dispatch path where a `Problem` value is composed
 once and wrapped per Category branch. Outside that pattern, new code
 should reach for the builder.
 
-When the validation logic outgrows a single range check — multiple flags,
-format parsing, conditional rules — extract it into a helper that also returns
-the typed `*errs.ValidationError`; the helper, not `Execute`, sets `Param` (a
-helper bound to one shortcut is normal in this codebase; see `parseTimeRange`
-in `shortcuts/calendar/calendar_agenda.go`).
+Legacy helpers (`output.ErrValidation`, `output.ErrAuth`, `output.ErrNetwork`)
+remain callable during migration but are `// Deprecated:` — new code goes
+through the builder.
+
+#### Shortcut `Execute` walkthrough
+
+Adapted from `shortcuts/calendar/calendar_suggestion.go:222`, whose legacy
+form is `output.ErrValidation("--duration-minutes must be between 1 and
+1440")`. The typed migration target (builder form):
+
+```go
+Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
+    duration := runtime.Int("duration-minutes")
+    if duration < 1 || duration > 1440 {
+        return errs.NewValidationError(errs.SubtypeInvalidArgument,
+            "--duration-minutes must be between 1 and 1440, got %d", duration).
+            WithHint("pass a value in [1, 1440]").
+            WithParam("--duration-minutes")
+    }
+
+    _, err := runtime.DoAPI(req, opts)
+    if err != nil {
+        return err // already typed by the framework boundary; propagate
+    }
+    return nil
+}
+```
+
+Two patterns visible: a producer site (the typed `*errs.ValidationError`
+above) and a propagation site (the `return err` after `runtime.DoAPI`,
+applying [Propagate typed errors unchanged](#propagate-typed-errors-unchanged)).
+
+When the validation logic outgrows a single range check — multiple
+flags, format parsing, conditional rules — extract it into a helper that
+also returns the typed `*errs.ValidationError`. The helper, not
+`Execute`, sets `Param` (a helper bound to one shortcut is normal in
+this codebase; see `parseTimeRange` in
+`shortcuts/calendar/calendar_agenda.go:144`).
 
 ### Wrapping upstream errors
 
@@ -486,7 +479,7 @@ Rare; the existing structs cover the 9 Categories with room. If you must:
 
 1. In `errs/types.go`, add a new section with: the struct embedding `errs.Problem`, a nil-receiver-safe `Unwrap()` if it carries `Cause`, a `NewXxxError(subtype, format, args...)` constructor, and one chained `WithX` setter per extension field.
 2. Add an `IsXxx` predicate in `errs/predicates.go`.
-3. Add a wire-format pin in `errs/marshal_test.go` and a builder-chain pin in `errs/types_test.go`.
+3. Add a wire-format pin in `errs/marshal_test.go` and a builder-chain pin in `errs/types_builder_test.go`.
 
 `CheckProblemEmbed` enforces the `Problem` embed at lint time. New
 top-level wire fields are forbidden — per-Subtype data goes into the
@@ -495,33 +488,19 @@ top level.
 
 ## CI guards
 
-Two golangci-lint rules and the custom `errscontract` AST module enforce the
-contract; CI runs all three on every PR.
+| Check | Enforces | Where |
+|-------|----------|-------|
+| forbidigo | business path (`shortcuts/**`, `cmd/service/**`) must not call legacy `output.*` error constructors — route through the typed classifier | `.golangci.yml` |
+| `CheckProblemEmbed` | every exported `*Error` embeds `errs.Problem` | `lint/errscontract/` AST |
+| `CheckNoRegistrar` | no `mergeCodeMeta` / `RegisterServiceMap` from service code | `lint/errscontract/` AST |
+| `CheckAdHocSubtype` | `ad_hoc_*` Subtypes labeled for promotion (warn) | `lint/errscontract/` AST |
+| `CheckDeclaredSubtype` | every `Subtype:` value is a declared constant or `ad_hoc_*` | `lint/errscontract/` AST |
+| `CheckTypedErrorCompleteness` | every `*errs.<X>Error{Problem: errs.Problem{...}}` literal must set `Category`, `Subtype`, and `Message` | `lint/errscontract/` AST |
 
-**golangci-lint** — scopes are defined in `.golangci.yml` (not duplicated here,
-so this spec cannot drift from the lint config):
-
-| Rule | Enforces |
-|------|----------|
-| forbidigo `errs-no-bare-wrap` | a command / wire-boundary final error must be typed (`errs.NewXxxError`), never a bare `fmt.Errorf` / `errors.New`; a genuine intermediate wrap opts out with `//nolint:forbidigo` + a reason |
-| errorlint | every error wrap uses `%w` and every comparison uses `errors.Is` / `errors.As` — interior wraps stay legal but cannot break the `errors.Unwrap` chain the typed boundary relies on |
-
-**errscontract** (`lint/errscontract/`, a separate Go module so its
-`golang.org/x/tools` dependency stays out of the shipped binary; run locally
-with `go run -C lint . ..`):
-
-| Check | Enforces |
-|-------|----------|
-| `CheckNoLegacyEnvelopeLiteral` / `CheckNoLegacyCommonHelperCall` / `CheckNoLegacyRuntimeAPICall` | the removed `output.*` legacy error surface cannot be reintroduced anywhere |
-| `CheckProblemEmbed` | every exported `*Error` embeds `errs.Problem` |
-| `CheckDeclaredSubtype` | every `Subtype:` value is a declared constant (or `ad_hoc_*`) |
-| `CheckTypedErrorCompleteness` | every typed-error struct literal sets `Category`, `Subtype`, and `Message` |
-| `CheckAdHocSubtype` | `ad_hoc_*` Subtypes flagged for promotion (warning) |
-| `CheckNoRegistrar` | no `mergeCodeMeta` / `RegisterServiceMap` from service code |
-
-`errscontract` also carries framework-internal invariants (nil-safe `Unwrap`,
-builder immutability, unwrap symmetry); see `lint/errscontract/` for the full
-set and `lint/README.md` for adding a new lint domain.
+CI runs `lint/` on every PR. Locally: `go run -C lint . ..`. The
+lintcheck CLI lives in its own Go module so its `golang.org/x/tools`
+dependency stays out of the shipped `lark-cli` binary's module graph;
+see `lint/README.md` for how to add a new lint domain.
 
 ## Stability
 
@@ -531,13 +510,67 @@ set and `lint/README.md` for adding a new lint domain.
 | Additive | new Category, new declared Subtype, new extension field on an existing struct | minor release; consumers ignore unknown fields by contract |
 | Experimental | `ad_hoc_*` Subtypes; fields documented as such in `errs/types.go` | may change or be promoted/removed within one release |
 
+The deprecated `*output.ExitError` surface is outside these tiers — it
+will be removed once business migration completes.
+
+## Migration
+
+**Strategy shift (2026-05-26).** The original plan (`docs/design/errors-refactor/spec.md` v2.12 §9) was a centrally-driven 4-PR rollout — framework → auth domain → multi-pilot → full-repo + legacy removal. That plan is **superseded** by a hybrid model: framework owner ships framework-level hardening (including a typed `*errs.*Error` migration of `internal/**`) as one focused PR; business-domain typed migration is **self-service** via [`docs/errors-guide.md`](../docs/errors-guide.md) and the builder API, with no central sweep timeline.
+
+Why the shift: 800+ legacy call sites split across 8+ business domains do not all share a single reviewer's bandwidth, and the contract is now expressive enough that each domain owner can migrate their own code from the guide without coordinating with framework owner.
+
+### Current state
+
+1. **Framework slice — ✅ shipped (PR #984).** The `errs/` typed taxonomy, classifier (`internal/errclass`), promotion stub (`internal/errcompat`, passthrough), dispatcher hook (`WriteTypedErrorEnvelope`), and the `lint/errscontract` AST guards. Wire shapes preserved byte-for-byte versus pre-PR, with **one intentional semantic fix**: config-class errors (`*core.ConfigError`) now exit `3` instead of `2`, aligning with `ExitCodeForCategory` (config errors share the auth exit slot per the taxonomy). The classifier and promote helpers are *shipped but unused* in production paths — they exist so framework migration can plug in without re-architecting.
+
+2. **Builder API — ✅ shipped (this branch).** `errs/types.go` adds the canonical producer surface (`errs.NewXxxError(subtype, format, args...).WithX(...)`) for all 10 typed types, alongside each struct declaration. Constructor signature pins `Category` (via function name) and `Subtype` + `Message` (positional), so the producer cannot mis-specify any of the three identity fields. Optional fields chain through `.WithX(...)` setters that preserve the concrete pointer type.
+
+### Next: framework migration PR (planned)
+
+A single PR consolidates the work the original §9 spec split across PRs 2–4 — restricted to framework code, no business sweep:
+
+- **Migrate `internal/**` typed construction to the builder API.** ~16 call sites in `internal/errclass/classify.go` (BuildAPIError fanout), `internal/auth/transport.go` (SecurityPolicy), `internal/auth/uat_client.go`, `internal/errcompat/promote*.go`, `internal/client/client.go`, `internal/client/api_errors.go`.
+- **Land the framework-side semantic changes** previously scoped to spec §9 PR 2: `SecurityPolicyError` exit `1→6`, `WrapDoAPIError` typed (`*NetworkError` with subtype timeout/tls/dns/server_error/transport, `*InternalError` for JSON-decode), `WrapJSONResponseParseError` typed, `errcompat.PromoteConfigError` real Type routing, `PromoteAuthError` helper + dispatcher wiring, 10 credential Lark codes registered in codeMeta, 99991543 config classification, `resolveAccessToken` typed `*AuthenticationError`, `BuildAPIError` filling `*PermissionError.MissingScopes` / `Identity` / `ConsoleURL`, deletion of `scopeAwareChecker`.
+- **Add `forbidigo` rule** banning `output.Err*` constructors in `shortcuts/**` and `cmd/**` (mirrors the contract that new business code must use the builder).
+- **CHANGELOG** lists the resulting ~10 shell-exit-code shifts in one release entry (vs the spec §1 spread of 11 — the remaining one site lives in `task` business code).
+
+### Business-domain migration (self-service, no central timeline)
+
+Each business package migrates its own `output.Err*` call sites to the builder when convenient — typically batched within one domain. The guide at [`docs/errors-guide.md`](../docs/errors-guide.md) walks owners through the 8 typical error modes (validation / authorization / authentication / config / network / api / internal / policy) with real `file:line` examples from main. The three-layer extension model (add Subtype / add field / add Category) handles cases the existing taxonomy does not cover.
+
+Helper assertions accept both shapes during migration (see `shortcuts/mail/mail_shortcut_validation_test.go` `assertValidationError`) so domain migrations stay green incrementally.
+
+### Legacy removal
+
+Deferred until business migration completion approaches the asymptote. `Errorf`, `ErrAPI`, `ErrAuth`, `ErrWithHint`, `ErrBare`, `ClassifyLarkError`, `ErrDetail`, `ExitError`, and `ErrorEnvelope` are `// Deprecated:` today and stay callable. No fixed removal date.
+
+### Before / after at a call site
+
+```go
+// before (legacy)
+return output.ErrAPI(larkCode, "create event failed", resp.RawBody())
+
+// after (typed) — cc carries Brand / AppID / Identity from the caller's context
+return errclass.BuildAPIError(parsedResp, cc)
+```
+
+```go
+// before (legacy validation)
+return output.ErrValidation("--duration-minutes must be between 1 and 1440")
+
+// after (builder)
+return errs.NewValidationError(errs.SubtypeInvalidArgument,
+    "--duration-minutes must be between 1 and 1440, got %d", duration).
+    WithParam("--duration-minutes")
+```
+
 ## Troubleshooting
 
 **Envelope shows `type=api subtype=unknown` for what should be a more
 specific category.** The Lark code is unknown to `LookupCodeMeta` and fell
 through to the generic bucket (`internal/errclass/classify.go`). Add the
 code to `internal/errclass/codemeta_<service>.go` with the right Category
-and Subtype, plus a dispatch test in `internal/errclass/classify_test.go`.
+and Subtype, plus a dispatch test in `classify_test.go`.
 
 **Envelope shows `type=internal subtype=sdk_error`.** Origin is
 `client.WrapDoAPIError` taking the non-transport branch
@@ -580,6 +613,8 @@ string cannot be classified retroactively.
 - *Add a new condition?* → **Add a Subtype**
 - *Consume from a shell script?* → **Consumers / Shell / AI**
 - *Understand or fix a CI failure?* → **CI guards**
+- *Migrate a legacy `ExitError` call site?* → **Migration** + the
+  Deprecated note on the symbol being replaced.
 - *Read source.* → `errs/doc.go` → `errs/category.go` → `errs/types.go`
   → `errs/predicates.go` → `internal/errclass/` →
   `cmd/root.go` `handleRootError`.

errs/raw.go

@@ -1,29 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package errs
-
-import "errors"
-
-// rawPassthrough marks an error as raw passthrough: the dispatcher must not
-// rewrite its message or hint with local enrichment. Raw is
-// dispatcher-internal routing state, not a wire field. It is deliberately not
-// a typed taxonomy error (no embedded Problem) — it only wraps one.
-type rawPassthrough struct{ err error }
-
-func (e *rawPassthrough) Error() string { return e.err.Error() }
-func (e *rawPassthrough) Unwrap() error { return e.err }
-
-// MarkRaw wraps err as raw passthrough. MarkRaw(nil) returns nil.
-func MarkRaw(err error) error {
-	if err == nil {
-		return nil
-	}
-	return &rawPassthrough{err: err}
-}
-
-// IsRaw reports whether err or any error in its chain is marked raw.
-func IsRaw(err error) bool {
-	var raw *rawPassthrough
-	return errors.As(err, &raw)
-}

errs/raw_test.go

@@ -1,96 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package errs_test
-
-import (
-	"encoding/json"
-	"errors"
-	"fmt"
-	"testing"
-
-	"github.com/larksuite/cli/errs"
-)
-
-func TestMarkRawNilReturnsNil(t *testing.T) {
-	if got := errs.MarkRaw(nil); got != nil {
-		t.Fatalf("MarkRaw(nil) = %v, want nil", got)
-	}
-}
-
-func TestIsRaw(t *testing.T) {
-	base := fmt.Errorf("boom")
-
-	if !errs.IsRaw(errs.MarkRaw(base)) {
-		t.Errorf("IsRaw(MarkRaw(err)) = false, want true")
-	}
-	if errs.IsRaw(base) {
-		t.Errorf("IsRaw(bare err) = true, want false")
-	}
-	if errs.IsRaw(nil) {
-		t.Errorf("IsRaw(nil) = true, want false")
-	}
-
-	// Raw marking survives further wrapping above it in the chain.
-	wrapped := fmt.Errorf("outer: %w", errs.MarkRaw(base))
-	if !errs.IsRaw(wrapped) {
-		t.Errorf("IsRaw(wrap(MarkRaw(err))) = false, want true")
-	}
-}
-
-func TestMarkRawPreservesErrorMessage(t *testing.T) {
-	base := fmt.Errorf("boom")
-	if got := errs.MarkRaw(base).Error(); got != "boom" {
-		t.Fatalf("MarkRaw(err).Error() = %q, want %q", got, "boom")
-	}
-}
-
-func TestMarkRawPreservesErrorsIsChain(t *testing.T) {
-	sentinel := errors.New("sentinel")
-	wrapped := fmt.Errorf("ctx: %w", sentinel)
-
-	if !errors.Is(errs.MarkRaw(wrapped), sentinel) {
-		t.Fatalf("errors.Is(MarkRaw(err), sentinel) = false, want true")
-	}
-}
-
-func TestProblemOfPunchesThroughMarkRaw(t *testing.T) {
-	typed := errs.NewValidationError(errs.SubtypeInvalidArgument, "bad flag")
-	raw := errs.MarkRaw(typed)
-
-	p, ok := errs.ProblemOf(raw)
-	if !ok {
-		t.Fatalf("ProblemOf(MarkRaw(typed)) ok = false, want true")
-	}
-	if p.Category != errs.CategoryValidation {
-		t.Errorf("ProblemOf(MarkRaw(typed)).Category = %v, want %v", p.Category, errs.CategoryValidation)
-	}
-
-	// errors.As still finds the concrete typed error through the raw wrapper.
-	var ve *errs.ValidationError
-	if !errors.As(raw, &ve) {
-		t.Errorf("errors.As(MarkRaw(typed), *ValidationError) = false, want true")
-	}
-}
-
-// TestMarkRawUnwrapsToInnerTypedError pins the envelope-serialization
-// contract: UnwrapTypedError must return the inner concrete typed error,
-// not the rawPassthrough wrapper. The wrapper has no exported fields, so if it
-// were returned the JSON envelope would marshal to an empty "{}" error.
-func TestMarkRawUnwrapsToInnerTypedError(t *testing.T) {
-	base := errs.NewValidationError(errs.SubtypeInvalidArgument, "bad flag")
-	typed, ok := errs.UnwrapTypedError(errs.MarkRaw(base))
-	if !ok {
-		t.Fatal("UnwrapTypedError(MarkRaw(typed)) must find a typed error")
-	}
-	out, err := json.Marshal(typed)
-	if err != nil {
-		t.Fatal(err)
-	}
-	if string(out) == "{}" {
-		t.Fatalf("UnwrapTypedError returned the opaque rawPassthrough wrapper; envelope would be empty: %s", out)
-	}
-	if got := errs.CategoryOf(typed); got != errs.CategoryValidation {
-		t.Fatalf("unwrapped category = %q, want validation", got)
-	}
-}

errs/types.go

@@ -77,10 +77,6 @@ type ValidationError struct {
 type InvalidParam struct {
 	Name   string `json:"name"`
 	Reason string `json:"reason"`
-	// Suggestions holds machine-readable, ranked candidate corrections for this
-	// parameter (e.g. did-you-mean flags or subcommands), so an agent can retry
-	// without parsing the human-facing hint. Omitted when there are none.
-	Suggestions []string `json:"suggestions,omitempty"`
 }
 
 // Unwrap exposes the wrapped cause so errors.Unwrap / errors.Is can traverse

errs/types_test.go

@@ -101,9 +101,9 @@ func TestSecurityPolicyErrorUnwrap(t *testing.T) {
 // interface would panic when the root dispatcher or any caller walks the
 // errors.Is / errors.Unwrap chain.
 //
-// The doc comments on these types claim "nil-receiver safe"; this test
-// pins that claim so the behavioral comment cannot silently drift from the
-// implementation.
+// The doc comments on these types claim "nil-receiver safe" but until this
+// test landed nothing actually pinned that claim — exactly the
+// behavioral-comment-without-test footgun caught in PR #984 review.
 func TestTypedErrors_UnwrapNilReceiver(t *testing.T) {
 	t.Helper()
 	checks := []struct {

events/im/message_receive.go

@@ -23,7 +23,7 @@ type ImMessageReceiveOutput struct {
 	ChatType    string `json:"chat_type,omitempty"    desc:"Conversation type"                                                                                                                                                                   enum:"p2p,group"`
 	MessageType string `json:"message_type,omitempty" desc:"Message type"`
 	SenderID    string `json:"sender_id,omitempty"    desc:"Sender open_id; prefixed with ou_"                                                                                                                                                   kind:"open_id"`
-	Content     string `json:"content,omitempty"      desc:"Message content. For most types (text/post/image/file/audio, etc.) this is pre-rendered human-readable text."`
+	Content     string `json:"content,omitempty"      desc:"Message content. For most types (text/post/image/file/audio, etc.) this is pre-rendered human-readable text. For interactive (cards) it stays as the raw JSON string and callers must fromjson to parse it."`
 }
 
 func processImMessageReceive(_ context.Context, _ event.APIClient, raw *event.RawEvent, _ map[string]string) (json.RawMessage, error) {
@@ -55,10 +55,8 @@ func processImMessageReceive(_ context.Context, _ event.APIClient, raw *event.Ra
 	}
 
 	msg := envelope.Event.Message
-	var content string
-	if msg.MessageType == "interactive" {
-		content = convertlib.ConvertInteractiveEventContent(msg.Content, msg.Mentions)
-	} else {
+	content := msg.Content
+	if msg.MessageType != "interactive" {
 		content = convertlib.ConvertBodyContent(msg.MessageType, &convertlib.ConvertContext{
 			RawContent: msg.Content,
 			MentionMap: convertlib.BuildMentionKeyMap(msg.Mentions),

extension/platform/abort.go

@@ -7,8 +7,8 @@ import "fmt"
 
 // AbortError is returned by a Wrapper that wants to short-circuit the
 // command chain (instead of calling next). The framework converts it
-// to a typed errs.* error so the JSON envelope carries the structured
-// fields agents expect.
+// to an *output.ExitError with type "hook" so the JSON envelope carries
+// the structured fields agents expect.
 //
 // HookName is the framework-namespaced name ("secaudit.approval"); the
 // Registrar adds the plugin-name prefix automatically.

extension/platform/errors.go

@@ -7,9 +7,9 @@ import "fmt"
 
 // CommandDeniedError is the structured error returned by a denyStub. Every
 // pruned-command execution path -- direct invocation, alias expansion,
-// internal call -- returns this exact type. The dispatcher converts it to a
-// typed errs.* error; the Layer field carries the denial layer for the
-// envelope.
+// internal call -- returns this exact type. It is wire-compatible with the
+// output.ExitError envelope via the Layer (== error.type) field and the
+// detail map produced by ExitError().
 //
 // Layer values:
 //

harness-opt/.gitignore

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

harness-opt/baseline/baseline-tokens.json

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

harness-opt/baseline/noise-floor.json

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

harness-opt/baseline/summary.json

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

harness-opt/coverage.json

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

harness-opt/objective.json

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

harness-opt/opt-state.json

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

harness-opt/opt-state.md

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

harness-opt/pool/head.json

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

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

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

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

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

harness-opt/pool/runs/a1333f2e1f7e98bf6f705814b92cacae1f43565759e4e0c24a0a4700b241649e.json

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

harness-opt/pool/runs/phi0.json

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

harness-opt/trend.json

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

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

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