feat: 更新 flag

skip-hidden defaults to false (lossless reads), but the read primitives don't mark which rows/cols are hidden. Cross-reference +sheet-info --include hidden_rows,hidden_cols + row_indices/col_indices so agents can identify hidden ranges when they need to filter or interpret hidden data.

Synced from sheet-skill-spec.

cmd/api/api_test.go

@@ -66,6 +66,24 @@ func TestApiCmd_DryRun(t *testing.T) {
 	}
 }
 
+// Regression: --params null parses to a nil map; writing page_size onto it must
+// not panic. Symmetric to the typed-flag overlay path in cmd/service — both
+// write into the map ParseJSONMap returns.
+func TestApiCmd_NullParamsWithPageSize(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
+		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
+	})
+
+	cmd := NewCmdApi(f, nil)
+	cmd.SetArgs([]string{"GET", "/open-apis/test", "--params", "null", "--page-size", "50", "--as", "bot", "--dry-run"})
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("--params null with --page-size should not error, got: %v", err)
+	}
+	if out := stdout.String(); !strings.Contains(out, "page_size") {
+		t.Errorf("expected page_size applied over null --params, got:\n%s", out)
+	}
+}
+
 func TestApiCmd_BotMode(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,

cmd/auth/login_interactive.go

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

cmd/command_catalog_path_test.go

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

cmd/error_auth_hint.go

@@ -11,6 +11,7 @@ import (
 	"github.com/spf13/cobra"
 
 	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/apicatalog"
 	internalauth "github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
@@ -118,38 +119,37 @@ func resolveDeclaredShortcutScopes(cmd *cobra.Command, identity string) []string
 }
 
 // resolveDeclaredServiceMethodScopes returns the scopes declared by a
-// service/resource/method command from the embedded from_meta registry.
+// service/resource/method command. It reconstructs the catalog path from the
+// command ancestry and resolves it through the same navigation Module the
+// command tree is built from (apicatalog), so it stays correct for nested
+// resources instead of hard-coding a root->service->resource->method depth.
+// Non-method commands (services, resources, shortcuts) resolve to a non-method
+// target and yield no scopes.
 func resolveDeclaredServiceMethodScopes(cmd *cobra.Command, identity string) []string {
-	// Service-method scope lookup only applies to commands mounted as
-	// root -> service -> resource -> method. Non-resource/method commands
-	// intentionally return no scopes here so auth-hint enrichment does not
-	// change runtime semantics for other command shapes.
-	if cmd == nil || cmd.Parent() == nil || cmd.Parent().Parent() == nil || cmd.Parent().Parent().Parent() == nil {
+	if cmd == nil || strings.HasPrefix(cmd.Name(), "+") {
 		return nil
 	}
-	if strings.HasPrefix(cmd.Name(), "+") {
+	path := commandCatalogPath(cmd)
+	if len(path) == 0 {
 		return nil
 	}
+	target, err := registry.RuntimeCatalog().Resolve(path)
+	if err != nil || target.Kind != apicatalog.TargetMethod {
+		return nil
+	}
+	return registry.DeclaredScopesForMethod(target.Method.Method, identity)
+}
 
-	service := cmd.Parent().Parent().Name()
-	resource := cmd.Parent().Name()
-	method := cmd.Name()
-
-	spec := registry.LoadFromMeta(service)
-	if spec == nil {
-		return nil
+// commandCatalogPath reconstructs the catalog path [service, resource..., method]
+// from a command's ancestry, excluding the root command. It is the inverse of
+// the service command tree's construction, so any depth (flat or nested)
+// round-trips through apicatalog.Resolve.
+func commandCatalogPath(cmd *cobra.Command) []string {
+	var path []string
+	for c := cmd; c != nil && c.Parent() != nil; c = c.Parent() {
+		path = append([]string{c.Name()}, path...)
 	}
-	resources, _ := spec["resources"].(map[string]interface{})
-	resMap, _ := resources[resource].(map[string]interface{})
-	if resMap == nil {
-		return nil
-	}
-	methods, _ := resMap["methods"].(map[string]interface{})
-	methodMap, _ := methods[method].(map[string]interface{})
-	if methodMap == nil {
-		return nil
-	}
-	return registry.DeclaredScopesForMethod(methodMap, identity)
+	return path
 }
 
 // shortcutSupportsIdentity reports whether a shortcut supports the requested

cmd/root.go

@@ -36,7 +36,7 @@ const rootLong = `lark-cli — Lark/Feishu CLI tool.
 USAGE:
     lark-cli <command> [subcommand] [method] [options]
     lark-cli api <method> <path> [--params <json>] [--data <json>]
-    lark-cli schema <service.resource.method> [--format pretty]
+    lark-cli schema <service.resource.method>
 
 EXAMPLES:
     # View upcoming events

cmd/schema/schema.go

@@ -5,17 +5,17 @@ package schema
 
 import (
 	"context"
-	"fmt"
+	"errors"
 	"io"
-	"sort"
 	"strings"
 
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/apicatalog"
 	"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/internal/schema"
-	"github.com/larksuite/cli/internal/util"
 	"github.com/spf13/cobra"
 )
 
@@ -24,336 +24,10 @@ type SchemaOptions struct {
 	Factory *cmdutil.Factory
 	Ctx     context.Context
 
-	// Positional args
-	Path      string   // first positional, when only one is given
-	ExtraArgs []string // 2nd+ positional args (space-separated form)
-
-	// Flags
-	Format string
-}
-
-func printServices(w io.Writer) {
-	services := registry.ListFromMetaProjects()
-	fmt.Fprintf(w, "%sAvailable services:%s\n\n", output.Bold, output.Reset)
-	for _, s := range services {
-		spec := registry.LoadFromMeta(s)
-		title := registry.GetStrFromMap(spec, "title")
-		if title == "" {
-			title = registry.GetStrFromMap(spec, "description")
-		}
-		fmt.Fprintf(w, "  %s%s%s  %s%s%s\n", output.Cyan, s, output.Reset, output.Dim, title, output.Reset)
-	}
-	fmt.Fprintf(w, "\n%sUsage: lark-cli schema <service>.<resource>.<method>%s\n", output.Dim, output.Reset)
-}
-
-func printResourceList(w io.Writer, spec map[string]interface{}, mode core.StrictMode) {
-	name := registry.GetStrFromMap(spec, "name")
-	version := registry.GetStrFromMap(spec, "version")
-	title := registry.GetStrFromMap(spec, "title")
-	if title == "" {
-		title = registry.GetStrFromMap(spec, "description")
-	}
-	servicePath := registry.GetStrFromMap(spec, "servicePath")
-
-	fmt.Fprintf(w, "%s%s%s (%s) — %s\n\n", output.Bold, name, output.Reset, version, title)
-	fmt.Fprintf(w, "%sBase path: %s%s\n\n", output.Dim, servicePath, output.Reset)
-
-	resources, _ := spec["resources"].(map[string]interface{})
-	for _, resName := range sortedKeys(resources) {
-		resMap, _ := resources[resName].(map[string]interface{})
-		methods, _ := resMap["methods"].(map[string]interface{})
-		methods = filterMethodsByStrictMode(methods, mode)
-		if len(methods) == 0 {
-			continue
-		}
-		fmt.Fprintf(w, "  %s%s%s\n", output.Cyan, resName, output.Reset)
-		for _, methodName := range sortedKeys(methods) {
-			m, _ := methods[methodName].(map[string]interface{})
-			httpMethod := registry.GetStrFromMap(m, "httpMethod")
-			desc := registry.GetStrFromMap(m, "description")
-			danger := ""
-			if d, _ := m["danger"].(bool); d {
-				danger = fmt.Sprintf(" %s[danger]%s", output.Red, output.Reset)
-			}
-			fmt.Fprintf(w, "    %-7s %s%s%s  %s%s%s%s\n", httpMethod, output.Bold, methodName, output.Reset, output.Dim, desc, output.Reset, danger)
-		}
-		fmt.Fprintln(w)
-	}
-	fmt.Fprintf(w, "%sUsage: lark-cli schema %s.<resource>.<method>%s\n", output.Dim, name, output.Reset)
-}
-
-// hasFileFields returns true if any requestBody field has type "file".
-func hasFileFields(method map[string]interface{}) (bool, []string) {
-	names := cmdutil.DetectFileFields(method)
-	return len(names) > 0, names
-}
-
-func printMethodDetail(w io.Writer, spec map[string]interface{}, resName, methodName string, method map[string]interface{}) {
-	servicePath := registry.GetStrFromMap(spec, "servicePath")
-	specName := registry.GetStrFromMap(spec, "name")
-	methodPath := registry.GetStrFromMap(method, "path")
-	fullPath := servicePath + "/" + methodPath
-	httpMethod := registry.GetStrFromMap(method, "httpMethod")
-	desc := registry.GetStrFromMap(method, "description")
-	isFileUpload, fileFieldNames := hasFileFields(method)
-
-	fmt.Fprintf(w, "%s%s.%s.%s%s\n\n", output.Bold, specName, resName, methodName, output.Reset)
-
-	httpColor := output.Yellow
-	if httpMethod == "GET" {
-		httpColor = output.Green
-	} else if httpMethod == "DELETE" {
-		httpColor = output.Red
-	}
-	fmt.Fprintf(w, "  %s%s%s %s\n", httpColor, httpMethod, output.Reset, fullPath)
-	if desc != "" {
-		fmt.Fprintf(w, "  %s\n", desc)
-	}
-	fmt.Fprintln(w)
-
-	// Parameters
-	params, _ := method["parameters"].(map[string]interface{})
-	if len(params) > 0 {
-		fmt.Fprintf(w, "%sParameters:%s\n\n", output.Bold, output.Reset)
-		fmt.Fprintf(w, "  %s--params%s  <json>  %soptional%s\n", output.Cyan, output.Reset, output.Dim, output.Reset)
-		for _, paramName := range sortedParamKeys(params) {
-			p, _ := params[paramName].(map[string]interface{})
-			pType := registry.GetStrFromMap(p, "type")
-			if pType == "" {
-				pType = "string"
-			}
-			location := registry.GetStrFromMap(p, "location")
-			required, _ := p["required"].(bool)
-			reqStr := fmt.Sprintf("%soptional%s", output.Dim, output.Reset)
-			if required {
-				reqStr = fmt.Sprintf("%srequired%s", output.Red, output.Reset)
-			}
-			locColor := output.Dim
-			if location == "path" {
-				locColor = output.Yellow
-			}
-			// Options (enum values)
-			optStr := formatOptions(p)
-			fmt.Fprintf(w, "      - %s%s%s (%s, %s%s%s, %s)%s\n", output.Cyan, paramName, output.Reset, pType, locColor, location, output.Reset, reqStr, optStr)
-			if pdesc := registry.GetStrFromMap(p, "description"); pdesc != "" {
-				pdesc = util.TruncateStrWithEllipsis(pdesc, 100)
-				fmt.Fprintf(w, "        %s%s%s\n", output.Dim, pdesc, output.Reset)
-			}
-			if ex := registry.GetStrFromMap(p, "example"); ex != "" {
-				fmt.Fprintf(w, "        %se.g. %s%s\n", output.Dim, ex, output.Reset)
-			}
-			if rangeStr := formatRange(p); rangeStr != "" {
-				fmt.Fprintf(w, "        %srange: %s%s\n", output.Dim, rangeStr, output.Reset)
-			}
-		}
-		fmt.Fprintln(w)
-	}
-
-	// --data for write methods
-	if httpMethod == "POST" || httpMethod == "PUT" || httpMethod == "PATCH" || httpMethod == "DELETE" {
-		if len(params) == 0 {
-			fmt.Fprintf(w, "%sParameters:%s\n\n", output.Bold, output.Reset)
-		}
-		fileUploadTag := ""
-		if isFileUpload {
-			fileUploadTag = fmt.Sprintf("  %s[file upload]%s", output.Yellow, output.Reset)
-		}
-		fmt.Fprintf(w, "  %s--data%s  <json>  %soptional%s%s\n", output.Cyan, output.Reset, output.Dim, output.Reset, fileUploadTag)
-		requestBody, _ := method["requestBody"].(map[string]interface{})
-		if len(requestBody) > 0 {
-			printNestedFields(w, requestBody, "      ", "")
-		}
-
-		if isFileUpload {
-			if len(fileFieldNames) == 1 {
-				fmt.Fprintf(w, "\n  %s--file%s  <[field=]path>  %sfile upload%s\n", output.Cyan, output.Reset, output.Dim, output.Reset)
-				fmt.Fprintf(w, "      Upload file as multipart/form-data. Default field: %q\n", fileFieldNames[0])
-			} else {
-				fmt.Fprintf(w, "\n  %s--file%s  <field=path>  %sfile upload%s\n", output.Cyan, output.Reset, output.Dim, output.Reset)
-				fmt.Fprintf(w, "      Upload file as multipart/form-data. Fields: %s\n", strings.Join(fileFieldNames, ", "))
-			}
-		}
-		fmt.Fprintln(w)
-	}
-
-	// Response
-	responseBody, _ := method["responseBody"].(map[string]interface{})
-	if len(responseBody) > 0 {
-		fmt.Fprintf(w, "%sResponse:%s\n\n", output.Bold, output.Reset)
-		printNestedFields(w, responseBody, "  ", "")
-		fmt.Fprintln(w)
-	}
-
-	// Identity
-	if tokens, ok := method["accessTokens"].([]interface{}); ok && len(tokens) > 0 {
-		var identities []string
-		for _, t := range tokens {
-			if s, ok := t.(string); ok {
-				switch s {
-				case "user":
-					identities = append(identities, "user")
-				case "tenant":
-					identities = append(identities, "bot")
-				}
-			}
-		}
-		if len(identities) > 0 {
-			fmt.Fprintf(w, "%sIdentity:%s %s\n", output.Bold, output.Reset, strings.Join(identities, ", "))
-		}
-	}
-
-	// Scopes (all)
-	if scopes, ok := method["scopes"].([]interface{}); ok && len(scopes) > 0 {
-		var scopeStrs []string
-		for _, s := range scopes {
-			if str, ok := s.(string); ok {
-				scopeStrs = append(scopeStrs, str)
-			}
-		}
-		fmt.Fprintf(w, "%sScopes:%s   %s\n", output.Bold, output.Reset, strings.Join(scopeStrs, ", "))
-	}
-
-	// CLI example
-	if isFileUpload && len(fileFieldNames) == 1 {
-		fmt.Fprintf(w, "%sCLI:%s      lark-cli %s %s %s --file <path>\n", output.Bold, output.Reset, specName, resName, methodName)
-	} else if isFileUpload {
-		fmt.Fprintf(w, "%sCLI:%s      lark-cli %s %s %s --file <field=path>\n", output.Bold, output.Reset, specName, resName, methodName)
-	} else {
-		fmt.Fprintf(w, "%sCLI:%s      lark-cli %s %s %s\n", output.Bold, output.Reset, specName, resName, methodName)
-	}
-
-	// Docs
-	if docUrl := registry.GetStrFromMap(method, "docUrl"); docUrl != "" {
-		fmt.Fprintf(w, "%sDocs:%s     %s\n", output.Bold, output.Reset, docUrl)
-	}
-}
-
-func printNestedFields(w io.Writer, fields map[string]interface{}, indent, prefix string) {
-	for _, fieldName := range sortedFieldKeys(fields) {
-		f, _ := fields[fieldName].(map[string]interface{})
-		fullName := fieldName
-		if prefix != "" {
-			fullName = prefix + "." + fieldName
-		}
-		fType := registry.GetStrFromMap(f, "type")
-		required, _ := f["required"].(bool)
-		reqStr := fmt.Sprintf("%soptional%s", output.Dim, output.Reset)
-		if required {
-			reqStr = fmt.Sprintf("%srequired%s", output.Red, output.Reset)
-		}
-		optStr := formatOptions(f)
-		fmt.Fprintf(w, "%s- %s%s%s (%s, %s)%s\n", indent, output.Cyan, fullName, output.Reset, fType, reqStr, optStr)
-		desc := registry.GetStrFromMap(f, "description")
-		if desc != "" {
-			desc = util.TruncateStrWithEllipsis(desc, 100)
-			fmt.Fprintf(w, "%s  %s%s%s\n", indent, output.Dim, desc, output.Reset)
-		}
-		if ex := registry.GetStrFromMap(f, "example"); ex != "" {
-			fmt.Fprintf(w, "%s  %se.g. %s%s\n", indent, output.Dim, ex, output.Reset)
-		}
-		if rangeStr := formatRange(f); rangeStr != "" {
-			fmt.Fprintf(w, "%s  %srange: %s%s\n", indent, output.Dim, rangeStr, output.Reset)
-		}
-		if props, ok := f["properties"].(map[string]interface{}); ok && len(props) > 0 {
-			printNestedFields(w, props, indent+"  ", fullName)
-		}
-	}
-}
-
-// formatOptions returns " — val1 | val2 | ..." if field has options, else "".
-func formatOptions(f map[string]interface{}) string {
-	opts, ok := f["options"].([]interface{})
-	if !ok || len(opts) == 0 {
-		return ""
-	}
-	var vals []string
-	for _, o := range opts {
-		if om, ok := o.(map[string]interface{}); ok {
-			if v := registry.GetStrFromMap(om, "value"); v != "" {
-				vals = append(vals, v)
-			}
-		}
-	}
-	if len(vals) == 0 {
-		return ""
-	}
-	return fmt.Sprintf(" %s— %s%s", output.Dim, strings.Join(vals, " | "), output.Reset)
-}
-
-// formatRange returns "min..max" if field has min/max, else "".
-func formatRange(f map[string]interface{}) string {
-	minVal := registry.GetStrFromMap(f, "min")
-	maxVal := registry.GetStrFromMap(f, "max")
-	if minVal == "" && maxVal == "" {
-		return ""
-	}
-	if minVal != "" && maxVal != "" {
-		return minVal + ".." + maxVal
-	}
-	if minVal != "" {
-		return ">=" + minVal
-	}
-	return "<=" + maxVal
-}
-
-// sortedKeys returns map keys in alphabetical order.
-func sortedKeys(m map[string]interface{}) []string {
-	keys := make([]string, 0, len(m))
-	for k := range m {
-		keys = append(keys, k)
-	}
-	sort.Strings(keys)
-	return keys
-}
-
-// sortedParamKeys returns parameter keys sorted: required first, then alphabetical.
-func sortedParamKeys(params map[string]interface{}) []string {
-	keys := make([]string, 0, len(params))
-	for k := range params {
-		keys = append(keys, k)
-	}
-	sort.Slice(keys, func(i, j int) bool {
-		pi, _ := params[keys[i]].(map[string]interface{})
-		pj, _ := params[keys[j]].(map[string]interface{})
-		ri, _ := pi["required"].(bool)
-		rj, _ := pj["required"].(bool)
-		if ri != rj {
-			return ri
-		}
-		return keys[i] < keys[j]
-	})
-	return keys
-}
-
-// sortedFieldKeys returns field keys sorted: required first, then alphabetical.
-func sortedFieldKeys(fields map[string]interface{}) []string {
-	keys := make([]string, 0, len(fields))
-	for k := range fields {
-		keys = append(keys, k)
-	}
-	sort.Slice(keys, func(i, j int) bool {
-		fi, _ := fields[keys[i]].(map[string]interface{})
-		fj, _ := fields[keys[j]].(map[string]interface{})
-		ri, _ := fi["required"].(bool)
-		rj, _ := fj["required"].(bool)
-		if ri != rj {
-			return ri
-		}
-		return keys[i] < keys[j]
-	})
-	return keys
-}
-
-func findResourceByPath(resources map[string]interface{}, parts []string) (map[string]interface{}, string, []string) {
-	for i := len(parts); i >= 1; i-- {
-		candidateName := strings.Join(parts[:i], ".")
-		if res, ok := resources[candidateName]; ok {
-			if resMap, ok := res.(map[string]interface{}); ok {
-				return resMap, candidateName, parts[i:]
-			}
-		}
-	}
-	return nil, "", nil
+	// Args are the positional path segments, in either the dotted single-arg
+	// form ("im.messages.reply") or the space-separated form ("im messages
+	// reply"); apicatalog.ParsePath normalizes both.
+	Args []string
 }
 
 // NewCmdSchema creates the schema command. If runF is non-nil it is called instead of schemaRun (test hook).
@@ -365,12 +39,7 @@ func NewCmdSchema(f *cmdutil.Factory, runF func(*SchemaOptions) error) *cobra.Co
 		Short: "View API method parameters, types, and scopes",
 		Args:  cobra.MaximumNArgs(8),
 		RunE: func(cmd *cobra.Command, args []string) error {
-			if len(args) > 0 {
-				opts.Path = args[0]
-			}
-			if len(args) > 1 {
-				opts.ExtraArgs = args[1:]
-			}
+			opts.Args = append([]string(nil), args...)
 			opts.Ctx = cmd.Context()
 			if runF != nil {
 				return runF(opts)
@@ -380,433 +49,89 @@ func NewCmdSchema(f *cmdutil.Factory, runF func(*SchemaOptions) error) *cobra.Co
 	}
 	cmdutil.DisableAuthCheck(cmd)
 
+	// Tolerated for agent compatibility; ignored — schema only emits the JSON
+	// envelope, and its output is identity-independent (strict-mode filtering
+	// comes from ResolveStrictMode, never from --as).
+	cmd.Flags().String("format", "json", "")
+	cmd.Flags().Bool("json", true, "")
+	cmd.Flags().String("as", "", "")
+	_ = cmd.Flags().MarkHidden("format")
+	_ = cmd.Flags().MarkHidden("json")
+	_ = cmd.Flags().MarkHidden("as")
+
 	cmd.ValidArgsFunction = completeSchemaPath(f)
-	cmd.Flags().StringVar(&opts.Format, "format", "json", "output format: json (default) | pretty")
-	cmdutil.RegisterFlagCompletion(cmd, "format", func(_ *cobra.Command, _ []string, _ string) ([]string, cobra.ShellCompDirective) {
-		return []string{"json", "pretty"}, cobra.ShellCompDirectiveNoFileComp
-	})
 	cmdutil.SetRisk(cmd, cmdutil.RiskRead)
 
 	return cmd
 }
 
-// completeSchemaPath provides tab-completion for the schema path argument.
-// It handles both legacy dotted resource names (e.g. app.table.fields) and the
-// newer space-separated form (e.g. `schema im messages reply`).
+// completeSchemaPath is a thin adapter over the embedded catalog's Complete.
+// It uses the embedded source so completion candidates match what `schema`
+// execution can resolve (both overlay-free).
 func completeSchemaPath(f *cmdutil.Factory) func(*cobra.Command, []string, string) ([]string, cobra.ShellCompDirective) {
 	return func(cmd *cobra.Command, args []string, toComplete string) ([]string, cobra.ShellCompDirective) {
 		mode := f.ResolveStrictMode(cmd.Context())
-
-		// Case 1: legacy "single dotted arg" path — no previous args yet
-		if len(args) == 0 {
-			parts := strings.Split(toComplete, ".")
-			if len(parts) <= 1 {
-				var completions []string
-				for _, s := range registry.ListFromMetaProjects() {
-					if strings.HasPrefix(s, toComplete) {
-						completions = append(completions, s+".")
-					}
-				}
-				return completions, cobra.ShellCompDirectiveNoFileComp | cobra.ShellCompDirectiveNoSpace
-			}
-			serviceName := parts[0]
-			spec := registry.LoadFromMeta(serviceName)
-			if spec == nil {
-				return nil, cobra.ShellCompDirectiveNoFileComp
-			}
-			spec = filterSpecByStrictMode(spec, mode)
-			resources, _ := spec["resources"].(map[string]interface{})
-			if resources == nil {
-				return nil, cobra.ShellCompDirectiveNoFileComp
-			}
-			afterService := strings.Join(parts[1:], ".")
-			completions := completeSchemaPathForSpec(serviceName, resources, afterService)
-			allTrailingDot := len(completions) > 0
-			for _, c := range completions {
-				if !strings.HasSuffix(c, ".") {
-					allTrailingDot = false
-					break
-				}
-			}
-			directive := cobra.ShellCompDirectiveNoFileComp
-			if allTrailingDot {
-				directive |= cobra.ShellCompDirectiveNoSpace
-			}
-			return completions, directive
+		completions, noSpace := registry.EmbeddedCatalog().Complete(args, toComplete, registry.FilterForStrictMode(mode))
+		directive := cobra.ShellCompDirectiveNoFileComp
+		if noSpace {
+			directive |= cobra.ShellCompDirectiveNoSpace
 		}
-
-		// Case 2: space-form, args already has segments
-		// Walk down service -> resource(s) -> method based on existing args
-		serviceName := args[0]
-		spec := registry.LoadFromMeta(serviceName)
-		if spec == nil {
-			return nil, cobra.ShellCompDirectiveNoFileComp
-		}
-		spec = filterSpecByStrictMode(spec, mode)
-		resources, _ := spec["resources"].(map[string]interface{})
-		if resources == nil {
-			return nil, cobra.ShellCompDirectiveNoFileComp
-		}
-
-		// args[1:] are resource path segments (possibly partial); current
-		// toComplete is the next segment under cursor.
-		consumed := args[1:]
-		resource, _, remaining := findResourceByPath(resources, consumed)
-		if resource == nil {
-			// Suggest top-level resource names that match toComplete
-			var completions []string
-			for resName := range resources {
-				if strings.HasPrefix(resName, toComplete) {
-					completions = append(completions, resName)
-				}
-			}
-			sort.Strings(completions)
-			return completions, cobra.ShellCompDirectiveNoFileComp
-		}
-		if len(remaining) > 0 {
-			// Already typed past the resource — suggest methods
-			methods, _ := resource["methods"].(map[string]interface{})
-			methods = filterMethodsByStrictMode(methods, mode)
-			var completions []string
-			for mName := range methods {
-				if strings.HasPrefix(mName, toComplete) {
-					completions = append(completions, mName)
-				}
-			}
-			sort.Strings(completions)
-			return completions, cobra.ShellCompDirectiveNoFileComp
-		}
-		// Resource matched exactly, suggest methods
-		methods, _ := resource["methods"].(map[string]interface{})
-		methods = filterMethodsByStrictMode(methods, mode)
-		var completions []string
-		for mName := range methods {
-			if strings.HasPrefix(mName, toComplete) {
-				completions = append(completions, mName)
-			}
-		}
-		sort.Strings(completions)
-		return completions, cobra.ShellCompDirectiveNoFileComp
+		return completions, directive
 	}
 }
 
-func completeSchemaPathForSpec(serviceName string, resources map[string]interface{}, afterService string) []string {
-	var completions []string
-
-	for resName, resVal := range resources {
-		if strings.HasPrefix(resName, afterService) {
-			completions = append(completions, serviceName+"."+resName+".")
-			continue
-		}
-		if !strings.HasPrefix(afterService, resName+".") {
-			continue
-		}
-		methodPrefix := afterService[len(resName)+1:]
-		resMap, _ := resVal.(map[string]interface{})
-		if resMap == nil {
-			continue
-		}
-		methods, _ := resMap["methods"].(map[string]interface{})
-		for methodName := range methods {
-			if strings.HasPrefix(methodName, methodPrefix) {
-				completions = append(completions, serviceName+"."+resName+"."+methodName)
-			}
-		}
-	}
-
-	sort.Strings(completions)
-	return completions
-}
-
 func schemaRun(opts *SchemaOptions) error {
 	out := opts.Factory.IOStreams.Out
 	mode := opts.Factory.ResolveStrictMode(opts.Ctx)
-
-	// args may have arrived as a single string (legacy single-arg path) or
-	// split into multiple — normalize to a single args slice.
-	var rawArgs []string
-	if opts.Path != "" {
-		rawArgs = []string{opts.Path}
-	}
-	if len(opts.ExtraArgs) > 0 {
-		if opts.Path != "" {
-			rawArgs = append([]string{opts.Path}, opts.ExtraArgs...)
-		} else {
-			rawArgs = append([]string(nil), opts.ExtraArgs...)
-		}
-	}
-	parts := schema.ParsePath(rawArgs)
-
-	if opts.Format == "pretty" {
-		return runPrettyMode(out, parts, mode)
-	}
-	return runJSONMode(out, parts, mode)
+	return runSchema(out, apicatalog.ParsePath(opts.Args), mode)
 }
 
-// runJSONMode dispatches list/single envelope output based on parts.
-// JSON mode uses embedded data only (bypasses remote overlay) so envelope
-// output is deterministic across machines.
-func runJSONMode(out io.Writer, parts []string, mode core.StrictMode) error {
-	filter := strictModeFilter(mode)
-
-	switch len(parts) {
-	case 0:
-		envs := schema.AssembleAll(filter)
-		output.PrintJson(out, envs)
-		return nil
-	case 1:
-		spec := registry.EmbeddedSpec(parts[0])
-		if spec == nil {
-			return errUnknownEmbeddedService(parts[0])
+// runSchema resolves the path through the embedded catalog and renders the
+// matching envelope(s). The catalog owns navigation (Resolve + MethodRefs) and
+// schema owns rendering (Envelope/Envelopes); this adapter only chooses the
+// output shape — a single resolved method renders as one envelope object,
+// anything broader as an array — and maps resolve failures to hints.
+func runSchema(out io.Writer, parts []string, mode core.StrictMode) error {
+	catalog := registry.EmbeddedCatalog()
+	target, err := catalog.Resolve(parts)
+	if err != nil {
+		return resolveError(err)
+	}
+	refs := catalog.MethodRefs(target, registry.FilterForStrictMode(mode))
+	if target.Kind == apicatalog.TargetMethod {
+		if len(refs) == 0 {
+			return errs.NewValidationError(errs.SubtypeInvalidArgument,
+				"Method %s not available in current identity mode", target.Method.SchemaPath()).
+				WithHint("strict mode hides methods the active account identity cannot call; it is shown for an identity (user or bot) that has the required access token")
 		}
-		envs := schema.AssembleService(parts[0], spec, filter)
-		output.PrintJson(out, envs)
-		return nil
-	default:
-		return runJSONForPath(out, parts, filter)
-	}
-}
-
-// runJSONForPath handles len(parts) >= 2: try resource match first, fallback
-// to single-method match. Uses embedded data only.
-func runJSONForPath(out io.Writer, parts []string, filter schema.MethodFilter) error {
-	serviceName := parts[0]
-	spec := registry.EmbeddedSpec(serviceName)
-	if spec == nil {
-		return errUnknownEmbeddedService(serviceName)
-	}
-	resources, _ := spec["resources"].(map[string]interface{})
-	resource, resName, remaining := findResourceByPath(resources, parts[1:])
-	if resource == nil {
-		var names []string
-		for k := range resources {
-			names = append(names, k)
-		}
-		sort.Strings(names)
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Unknown resource: %s.%s", serviceName, strings.Join(parts[1:], ".")),
-			fmt.Sprintf("Available: %s", strings.Join(names, ", ")))
-	}
-	if len(remaining) == 0 {
-		// Resource-scoped envelope array
-		envs := assembleResource(serviceName, resName, resource, filter)
-		output.PrintJson(out, envs)
+		output.PrintJson(out, schema.EnvelopeOf(refs[0]))
 		return nil
 	}
-	methodName := remaining[0]
-	methods, _ := resource["methods"].(map[string]interface{})
-	method, ok := methods[methodName].(map[string]interface{})
-	if !ok {
-		var names []string
-		for k := range methods {
-			names = append(names, k)
-		}
-		sort.Strings(names)
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Unknown method: %s.%s.%s", serviceName, resName, methodName),
-			fmt.Sprintf("Available: %s", strings.Join(names, ", ")))
-	}
-	if len(remaining) > 1 {
-		// Method exists but caller appended extra segments — reject so they
-		// don't silently get this method's schema when they typo'd the path.
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Unknown path: %s.%s.%s",
-				serviceName, resName, strings.Join(remaining, ".")),
-			fmt.Sprintf("Method %q exists but the trailing segments %q do not resolve",
-				methodName, strings.Join(remaining[1:], ".")))
-	}
-	if filter != nil && !filter(method) {
-		// Method exists in spec but filtered out by strict mode
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Method %s.%s.%s not available in current identity mode", serviceName, resName, methodName),
-			"Use --as user / --as bot to switch")
-	}
-	env := schema.AssembleEnvelope(serviceName, []string{resName}, methodName, method)
-	output.PrintJson(out, env)
+	output.PrintJson(out, schema.Envelopes(refs))
 	return nil
 }
 
-func assembleResource(serviceName, resName string, resource map[string]interface{}, filter schema.MethodFilter) []schema.Envelope {
-	methods, _ := resource["methods"].(map[string]interface{})
-	resourcePath := []string{resName}
-	var envs []schema.Envelope
-	for methodName, raw := range methods {
-		method, ok := raw.(map[string]interface{})
-		if !ok {
-			continue
-		}
-		if filter != nil && !filter(method) {
-			continue
-		}
-		envs = append(envs, schema.AssembleEnvelope(serviceName, resourcePath, methodName, method))
+// resolveError maps a catalog *ResolveError to a typed *errs.ValidationError
+// (CategoryValidation drives the exit code; Hint promotes to the envelope),
+// preserving the historical message + hint text.
+func resolveError(err error) error {
+	var re *apicatalog.ResolveError
+	if !errors.As(err, &re) {
+		return err
 	}
-	sort.Slice(envs, func(i, j int) bool { return envs[i].Name < envs[j].Name })
-	return envs
-}
-
-// runPrettyMode preserves the existing legacy pretty rendering verbatim.
-// All printServices/printResourceList/printMethodDetail calls stay unchanged.
-func runPrettyMode(out io.Writer, parts []string, mode core.StrictMode) error {
-	if len(parts) == 0 {
-		printServices(out)
-		return nil
-	}
-	serviceName := parts[0]
-	spec := registry.LoadFromMeta(serviceName)
-	if spec == nil {
-		return errUnknownService(serviceName)
-	}
-	if len(parts) == 1 {
-		printResourceList(out, spec, mode)
-		return nil
-	}
-	resources, _ := spec["resources"].(map[string]interface{})
-	resource, resName, remaining := findResourceByPath(resources, parts[1:])
-	if resource == nil {
-		var names []string
-		for k := range resources {
-			names = append(names, k)
-		}
-		sort.Strings(names)
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Unknown resource: %s.%s", serviceName, strings.Join(parts[1:], ".")),
-			fmt.Sprintf("Available: %s", strings.Join(names, ", ")))
-	}
-	if len(remaining) == 0 {
-		fmt.Fprintf(out, "%s%s.%s%s\n\n", output.Bold, serviceName, resName, output.Reset)
-		methods, _ := resource["methods"].(map[string]interface{})
-		methods = filterMethodsByStrictMode(methods, mode)
-		for _, mName := range sortedKeys(methods) {
-			m, _ := methods[mName].(map[string]interface{})
-			httpMethod := registry.GetStrFromMap(m, "httpMethod")
-			desc := registry.GetStrFromMap(m, "description")
-			fmt.Fprintf(out, "  %-7s %s%s%s  %s%s%s\n", httpMethod, output.Bold, mName, output.Reset, output.Dim, desc, output.Reset)
-		}
-		fmt.Fprintf(out, "\n%sUsage: lark-cli schema %s.%s.<method>%s\n", output.Dim, serviceName, resName, output.Reset)
-		return nil
-	}
-	methodName := remaining[0]
-	methods, _ := resource["methods"].(map[string]interface{})
-	methods = filterMethodsByStrictMode(methods, mode)
-	method, ok := methods[methodName].(map[string]interface{})
-	if !ok {
-		var names []string
-		for k := range methods {
-			names = append(names, k)
-		}
-		sort.Strings(names)
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Unknown method: %s.%s.%s", serviceName, resName, methodName),
-			fmt.Sprintf("Available: %s", strings.Join(names, ", ")))
-	}
-	if len(remaining) > 1 {
-		return output.ErrWithHint(output.ExitValidation, "validation",
-			fmt.Sprintf("Unknown path: %s.%s.%s",
-				serviceName, resName, strings.Join(remaining, ".")),
-			fmt.Sprintf("Method %q exists but the trailing segments %q do not resolve",
-				methodName, strings.Join(remaining[1:], ".")))
-	}
-	printMethodDetail(out, spec, resName, methodName, method)
-	return nil
-}
-
-// strictModeFilter adapts core.StrictMode into a schema.MethodFilter, or returns
-// nil if strict mode is not active.
-func strictModeFilter(mode core.StrictMode) schema.MethodFilter {
-	if !mode.IsActive() {
-		return nil
-	}
-	token := registry.IdentityToAccessToken(string(mode.ForcedIdentity()))
-	return func(method map[string]interface{}) bool {
-		tokens, _ := method["accessTokens"].([]interface{})
-		if tokens == nil {
-			return true // permissive when meta_data lacks accessTokens
-		}
-		for _, t := range tokens {
-			if s, _ := t.(string); s == token {
-				return true
-			}
-		}
-		return false
-	}
-}
-
-func errUnknownService(name string) error {
-	return output.ErrWithHint(output.ExitValidation, "validation",
-		fmt.Sprintf("Unknown service: %s", name),
-		fmt.Sprintf("Available: %s", strings.Join(registry.ListFromMetaProjects(), ", ")))
-}
-
-// errUnknownEmbeddedService is the JSON-mode variant: it lists only embedded
-// services (no overlay) because JSON mode itself bypasses overlay; suggesting
-// overlay-only services would mislead callers when those services subsequently
-// fail to resolve in envelope output.
-func errUnknownEmbeddedService(name string) error {
-	return output.ErrWithHint(output.ExitValidation, "validation",
-		fmt.Sprintf("Unknown service: %s", name),
-		fmt.Sprintf("Available: %s", strings.Join(registry.EmbeddedServiceNames(), ", ")))
-}
-
-// filterSpecByStrictMode returns a shallow copy of spec with each resource's methods
-// filtered by strict mode. Returns the original spec when strict mode is off.
-func filterSpecByStrictMode(spec map[string]interface{}, mode core.StrictMode) map[string]interface{} {
-	if !mode.IsActive() {
-		return spec
-	}
-	result := make(map[string]interface{}, len(spec))
-	for k, v := range spec {
-		result[k] = v
-	}
-	resources, _ := spec["resources"].(map[string]interface{})
-	if resources == nil {
-		return result
-	}
-	filteredRes := make(map[string]interface{}, len(resources))
-	for resName, resVal := range resources {
-		resMap, ok := resVal.(map[string]interface{})
-		if !ok {
-			continue
-		}
-		methods, _ := resMap["methods"].(map[string]interface{})
-		filtered := filterMethodsByStrictMode(methods, mode)
-		if len(filtered) == 0 {
-			continue
-		}
-		resCopy := make(map[string]interface{}, len(resMap))
-		for k, v := range resMap {
-			resCopy[k] = v
-		}
-		resCopy["methods"] = filtered
-		filteredRes[resName] = resCopy
-	}
-	result["resources"] = filteredRes
-	return result
-}
-
-// filterMethodsByStrictMode removes methods incompatible with the active strict mode.
-// Returns the original map unmodified when strict mode is off.
-func filterMethodsByStrictMode(methods map[string]interface{}, mode core.StrictMode) map[string]interface{} {
-	if !mode.IsActive() || methods == nil {
-		return methods
-	}
-	token := registry.IdentityToAccessToken(string(mode.ForcedIdentity()))
-	filtered := make(map[string]interface{}, len(methods))
-	for name, val := range methods {
-		m, ok := val.(map[string]interface{})
-		if !ok {
-			continue
-		}
-		tokens, _ := m["accessTokens"].([]interface{})
-		if tokens == nil {
-			filtered[name] = val
-			continue
-		}
-		for _, t := range tokens {
-			if ts, ok := t.(string); ok && ts == token {
-				filtered[name] = val
-				break
-			}
-		}
-	}
-	return filtered
+	switch re.Kind {
+	case apicatalog.ErrService:
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "Unknown service: %s", re.Subject).
+			WithHint("Available: %s", strings.Join(re.Candidates, ", "))
+	case apicatalog.ErrResource:
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "Unknown resource: %s", re.Subject).
+			WithHint("Available: %s", strings.Join(re.Candidates, ", "))
+	case apicatalog.ErrMethod:
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "Unknown method: %s", re.Subject).
+			WithHint("Available: %s", strings.Join(re.Candidates, ", "))
+	case apicatalog.ErrPath:
+		return errs.NewValidationError(errs.SubtypeInvalidArgument, "Unknown path: %s", re.Subject).
+			WithHint("Method %q exists but the trailing segments %q do not resolve", re.Method, re.Trailing)
+	}
+	return err
 }

cmd/schema/schema_test.go

@@ -4,7 +4,6 @@
 package schema
 
 import (
-	"bytes"
 	"encoding/json"
 	"strings"
 	"testing"
@@ -21,29 +20,46 @@ func TestSchemaCmd_FlagParsing(t *testing.T) {
 		gotOpts = opts
 		return nil
 	})
-	cmd.SetArgs([]string{"calendar.events.list", "--format", "pretty"})
+	cmd.SetArgs([]string{"calendar.events.list"})
 	err := cmd.Execute()
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
-	if gotOpts.Path != "calendar.events.list" {
-		t.Errorf("expected path calendar.events.list, got %s", gotOpts.Path)
-	}
-	if gotOpts.Format != "pretty" {
-		t.Errorf("expected Format=pretty, got %s", gotOpts.Format)
+	if len(gotOpts.Args) != 1 || gotOpts.Args[0] != "calendar.events.list" {
+		t.Errorf("expected args [calendar.events.list], got %v", gotOpts.Args)
 	}
 }
 
-func TestSchemaCmd_NoArgs_Pretty(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, nil)
-
-	cmd := NewCmdSchema(f, nil)
-	cmd.SetArgs([]string{"--format", "pretty"})
-	if err := cmd.Execute(); err != nil {
-		t.Fatalf("unexpected error: %v", err)
+func TestSchemaCmd_OutputFlagsAcceptedForCompat(t *testing.T) {
+	// Agents are habituated to --format/--json/--as from api/service commands.
+	// schema must accept them without erroring and always emit the JSON envelope —
+	// its output is structured JSON and identity-independent, so the values have
+	// no effect.
+	argSets := [][]string{
+		{"--format", "json"},
+		{"--format", "pretty"},
+		{"--format", "table"}, // no table rendering for a nested schema -> JSON
+		{"--format", "csv"},
+		{"--json"},
+		{"--json", "--format", "ndjson"},
+		{"--as", "user"},
+		{"--as", "bot"},
+		{"--as", "user", "--json"},
 	}
-	if !strings.Contains(stdout.String(), "Available services") {
-		t.Error("expected service list in pretty mode")
+	for _, extra := range argSets {
+		f, stdout, _, _ := cmdutil.TestFactory(t, nil)
+		cmd := NewCmdSchema(f, nil)
+		cmd.SetArgs(append([]string{"im.images.create"}, extra...))
+		if err := cmd.Execute(); err != nil {
+			t.Fatalf("args %v should be accepted, got error: %v", extra, err)
+		}
+		var env map[string]interface{}
+		if err := json.Unmarshal(stdout.Bytes(), &env); err != nil {
+			t.Fatalf("args %v: output is not a JSON envelope: %v\n%s", extra, err, stdout.String())
+		}
+		if env["name"] != "im images create" {
+			t.Errorf("args %v: expected the im images create envelope, got name=%v", extra, env["name"])
+		}
 	}
 }
 
@@ -51,7 +67,7 @@ func TestSchemaCmd_NoArgs_JSON_IsArray(t *testing.T) {
 	f, stdout, _, _ := cmdutil.TestFactory(t, nil)
 
 	cmd := NewCmdSchema(f, nil)
-	cmd.SetArgs([]string{}) // default --format json
+	cmd.SetArgs([]string{})
 	if err := cmd.Execute(); err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
@@ -76,7 +92,7 @@ func TestSchemaCmd_JSONIsEnvelope(t *testing.T) {
 	f, stdout, _, _ := cmdutil.TestFactory(t, nil)
 
 	cmd := NewCmdSchema(f, nil)
-	cmd.SetArgs([]string{"im.images.create", "--format", "json"})
+	cmd.SetArgs([]string{"im.images.create"})
 	if err := cmd.Execute(); err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
@@ -179,23 +195,6 @@ func TestSchemaCmd_NoYesForReadRisk(t *testing.T) {
 	}
 }
 
-func TestSchemaCmd_PrettyUnchanged_KeyTextPresent(t *testing.T) {
-	f, stdout, _, _ := cmdutil.TestFactory(t, nil)
-
-	cmd := NewCmdSchema(f, nil)
-	cmd.SetArgs([]string{"im.images.create", "--format", "pretty"})
-	if err := cmd.Execute(); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	out := stdout.String()
-	// Existing pretty rendering surfaces these markers — they must still appear
-	for _, want := range []string{"Parameters:", "Response:", "Identity:", "Scopes:", "CLI:"} {
-		if !strings.Contains(out, want) {
-			t.Errorf("pretty output missing marker %q", want)
-		}
-	}
-}
-
 func TestSchemaCmd_UnknownService(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, &core.CliConfig{
 		AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
@@ -212,168 +211,6 @@ func TestSchemaCmd_UnknownService(t *testing.T) {
 	}
 }
 
-func TestPrintMethodDetail_FileUpload(t *testing.T) {
-	spec := map[string]interface{}{
-		"name":        "im",
-		"servicePath": "/open-apis/im/v1",
-	}
-	method := map[string]interface{}{
-		"path":        "images",
-		"httpMethod":  "POST",
-		"description": "Upload an image",
-		"requestBody": map[string]interface{}{
-			"image_type": map[string]interface{}{
-				"type":     "string",
-				"required": true,
-			},
-			"image": map[string]interface{}{
-				"type":     "file",
-				"required": true,
-			},
-		},
-		"accessTokens": []interface{}{"user", "tenant"},
-	}
-
-	var buf bytes.Buffer
-	printMethodDetail(&buf, spec, "images", "create", method)
-	out := buf.String()
-
-	if !strings.Contains(out, "file upload") {
-		t.Errorf("expected 'file upload' marker in output, got:\n%s", out)
-	}
-	if !strings.Contains(out, "--file") {
-		t.Errorf("expected '--file' in output, got:\n%s", out)
-	}
-	if !strings.Contains(out, `"image"`) {
-		t.Errorf("expected default field name 'image' in output, got:\n%s", out)
-	}
-	if !strings.Contains(out, "--file <path>") {
-		t.Errorf("expected CLI example with --file <path>, got:\n%s", out)
-	}
-}
-
-func TestPrintMethodDetail_NoFileUpload(t *testing.T) {
-	spec := map[string]interface{}{
-		"name":        "calendar",
-		"servicePath": "/open-apis/calendar/v4",
-	}
-	method := map[string]interface{}{
-		"path":        "events",
-		"httpMethod":  "POST",
-		"description": "Create an event",
-		"requestBody": map[string]interface{}{
-			"summary": map[string]interface{}{
-				"type":     "string",
-				"required": true,
-			},
-		},
-	}
-
-	var buf bytes.Buffer
-	printMethodDetail(&buf, spec, "events", "create", method)
-	out := buf.String()
-
-	if strings.Contains(out, "file upload") {
-		t.Errorf("did not expect 'file upload' marker for non-file method, got:\n%s", out)
-	}
-	if strings.Contains(out, "--file") {
-		t.Errorf("did not expect '--file' for non-file method, got:\n%s", out)
-	}
-}
-
-func TestHasFileFields(t *testing.T) {
-	tests := []struct {
-		name       string
-		method     map[string]interface{}
-		wantBool   bool
-		wantFields []string
-	}{
-		{
-			name: "has file field",
-			method: map[string]interface{}{
-				"requestBody": map[string]interface{}{
-					"image": map[string]interface{}{"type": "file"},
-					"name":  map[string]interface{}{"type": "string"},
-				},
-			},
-			wantBool:   true,
-			wantFields: []string{"image"},
-		},
-		{
-			name: "no file field",
-			method: map[string]interface{}{
-				"requestBody": map[string]interface{}{
-					"name": map[string]interface{}{"type": "string"},
-				},
-			},
-			wantBool:   false,
-			wantFields: nil,
-		},
-		{
-			name:       "no requestBody",
-			method:     map[string]interface{}{},
-			wantBool:   false,
-			wantFields: nil,
-		},
-	}
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			got, names := hasFileFields(tt.method)
-			if got != tt.wantBool {
-				t.Errorf("hasFileFields() = %v, want %v", got, tt.wantBool)
-			}
-			if tt.wantFields == nil && names != nil {
-				t.Errorf("expected nil names, got %v", names)
-			}
-			if tt.wantFields != nil && len(names) != len(tt.wantFields) {
-				t.Errorf("expected %d field names, got %d", len(tt.wantFields), len(names))
-			}
-		})
-	}
-}
-
-func TestCompleteSchemaPathForSpec(t *testing.T) {
-	resources := map[string]interface{}{
-		"records": map[string]interface{}{
-			"methods": map[string]interface{}{
-				"create": map[string]interface{}{},
-				"list":   map[string]interface{}{},
-			},
-		},
-		"record_permissions": map[string]interface{}{
-			"methods": map[string]interface{}{
-				"get": map[string]interface{}{},
-			},
-		},
-	}
-
-	got := completeSchemaPathForSpec("base", resources, "records.cr")
-	if len(got) != 1 || got[0] != "base.records.create" {
-		t.Fatalf("completions = %v, want [base.records.create]", got)
-	}
-
-	got = completeSchemaPathForSpec("base", resources, "record")
-	if len(got) != 2 || got[0] != "base.record_permissions." || got[1] != "base.records." {
-		t.Fatalf("resource completions = %v", got)
-	}
-}
-
-func TestFilterSpecByStrictMode_RemovesIncompatibleMethodsFromCompletionSource(t *testing.T) {
-	spec := map[string]interface{}{
-		"resources": map[string]interface{}{
-			"records": map[string]interface{}{
-				"methods": map[string]interface{}{
-					"list":   map[string]interface{}{"accessTokens": []interface{}{"tenant"}},
-					"create": map[string]interface{}{"accessTokens": []interface{}{"user"}},
-				},
-			},
-		},
-	}
-
-	filtered := filterSpecByStrictMode(spec, core.StrictModeBot)
-	resources, _ := filtered["resources"].(map[string]interface{})
-	got := completeSchemaPathForSpec("base", resources, "records.")
-	if len(got) != 1 || got[0] != "base.records.list" {
-		t.Fatalf("filtered completions = %v, want [base.records.list]", got)
-	}
-}
+// Completion candidate generation (dotted + space forms, strict-mode filtering,
+// dotted-resource handling) now lives in internal/apicatalog and is covered by
+// apicatalog's TestComplete. cmd/schema only adapts catalog.Complete to cobra.

cmd/service/affordance.go

@@ -0,0 +1,80 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/larksuite/cli/internal/meta"
+)
+
+// methodLong composes a method command's long help in one place: the
+// description, the affordance guidance block (when the method has one), the
+// pointer to the full schema, and the params-only addendum (params whose flag
+// name is taken — paramFlagBinder.paramsOnlyHelp, "" when none). Affordance
+// sits near the top so an agent sees when-to-use and few-shot examples before
+// the flag list.
+func methodLong(description, affordance, schemaPath, paramsOnly string) string {
+	var b strings.Builder
+	b.WriteString(description)
+	if affordance != "" {
+		b.WriteString("\n\n")
+		b.WriteString(affordance)
+	}
+	fmt.Fprintf(&b, "\n\nView parameter definitions before calling:\n  lark-cli schema %s", schemaPath)
+	b.WriteString(paramsOnly)
+	return b.String()
+}
+
+// renderAffordance renders a method's affordance as a help block — when to use,
+// prerequisites, and (most importantly for agents) few-shot Examples — or "" when
+// the method carries no affordance. It reads the single typed model
+// (meta.Method.ParsedAffordance) so the help and the envelope agree on shape.
+func renderAffordance(m meta.Method) string {
+	a, ok := m.ParsedAffordance()
+	if !ok {
+		return ""
+	}
+
+	var b strings.Builder
+	bullets := func(title string, items []string) {
+		var nonEmpty []string
+		for _, it := range items {
+			if strings.TrimSpace(it) != "" {
+				nonEmpty = append(nonEmpty, it)
+			}
+		}
+		if len(nonEmpty) == 0 {
+			return
+		}
+		fmt.Fprintf(&b, "%s:\n", title)
+		for _, it := range nonEmpty {
+			fmt.Fprintf(&b, "  • %s\n", it)
+		}
+	}
+
+	bullets("When to use", a.UseWhen)
+	bullets("Avoid when", a.DoNotUseWhen)
+	bullets("Prerequisites", a.Prerequisites)
+	if len(a.Examples) > 0 {
+		var lines []string
+		for _, ex := range a.Examples {
+			if ex.Command == "" {
+				continue
+			}
+			if ex.Description != "" {
+				lines = append(lines, fmt.Sprintf("  • %s\n      %s", ex.Description, ex.Command))
+			} else {
+				lines = append(lines, fmt.Sprintf("  • %s", ex.Command))
+			}
+		}
+		if len(lines) > 0 {
+			fmt.Fprintf(&b, "Examples:\n%s\n", strings.Join(lines, "\n"))
+		}
+	}
+	bullets("Related", a.Related)
+
+	return strings.TrimRight(b.String(), "\n")
+}

cmd/service/affordance_test.go

@@ -0,0 +1,72 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"encoding/json"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/meta"
+)
+
+func TestRenderAffordance(t *testing.T) {
+	raw := json.RawMessage(`{
+		"use_when": ["发送文本消息"],
+		"do_not_use_when": ["群已解散"],
+		"prerequisites": ["已获取 chat_id"],
+		"examples": [
+			{"description":"发一条文本","command":"lark-cli im messages create --params '{...}'"},
+			{"command":"lark-cli im messages list"},
+			{"description":"no command, skipped","command":""}
+		],
+		"related": ["im.messages.list"]
+	}`)
+	out := renderAffordance(meta.Method{Affordance: raw})
+	for _, want := range []string{
+		"When to use:", "发送文本消息",
+		"Avoid when:", "群已解散",
+		"Prerequisites:", "已获取 chat_id",
+		"Examples:", "发一条文本", "lark-cli im messages create --params '{...}'",
+		"lark-cli im messages list", // example with no description -> bare command line
+		"Related:", "im.messages.list",
+	} {
+		if !strings.Contains(out, want) {
+			t.Errorf("renderAffordance missing %q in:\n%s", want, out)
+		}
+	}
+	if strings.Contains(out, "no command, skipped") {
+		t.Errorf("example with empty command should be skipped:\n%s", out)
+	}
+
+	// Absent or empty affordance renders nothing (so methods without an overlay
+	// add nothing to their help).
+	if renderAffordance(meta.Method{}) != "" || renderAffordance(meta.Method{Affordance: json.RawMessage(`{}`)}) != "" {
+		t.Error("empty affordance should render nothing")
+	}
+}
+
+func TestServiceMethod_AffordanceInLong(t *testing.T) {
+	withAff := map[string]interface{}{
+		"path": "messages", "httpMethod": "POST", "description": "发送消息",
+		"affordance": map[string]interface{}{
+			"examples": []interface{}{
+				map[string]interface{}{"description": "发文本", "command": "lark-cli im messages create ..."},
+			},
+		},
+	}
+	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(withAff), "create", "messages", nil)
+	if !strings.Contains(cmd.Long, "Examples:") || !strings.Contains(cmd.Long, "lark-cli im messages create ...") {
+		t.Errorf("affordance examples not in command Long:\n%s", cmd.Long)
+	}
+
+	// A method with no affordance adds no guidance block.
+	plain := map[string]interface{}{"path": "x", "httpMethod": "GET", "description": "d"}
+	cmd2 := NewCmdServiceMethod(f, imSpec(), meta.FromMap(plain), "list", "x", nil)
+	if strings.Contains(cmd2.Long, "Examples:") {
+		t.Errorf("no-affordance method should have no Examples in Long:\n%s", cmd2.Long)
+	}
+}

cmd/service/flaggroups.go

@@ -0,0 +1,211 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
+)
+
+// Flag annotations the grouped service-method help renderer reads.
+const (
+	flagGroupAnnotation = "lark_flag_group" // display group key
+	flagSubAnnotation   = "lark_flag_sub"   // "required" | "optional" within API Parameters
+	flagNoteAnnotation  = "lark_flag_note"  // extra lines shown indented under a flag
+
+	groupParams    = "params"    // typed path/query flags
+	groupBody      = "body"      // --data, --file
+	groupRaw       = "raw"       // --params
+	groupExecution = "execution" // --as/--dry-run/--page-*/--yes
+	groupOutput    = "output"    // --output/--format/--jq
+
+	subRequired = "required"
+	subOptional = "optional"
+)
+
+// serviceFlagGroupOrder is the display order + titles of the flag groups. API
+// Parameters carries only typed path/query flags; raw --params, request body and
+// execution/output controls each get their own group so an agent can tell the
+// distinct input kinds apart.
+var serviceFlagGroupOrder = []struct{ key, title string }{
+	{groupParams, "API Parameters"},
+	{groupBody, "Request Body"},
+	{groupRaw, "Raw Parameter Input"},
+	{groupExecution, "Execution"},
+	{groupOutput, "Output"},
+}
+
+// applyGroupedUsage installs the grouped usage renderer on a service method
+// cmd: local flags via the grouped renderer instead of cobra's flat Flags:
+// list; global (inherited) flags and the Risk/Tips sections appended by the
+// root help func are unaffected. Rendered by hand rather than via
+// cmd.SetUsageTemplate: cobra lazy-links text/template on the first
+// SetUsageTemplate call, whose executor reaches reflect.Value.MethodByName —
+// that disables the linker's method-level dead-code elimination and costs
+// ~19 MB of binary size.
+func applyGroupedUsage(cmd *cobra.Command) {
+	cmd.SetUsageFunc(func(c *cobra.Command) error {
+		w := c.OutOrStderr()
+		fmt.Fprintf(w, "Usage:\n  %s\n", c.UseLine())
+		if c.HasAvailableLocalFlags() {
+			fmt.Fprintf(w, "\n%s\n", renderServiceFlagGroups(c))
+		}
+		if c.HasAvailableInheritedFlags() {
+			fmt.Fprintf(w, "\nGlobal Flags:\n%s\n", strings.TrimRight(c.InheritedFlags().FlagUsages(), " \t\n"))
+		}
+		return nil
+	})
+}
+
+func annotate(f *pflag.Flag, key string, vals []string) {
+	if f.Annotations == nil {
+		f.Annotations = map[string][]string{}
+	}
+	f.Annotations[key] = vals
+}
+
+// tagFlagGroup records a flag's display group (no-op if the flag is absent).
+func tagFlagGroup(fs *pflag.FlagSet, name, group string) {
+	if f := fs.Lookup(name); f != nil {
+		annotate(f, flagGroupAnnotation, []string{group})
+	}
+}
+
+func annotationOf(f *pflag.Flag, key string) []string {
+	if f.Annotations != nil {
+		return f.Annotations[key]
+	}
+	return nil
+}
+
+func flagGroupOf(f *pflag.Flag) string {
+	if v := annotationOf(f, flagGroupAnnotation); len(v) > 0 {
+		return v[0]
+	}
+	return ""
+}
+
+func flagSubOf(f *pflag.Flag) string {
+	if v := annotationOf(f, flagSubAnnotation); len(v) > 0 {
+		return v[0]
+	}
+	return ""
+}
+
+// renderServiceFlagGroups renders the command's local flags into ordered,
+// titled groups; the API Parameters group is further split into Required /
+// Optional. It is the body of the usage func applyGroupedUsage installs.
+func renderServiceFlagGroups(cmd *cobra.Command) string {
+	var b strings.Builder
+	seen := map[*pflag.Flag]bool{}
+	for _, g := range serviceFlagGroupOrder {
+		flags := groupFlags(cmd, g.key, seen)
+		if len(flags) == 0 {
+			continue
+		}
+		fmt.Fprintf(&b, "%s:\n", g.title)
+		if g.key == groupParams {
+			writeSection(&b, "  Required:", subFlags(flags, subRequired))
+			writeSection(&b, "  Optional:", subFlags(flags, subOptional))
+		} else {
+			writeSection(&b, "", flags)
+		}
+		fmt.Fprintln(&b)
+	}
+	// Anything untagged (e.g. -h/--help) goes last under "Other".
+	var other []*pflag.Flag
+	cmd.LocalFlags().VisitAll(func(f *pflag.Flag) {
+		if f.Hidden || seen[f] {
+			return
+		}
+		other = append(other, f)
+	})
+	if len(other) > 0 {
+		fmt.Fprintln(&b, "Other:")
+		writeSection(&b, "", other)
+	}
+	return strings.TrimRight(b.String(), "\n")
+}
+
+// groupFlags returns the visible local flags tagged with group key, marking them
+// seen so the trailing "Other" bucket only catches genuinely untagged flags.
+func groupFlags(cmd *cobra.Command, key string, seen map[*pflag.Flag]bool) []*pflag.Flag {
+	var flags []*pflag.Flag
+	cmd.LocalFlags().VisitAll(func(f *pflag.Flag) {
+		if f.Hidden || flagGroupOf(f) != key {
+			return
+		}
+		flags = append(flags, f)
+		seen[f] = true
+	})
+	return flags
+}
+
+func subFlags(flags []*pflag.Flag, sub string) []*pflag.Flag {
+	var out []*pflag.Flag
+	for _, f := range flags {
+		s := flagSubOf(f)
+		// Untagged subgroup defaults to Optional so nothing is dropped.
+		if s == sub || (s == "" && sub == subOptional) {
+			out = append(out, f)
+		}
+	}
+	return out
+}
+
+// writeSection prints an optional (sub)header and the flags, aligned in a
+// column, each flag row followed by its note lines indented under the usage.
+func writeSection(b *strings.Builder, header string, flags []*pflag.Flag) {
+	if len(flags) == 0 {
+		return
+	}
+	if header != "" {
+		fmt.Fprintf(b, "%s\n", header)
+	}
+	specs := make([]string, len(flags))
+	maxSpec := 0
+	for i, f := range flags {
+		specs[i] = flagSpec(f)
+		if len(specs[i]) > maxSpec {
+			maxSpec = len(specs[i])
+		}
+	}
+	for i, f := range flags {
+		_, usage := pflag.UnquoteUsage(f)
+		if showsDefault(f) {
+			usage += fmt.Sprintf(" (default %s)", f.DefValue)
+		}
+		fmt.Fprintf(b, "%-*s   %s\n", maxSpec, specs[i], strings.TrimSpace(usage))
+		for _, note := range annotationOf(f, flagNoteAnnotation) {
+			fmt.Fprintf(b, "%*s%s\n", maxSpec+3+4, "", note)
+		}
+	}
+}
+
+// flagSpec is pflag's "      --name type" / "  -x, --name type" left column.
+func flagSpec(f *pflag.Flag) string {
+	typeName, _ := pflag.UnquoteUsage(f)
+	spec := "      --" + f.Name
+	if f.Shorthand != "" && f.ShorthandDeprecated == "" {
+		spec = "  -" + f.Shorthand + ", --" + f.Name
+	}
+	if typeName != "" {
+		spec += " " + typeName
+	}
+	return spec
+}
+
+// showsDefault mirrors pflag's "non-zero default" rule for the flag types these
+// commands use, so the grouped rendering shows the same "(default x)" hints as
+// cobra's flat list.
+func showsDefault(f *pflag.Flag) bool {
+	switch f.DefValue {
+	case "", "0", "false", "[]":
+		return false
+	}
+	return true
+}

cmd/service/flaggroups_test.go

@@ -0,0 +1,115 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/meta"
+)
+
+func TestServiceFlagGroups_AgentContract(t *testing.T) {
+	method := map[string]interface{}{
+		"path":       "chats/:chat_id/members",
+		"httpMethod": "POST",
+		"parameters": map[string]interface{}{
+			"chat_id": map[string]interface{}{"type": "string", "location": "path", "required": true},
+			"member_id_type": map[string]interface{}{
+				"type": "string", "location": "query",
+				"options": []interface{}{
+					map[string]interface{}{"value": "open_id", "description": "以 open_id 标识用户"},
+					map[string]interface{}{"value": "user_id", "description": "以 user_id 标识用户"},
+				},
+			},
+		},
+		// Documented body field -> --data belongs under Request Body.
+		"requestBody": map[string]interface{}{
+			"id_list": map[string]interface{}{"type": "list", "required": true},
+		},
+	}
+	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(method), "create", "chat.members", nil)
+	out := renderServiceFlagGroups(cmd)
+
+	idx := func(s string) int { return strings.Index(out, s) }
+
+	// Section order: API Parameters → Request Body → Raw Parameter Input → Execution → Output.
+	iParams, iBody, iRaw, iExec, iOut := idx("API Parameters:"), idx("Request Body:"), idx("Raw Parameter Input:"), idx("Execution:"), idx("Output:")
+	for name, i := range map[string]int{"API Parameters": iParams, "Request Body": iBody, "Raw Parameter Input": iRaw, "Execution": iExec, "Output": iOut} {
+		if i < 0 {
+			t.Fatalf("missing section %q in:\n%s", name, out)
+		}
+	}
+	if !(iParams < iBody && iBody < iRaw && iRaw < iExec && iExec < iOut) {
+		t.Errorf("section order wrong:\n%s", out)
+	}
+
+	// Required/Optional subsections under API Parameters.
+	if i := idx("  Required:"); i < iParams || i > iBody {
+		t.Errorf("Required subsection misplaced:\n%s", out)
+	}
+	if i := idx("  Optional:"); i < iParams || i > iBody {
+		t.Errorf("Optional subsection misplaced:\n%s", out)
+	}
+
+	// Typed flags are API Parameters; required path flag under Required, enum
+	// flag under Optional with an inline "enum: ..." (not multi-line meanings).
+	if i := idx("--chat-id"); i < iParams || i > iBody {
+		t.Errorf("--chat-id not under API Parameters:\n%s", out)
+	}
+	if !strings.Contains(out, "chat_id, required") {
+		t.Errorf("typed flag help format wrong:\n%s", out)
+	}
+	if !strings.Contains(out, "enum: open_id=以 open_id 标识用户|user_id=以 user_id 标识用户") {
+		t.Errorf("expected compact enum value=meaning inline:\n%s", out)
+	}
+
+	// --data is Request Body; --params is Raw Parameter Input (NOT API Parameters)
+	// and carries the precedence rule.
+	if i := idx("--data"); i < iBody || i > iRaw {
+		t.Errorf("--data not under Request Body:\n%s", out)
+	}
+	if i := idx("--params"); i < iRaw || i > iExec {
+		t.Errorf("--params not under Raw Parameter Input:\n%s", out)
+	}
+	if !strings.Contains(out, "typed flags override matching keys in --params") {
+		t.Errorf("missing --params precedence rule:\n%s", out)
+	}
+
+	// Control flags land in Execution/Output.
+	if i := idx("--dry-run"); i < iExec || i > iOut {
+		t.Errorf("--dry-run not under Execution:\n%s", out)
+	}
+	if idx("--format") < iOut {
+		t.Errorf("--format not under Output:\n%s", out)
+	}
+
+	// The usage template is wired to the grouped renderer (no flat Flags: list).
+	if u := cmd.UsageString(); !strings.Contains(u, "API Parameters:") || strings.Contains(u, "\nFlags:\n") {
+		t.Errorf("usage template not grouped:\n%s", u)
+	}
+}
+
+// TestServiceFlagGroups_UndocumentedBodyIsRaw: a POST with no documented body
+// fields still offers --data (escape hatch) but must NOT imply a declared body —
+// it goes under Raw Parameter Input, not "Request Body".
+func TestServiceFlagGroups_UndocumentedBodyIsRaw(t *testing.T) {
+	method := map[string]interface{}{"path": "things/do", "httpMethod": "POST"} // POST, no requestBody, no params
+	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(method), "do", "things", nil)
+	out := renderServiceFlagGroups(cmd)
+
+	if strings.Contains(out, "Request Body:") {
+		t.Errorf("undocumented body must not render a Request Body section:\n%s", out)
+	}
+	iRaw, iData := strings.Index(out, "Raw Parameter Input:"), strings.Index(out, "--data")
+	if iRaw < 0 || iData < iRaw {
+		t.Errorf("--data not under Raw Parameter Input:\n%s", out)
+	}
+	if !strings.Contains(out, "no documented fields") {
+		t.Errorf("--data should be labeled a raw escape hatch:\n%s", out)
+	}
+}

cmd/service/paramflags.go

@@ -0,0 +1,166 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/meta"
+	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
+)
+
+type boundParamFlag struct {
+	field meta.Field
+	read  func() interface{}
+}
+
+// paramsOnlyField is a path/query parameter that got no typed flag because its
+// kebab name is already taken by another flag (a standard flag like --format, or
+// a root persistent flag). It stays reachable via --params; the binder keeps it,
+// with the flag that claimed the name, so --help can show the exact --params form
+// and steer the reader off the wrong flag.
+type paramsOnlyField struct {
+	field   meta.Field
+	claimed *pflag.Flag
+}
+
+// paramFlagBinder owns one service method's generated typed param flags: it
+// registers them (kind, help, enum completion, reserved-name skip) and applies
+// the --params overlay, where a changed typed flag overrides its key in the
+// --params JSON. Holding the field<->flag binding here keeps the request builder
+// from re-deriving which flags map to which param keys.
+type paramFlagBinder struct {
+	bound      []boundParamFlag
+	paramsOnly []paramsOnlyField
+}
+
+// newParamFlagBinder registers one typed kebab flag per path/query parameter on
+// cmd and returns a binder for the --params overlay. A name already taken by
+// another flag is skipped — pflag panics on a local duplicate and a generated
+// flag would silently shadow a persistent one — and recorded as paramsOnly so
+// the parameter stays reachable (and discoverable) via --params. The taken set
+// is derived, not hand-listed: local flags (the standard set, registered before
+// this runs) via cmd, the lazily-added --help materialized here, and the root's
+// persistent flags via reserved (nil for direct callers that have no root).
+func newParamFlagBinder(cmd *cobra.Command, params []meta.Field, reserved *pflag.FlagSet) *paramFlagBinder {
+	cmd.InitDefaultHelpFlag() // materialize --help/-h so the local guard below sees it
+	b := &paramFlagBinder{}
+	for _, f := range params {
+		name := f.FlagName()
+		if claimed := flagClaiming(cmd, reserved, name); claimed != nil {
+			b.paramsOnly = append(b.paramsOnly, paramsOnlyField{field: f, claimed: claimed})
+			continue
+		}
+		read := registerTypedFlag(cmd.Flags(), name, f.CanonicalType(), paramFlagUsage(f))
+		if values := enumStrings(f.EnumValues()); len(values) > 0 {
+			cmdutil.RegisterFlagCompletion(cmd, name, func(_ *cobra.Command, _ []string, _ string) ([]string, cobra.ShellCompDirective) {
+				return values, cobra.ShellCompDirectiveNoFileComp
+			})
+		}
+		// Group as an API parameter and mark required/optional for the
+		// Required/Optional subsections of the grouped --help renderer.
+		if fl := cmd.Flags().Lookup(name); fl != nil {
+			annotate(fl, flagGroupAnnotation, []string{groupParams})
+			sub := subOptional
+			if f.Required {
+				sub = subRequired
+			}
+			annotate(fl, flagSubAnnotation, []string{sub})
+		}
+		b.bound = append(b.bound, boundParamFlag{field: f, read: read})
+	}
+	return b
+}
+
+// flagClaiming returns the flag already occupying name (so a typed param flag
+// would collide), or nil when the name is free. It checks the command's own
+// flags (the standard set + the materialized --help) and the root's persistent
+// flags — so the reserved set is whatever is actually registered, never a
+// hand-kept list that drifts when a global flag is added.
+func flagClaiming(cmd *cobra.Command, reserved *pflag.FlagSet, name string) *pflag.Flag {
+	if fl := cmd.Flags().Lookup(name); fl != nil {
+		return fl
+	}
+	if reserved != nil {
+		return reserved.Lookup(name)
+	}
+	return nil
+}
+
+// paramsOnlyHelp renders the --help addendum for parameters that have no typed
+// flag, or "" when there are none. Per field: a copy-pasteable --params form,
+// the same fieldFacts a typed flag would show on its usage line, and what the
+// colliding flag actually does — so neither a human nor an agent sets the
+// wrong one (e.g. --format, which is the output format, not the API parameter).
+func (b *paramFlagBinder) paramsOnlyHelp() string {
+	if len(b.paramsOnly) == 0 {
+		return ""
+	}
+	var sb strings.Builder
+	sb.WriteString("\nParameters set via --params (no typed flag; the name is taken by another flag):\n")
+	for _, p := range b.paramsOnly {
+		name := p.field.Name
+		fmt.Fprintf(&sb, "  %s: --params '{%q: %s}'\n", name, name, paramExample(p.field))
+		for _, fact := range fieldFacts(p.field) {
+			fmt.Fprintf(&sb, "      %s\n", fact)
+		}
+		if p.claimed != nil {
+			fmt.Fprintf(&sb, "      do not use --%s (%s)\n", p.claimed.Name, p.claimed.Usage)
+		}
+	}
+	return sb.String()
+}
+
+// hasTypedFlag reports whether the binder registered a typed flag for the
+// param named name. False for params-only fields — a flag with the same kebab
+// name may exist (that's the collision), but it is not this param's input.
+// Nil-safe for direct buildServiceRequest callers that have no binder.
+func (b *paramFlagBinder) hasTypedFlag(name string) bool {
+	if b == nil {
+		return false
+	}
+	for _, pf := range b.bound {
+		if pf.field.Name == name {
+			return true
+		}
+	}
+	return false
+}
+
+// overlay lets an explicit typed flag override the same key in --params
+// (--params is the base). Only changed flags apply, so the --params-only path is
+// unchanged. A nil binder or cmd is a no-op.
+func (b *paramFlagBinder) overlay(cmd *cobra.Command, params map[string]interface{}) {
+	if b == nil || cmd == nil {
+		return
+	}
+	for _, pf := range b.bound {
+		if cmd.Flags().Changed(pf.field.FlagName()) {
+			params[pf.field.Name] = pf.read()
+		}
+	}
+}
+
+// registerTypedFlag registers one flag of the given canonical JSON-Schema kind
+// and returns a reader for its parsed value; the kind→pflag-type switch lives
+// only here.
+func registerTypedFlag(fs *pflag.FlagSet, name, kind, usage string) func() interface{} {
+	switch kind {
+	case "integer":
+		return flagReader(fs.Int(name, 0, usage))
+	case "boolean":
+		return flagReader(fs.Bool(name, false, usage))
+	case "array":
+		return flagReader(fs.StringArray(name, nil, usage))
+	default:
+		return flagReader(fs.String(name, "", usage))
+	}
+}
+
+func flagReader[T any](p *T) func() interface{} {
+	return func() interface{} { return *p }
+}

cmd/service/paramflags_test.go

@@ -0,0 +1,626 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"errors"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/meta"
+	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
+)
+
+// imChatMembersCreate: POST chats/{chat_id}/members with one path param and one
+// optional enum query param — the canonical case from the screenshot feedback.
+func imChatMembersCreate() meta.Method {
+	return meta.FromMap(map[string]interface{}{
+		"path":       "chats/{chat_id}/members",
+		"httpMethod": "POST",
+		"parameters": map[string]interface{}{
+			"chat_id": map[string]interface{}{
+				"type": "string", "location": "path", "required": true,
+			},
+			"member_id_type": map[string]interface{}{
+				"type": "string", "location": "query", "required": false,
+				"options": []interface{}{
+					map[string]interface{}{"value": "open_id"},
+					map[string]interface{}{"value": "user_id"},
+				},
+			},
+		},
+	})
+}
+
+func TestServiceMethod_TypedFlagRegistered(t *testing.T) {
+	f := &cmdutil.Factory{}
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+
+	if cmd.Flags().Lookup("chat-id") == nil {
+		t.Error("expected generated --chat-id flag for path param chat_id")
+	}
+	if cmd.Flags().Lookup("member-id-type") == nil {
+		t.Error("expected generated --member-id-type flag for query param member_id_type")
+	}
+}
+
+// A query param literally named "format" kebab-collides with the global
+// --format flag. Generation must skip it (never re-register, never panic) and
+// leave the standard --format flag intact.
+func TestServiceMethod_TypedFlagReservedCollisionSkipped(t *testing.T) {
+	method := map[string]interface{}{
+		"path":       "messages",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"format": map[string]interface{}{"type": "string", "location": "query"},
+		},
+	}
+
+	var cmd *cobra.Command
+	func() {
+		defer func() {
+			if r := recover(); r != nil {
+				t.Fatalf("flag generation panicked on reserved-name collision: %v", r)
+			}
+		}()
+		cmd = NewCmdServiceMethod(&cmdutil.Factory{}, imSpec(), meta.FromMap(method), "list", "messages", nil)
+	}()
+
+	fl := cmd.Flags().Lookup("format")
+	if fl == nil || fl.DefValue != "json" {
+		t.Fatalf("standard --format flag must be preserved, got %+v", fl)
+	}
+}
+
+func TestServiceMethod_TypedFlag_DrivesPathParam(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--chat-id", "oc_abc123", "--data", `{"id_list":["ou_x"]}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !strings.Contains(stdout.String(), "chats/oc_abc123/members") {
+		t.Errorf("expected URL with chat_id substituted from --chat-id, got:\n%s", stdout.String())
+	}
+}
+
+func TestServiceMethod_TypedFlag_DrivesQueryParam(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--chat-id", "oc_abc123", "--member-id-type", "open_id", "--data", `{}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	if !strings.Contains(out, "member_id_type") || !strings.Contains(out, "open_id") {
+		t.Errorf("expected query param member_id_type=open_id from flag, got:\n%s", out)
+	}
+}
+
+func TestServiceMethod_TypedFlag_AgreesWithParams(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--chat-id", "oc_abc123", "--params", `{"chat_id":"oc_abc123"}`, "--data", `{}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("same value via flag and --params should be accepted, got: %v", err)
+	}
+	if !strings.Contains(stdout.String(), "chats/oc_abc123/members") {
+		t.Errorf("expected URL with chat_id, got:\n%s", stdout.String())
+	}
+}
+
+// --params is the base; an explicit typed flag overrides the same key.
+func TestServiceMethod_TypedFlag_OverridesParams(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--chat-id", "oc_flag", "--params", `{"chat_id":"oc_params"}`, "--data", `{}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	if !strings.Contains(out, "chats/oc_flag/members") {
+		t.Errorf("expected --chat-id to override --params chat_id, got:\n%s", out)
+	}
+	if strings.Contains(out, "oc_params") {
+		t.Errorf("--params value should have been overridden by the flag, got:\n%s", out)
+	}
+}
+
+// Override works for a non-string (integer) param too, exercising the int
+// register/read path end to end.
+func TestServiceMethod_TypedFlag_IntegerOverridesParams(t *testing.T) {
+	method := map[string]interface{}{
+		"path":       "messages",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"page_size": map[string]interface{}{"type": "integer", "location": "query"},
+		},
+	}
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(method), "list", "messages", nil)
+	cmd.SetArgs([]string{"--page-size", "100", "--params", `{"page_size":5}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	if !strings.Contains(out, "page_size") || !strings.Contains(out, "100") {
+		t.Errorf("expected --page-size 100 to override --params page_size=5, got:\n%s", out)
+	}
+}
+
+// Regression: with no typed flags passed, behavior is byte-identical to today.
+func TestServiceMethod_TypedFlag_OnlyParamsStillWorks(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--params", `{"chat_id":"oc_abc123"}`, "--data", `{}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !strings.Contains(stdout.String(), "chats/oc_abc123/members") {
+		t.Errorf("expected URL with chat_id from --params, got:\n%s", stdout.String())
+	}
+}
+
+// Regression: --params null is valid JSON that unmarshals to a nil map. A typed
+// flag overlaying onto it must not panic (assignment to a nil map) — null is
+// treated as "no base params", with the flag value applied on top.
+func TestServiceMethod_TypedFlag_OverridesNullParams(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--chat-id", "oc_abc123", "--params", "null", "--data", `{}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("--params null with a typed flag should not error, got: %v", err)
+	}
+	if !strings.Contains(stdout.String(), "chats/oc_abc123/members") {
+		t.Errorf("expected chat_id from --chat-id over null --params, got:\n%s", stdout.String())
+	}
+}
+
+// Startup smoke test: registering every embedded method must not panic on a
+// generated-flag name collision (pflag panics on duplicate registration, which
+// would crash the whole CLI at startup), and a known path param must surface as
+// a typed flag end to end.
+func TestRegisterServiceCommands_GeneratesFlagsNoPanic(t *testing.T) {
+	root := &cobra.Command{Use: "lark-cli"}
+	f := &cmdutil.Factory{}
+
+	defer func() {
+		if r := recover(); r != nil {
+			t.Fatalf("registering all service commands panicked: %v", r)
+		}
+	}()
+	RegisterServiceCommands(root, f)
+
+	create, _, err := root.Find([]string{"im", "chat.members", "create"})
+	if err != nil {
+		t.Fatalf("im chat.members create not registered: %v", err)
+	}
+	if create.Flags().Lookup("chat-id") == nil {
+		t.Error("expected generated --chat-id flag on im chat.members create")
+	}
+}
+
+// Locks the boolean and array branches of bindParamFlag end to end (string and
+// integer are covered above): a bool flag yields true and a repeatable array
+// flag yields all its elements in the request.
+func TestServiceMethod_TypedFlag_BoolAndArrayKinds(t *testing.T) {
+	method := map[string]interface{}{
+		"path":       "items",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"with_deleted": map[string]interface{}{"type": "boolean", "location": "query"},
+			"ids":          map[string]interface{}{"type": "list", "location": "query"},
+		},
+	}
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(method), "list", "items", nil)
+	cmd.SetArgs([]string{"--with-deleted", "--ids", "a", "--ids", "b", "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	for _, want := range []string{"with_deleted", "true", "ids", "\"a\"", "\"b\""} {
+		if !strings.Contains(out, want) {
+			t.Errorf("expected dry-run output to contain %q, got:\n%s", want, out)
+		}
+	}
+}
+
+// Override (--params base, typed flag wins) is covered for string and integer
+// above; this locks the same semantics for the boolean and array kinds.
+func TestServiceMethod_TypedFlag_BoolAndArrayOverrideParams(t *testing.T) {
+	method := map[string]interface{}{
+		"path":       "items",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"with_deleted": map[string]interface{}{"type": "boolean", "location": "query"},
+			"ids":          map[string]interface{}{"type": "list", "location": "query"},
+		},
+	}
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(method), "list", "items", nil)
+	cmd.SetArgs([]string{
+		"--params", `{"with_deleted":false,"ids":["from_params"]}`,
+		"--with-deleted", "--ids", "a", "--ids", "b",
+		"--dry-run",
+	})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	for _, want := range []string{"with_deleted", "true", "\"a\"", "\"b\""} {
+		if !strings.Contains(out, want) {
+			t.Errorf("expected flag to override --params (want %q), got:\n%s", want, out)
+		}
+	}
+	if strings.Contains(out, "from_params") {
+		t.Errorf("--params array value should have been overridden by --ids, got:\n%s", out)
+	}
+}
+
+// A param whose kebab name collides with a global flag (here "format" vs the
+// global --format) gets no typed flag, but the collision is no longer silent:
+// non-colliding params still get flags, the global --format is untouched, and
+// --help shows the exact --params form and steers the reader off --format.
+func TestServiceMethod_ParamsOnly_HelpSteersToParams(t *testing.T) {
+	method := map[string]interface{}{
+		"path":       "things/{thing_id}",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"thing_id": map[string]interface{}{"type": "string", "location": "path", "required": true},
+			"format": map[string]interface{}{"type": "string", "location": "query", "min": "1", "max": "64", "description": "返回的消息体格式。", "options": []interface{}{
+				map[string]interface{}{"value": "full"},
+				map[string]interface{}{"value": "metadata"},
+			}},
+		},
+	}
+	cmd := NewCmdServiceMethod(&cmdutil.Factory{}, imSpec(), meta.FromMap(method), "get", "things", nil)
+
+	if cmd.Flags().Lookup("thing-id") == nil {
+		t.Error("non-colliding param should still get a typed --thing-id flag")
+	}
+	if fl := cmd.Flags().Lookup("format"); fl == nil || fl.DefValue != "json" {
+		t.Fatalf("global --format must be preserved (not shadowed), got %+v", fl)
+	}
+	for _, want := range []string{`--params '{"format"`, "返回的消息体格式", "full", "metadata", "min: 1, max: 64", "do not use --format"} {
+		if !strings.Contains(cmd.Long, want) {
+			t.Errorf("help should contain %q so the reader uses --params, not --format; got:\n%s", want, cmd.Long)
+		}
+	}
+}
+
+// The collision guard derives reserved names from the actual flag sets — local
+// flags plus the root's persistent flags passed in — so a future persistent
+// flag is covered with no hand-maintained list. Here a param named "profile"
+// (a root persistent flag) is skipped while a normal param is bound.
+func TestParamFlagBinder_PersistentFlagReserved(t *testing.T) {
+	cmd := &cobra.Command{Use: "x"}
+	reserved := pflag.NewFlagSet("root", pflag.ContinueOnError)
+	reserved.String("profile", "", "use a specific profile")
+
+	m := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+		"profile": map[string]interface{}{"type": "string", "location": "query"},
+		"id":      map[string]interface{}{"type": "string", "location": "path"},
+	}})
+	b := newParamFlagBinder(cmd, m.Params(), reserved)
+
+	if cmd.Flags().Lookup("id") == nil {
+		t.Error("non-colliding param should get a typed flag")
+	}
+	if cmd.Flags().Lookup("profile") != nil {
+		t.Error("param colliding with a reserved persistent flag must not be registered")
+	}
+	found := false
+	for _, p := range b.paramsOnly {
+		if p.field.Name == "profile" {
+			found = true
+		}
+	}
+	if !found {
+		t.Error("colliding param should be recorded for the --params help note")
+	}
+}
+
+// boolIntQueryMethod is the fixture for the zero-value semantics tests: one
+// boolean and one integer query param, where false and 0 are meaningful values.
+func boolIntQueryMethod(required bool) meta.Method {
+	return meta.FromMap(map[string]interface{}{
+		"path":       "items",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"with_deleted": map[string]interface{}{"type": "boolean", "location": "query", "required": required},
+			"page_size":    map[string]interface{}{"type": "integer", "location": "query"},
+		},
+	})
+}
+
+// Presence is intent: a typed flag is only overlaid when explicitly Changed,
+// so --flag=false / --flag 0 are real values and must be sent — not silently
+// dropped as "empty", which would let the API default win over an explicit
+// user choice.
+func TestServiceMethod_TypedFlag_ExplicitFalseAndZeroAreSent(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), boolIntQueryMethod(false), "list", "items", nil)
+	cmd.SetArgs([]string{"--with-deleted=false", "--page-size", "0", "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	for _, want := range []string{`"with_deleted": false`, `"page_size": 0`} {
+		if !strings.Contains(out, want) {
+			t.Errorf("explicit zero value must be sent (want %s), got:\n%s", want, out)
+		}
+	}
+}
+
+// An explicitly provided false satisfies a required query parameter — the
+// pre-flight must not report "missing" for a value the user just set.
+func TestServiceMethod_TypedFlag_ExplicitFalseSatisfiesRequired(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), boolIntQueryMethod(true), "list", "items", nil)
+	cmd.SetArgs([]string{"--with-deleted=false", "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("required param explicitly set to false must pass pre-flight, got: %v", err)
+	}
+	if !strings.Contains(stdout.String(), `"with_deleted": false`) {
+		t.Errorf("explicit false must be sent, got:\n%s", stdout.String())
+	}
+}
+
+// The same presence-is-intent rule applies to the --params JSON base: a key
+// deliberately written as false/0 is sent. (Zero values used to be silently
+// dropped; this locks the corrected semantics as the contract.)
+func TestServiceMethod_Params_JSONZeroValuesAreSent(t *testing.T) {
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), boolIntQueryMethod(false), "list", "items", nil)
+	cmd.SetArgs([]string{"--params", `{"with_deleted":false,"page_size":0}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	for _, want := range []string{`"with_deleted": false`, `"page_size": 0`} {
+		if !strings.Contains(out, want) {
+			t.Errorf("--params zero value must be sent (want %s), got:\n%s", want, out)
+		}
+	}
+}
+
+// "" stays unusable: a required parameter fed an empty-string placeholder is
+// still caught by the friendly pre-flight error, not sent as an empty value.
+func TestServiceMethod_Params_EmptyStringStillMissing(t *testing.T) {
+	method := meta.FromMap(map[string]interface{}{
+		"path":       "items",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"user_id_type": map[string]interface{}{"type": "string", "location": "query", "required": true},
+		},
+	})
+	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), method, "list", "items", nil)
+	cmd.SetArgs([]string{"--params", `{"user_id_type":""}`, "--dry-run"})
+
+	err := cmd.Execute()
+	if err == nil || !strings.Contains(err.Error(), "missing required query parameter") {
+		t.Fatalf("empty string for a required param should still pre-flight error, got: %v", err)
+	}
+}
+
+// A declared optional query param fed "" is dropped (unusable value), not sent
+// as an empty query value — the declared-param loop owns the decision and the
+// undeclared passthrough must not resurrect it. Undeclared keys stay the
+// verbatim raw escape hatch.
+func TestServiceMethod_Params_EmptyOptionalDroppedUndeclaredKept(t *testing.T) {
+	method := meta.FromMap(map[string]interface{}{
+		"path":       "items",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"user_id_type": map[string]interface{}{"type": "string", "location": "query"},
+		},
+	})
+	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), method, "list", "items", nil)
+	cmd.SetArgs([]string{"--params", `{"user_id_type":"","custom_key":"v1"}`, "--dry-run"})
+
+	if err := cmd.Execute(); err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	out := stdout.String()
+	if strings.Contains(out, "user_id_type") {
+		t.Errorf("declared optional param with empty value must be dropped, got:\n%s", out)
+	}
+	if !strings.Contains(out, `"custom_key": "v1"`) {
+		t.Errorf("undeclared key must pass through verbatim, got:\n%s", out)
+	}
+}
+
+// min/max from the metadata surface on the typed flag's help line, in the same
+// vocabulary as the envelope's minimum/maximum.
+func TestParamFlagUsage_Bounds(t *testing.T) {
+	cases := []struct{ name, min, max, want string }{
+		{"both", "1", "100", "min: 1, max: 100"},
+		{"min only", "1", "", "min: 1"},
+		{"max only", "", "64", "max: 64"},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			fields := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+				"page_size": map[string]interface{}{"type": "integer", "location": "query", "min": tc.min, "max": tc.max},
+			}}).Params()
+			if usage := paramFlagUsage(fields[0]); !strings.Contains(usage, tc.want) {
+				t.Errorf("usage = %q, want contains %q", usage, tc.want)
+			}
+		})
+	}
+	t.Run("no bounds, no clause", func(t *testing.T) {
+		fields := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+			"page_token": map[string]interface{}{"type": "string", "location": "query"},
+		}}).Params()
+		if usage := paramFlagUsage(fields[0]); strings.Contains(usage, "min:") || strings.Contains(usage, "max:") {
+			t.Errorf("usage without bounds should not mention min/max, got %q", usage)
+		}
+	})
+}
+
+// The sanitized field description rides the help line — a bare name like
+// user_mailbox_id carries no meaning. The cut is at note separators (;), NOT
+// at sentence ends (。): the later sentence often holds the key affordance.
+func TestParamFlagUsage_Description(t *testing.T) {
+	fields := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+		"user_mailbox_id": map[string]interface{}{
+			"type": "string", "location": "path", "required": true,
+			"description": `用户邮箱地址。当使用用户身份访问时，可以输入"me"代表当前调用接口用户;后续补充说明不该出现`,
+		},
+	}}).Params()
+	usage := paramFlagUsage(fields[0])
+	if !strings.Contains(usage, `可以输入"me"代表当前调用接口用户`) {
+		t.Errorf("description must keep full sentences up to the note separator, got %q", usage)
+	}
+	if strings.Contains(usage, "补充说明") {
+		t.Errorf("text after the note separator must be cut, got %q", usage)
+	}
+
+	t.Run("long description truncated", func(t *testing.T) {
+		fields := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+			"x": map[string]interface{}{
+				"type": "string", "location": "query",
+				"description": strings.Repeat("长", 80),
+			},
+		}}).Params()
+		usage := paramFlagUsage(fields[0])
+		if !strings.Contains(usage, "...") {
+			t.Errorf("long description should be truncated with ellipsis, got %q", usage)
+		}
+		if strings.Contains(usage, strings.Repeat("长", 61)) {
+			t.Errorf("description should not exceed the cap, got %q", usage)
+		}
+	})
+
+	t.Run("trailing sentence punctuation trimmed", func(t *testing.T) {
+		fields := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+			"x": map[string]interface{}{
+				"type": "string", "location": "query", "description": "返回格式。",
+			},
+		}}).Params()
+		if usage := paramFlagUsage(fields[0]); strings.Contains(usage, "。.") {
+			t.Errorf("clause join must not double the punctuation, got %q", usage)
+		}
+	})
+}
+
+// Pins the convergence contract: the params-only addendum renders the SAME
+// fieldFacts list the typed flag's usage line joins inline — a fact added to
+// fieldFacts reaches both surfaces, and neither can drift over what a param's
+// help says (the addendum once rendered values-only enums and silently lacked
+// the API default).
+func TestParamHelp_BothSurfacesRenderFieldFacts(t *testing.T) {
+	f := meta.FromMap(map[string]interface{}{"parameters": map[string]interface{}{
+		"mode": map[string]interface{}{
+			"type": "string", "location": "query",
+			"description": "模式选择。",
+			"default":     "fast",
+			"min":         "1", "max": "8",
+			"options": []interface{}{
+				map[string]interface{}{"value": "fast", "description": "快速"},
+				map[string]interface{}{"value": "full"},
+			},
+		},
+	}}).Params()[0]
+
+	facts := fieldFacts(f)
+	if len(facts) != 4 { // description, enum, bounds, API default
+		t.Fatalf("fieldFacts = %v, want 4 facts", facts)
+	}
+	usage := paramFlagUsage(f)
+	help := (&paramFlagBinder{paramsOnly: []paramsOnlyField{{field: f}}}).paramsOnlyHelp()
+	for _, fact := range facts {
+		if !strings.Contains(usage, fact) {
+			t.Errorf("usage line missing fact %q: %q", fact, usage)
+		}
+		if !strings.Contains(help, fact) {
+			t.Errorf("params-only addendum missing fact %q:\n%s", fact, help)
+		}
+	}
+}
+
+// Bounds reach the registered flag's help end to end.
+func TestServiceMethod_TypedFlag_HelpShowsBounds(t *testing.T) {
+	method := meta.FromMap(map[string]interface{}{
+		"path":       "items",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"page_size": map[string]interface{}{"type": "integer", "location": "query", "min": "1", "max": "100", "default": "20"},
+		},
+	})
+	cmd := NewCmdServiceMethod(&cmdutil.Factory{}, imSpec(), method, "list", "items", nil)
+	fl := cmd.Flags().Lookup("page-size")
+	if fl == nil {
+		t.Fatal("expected generated --page-size flag")
+	}
+	if !strings.Contains(fl.Usage, "min: 1, max: 100") {
+		t.Errorf("flag usage should carry bounds, got %q", fl.Usage)
+	}
+}
+
+// The missing-required hint must name both recovery paths — the typed flag and
+// the --params fallback — so a reader who only knows one input style can
+// proceed without a round-trip through schema.
+func TestServiceMethod_MissingRequired_HintNamesFlagAndParams(t *testing.T) {
+	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), imChatMembersCreate(), "create", "chat.members", nil)
+	cmd.SetArgs([]string{"--data", `{"id_list":["ou_x"]}`, "--dry-run"})
+
+	err := cmd.Execute()
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) {
+		t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
+	}
+	for _, want := range []string{"--chat-id", `--params '{"chat_id": "<value>"}'`, "lark-cli schema im.chat.members.create"} {
+		if !strings.Contains(ve.Hint, want) {
+			t.Errorf("hint %q should contain %q", ve.Hint, want)
+		}
+	}
+}
+
+// A params-only required field (kebab name claimed by the standard --format
+// flag) has no typed flag to offer: the hint must give only the --params form,
+// never steer the reader to the colliding flag.
+func TestServiceMethod_MissingRequired_ParamsOnlyHintSkipsFlag(t *testing.T) {
+	method := meta.FromMap(map[string]interface{}{
+		"path":       "messages",
+		"httpMethod": "GET",
+		"parameters": map[string]interface{}{
+			"format": map[string]interface{}{"type": "string", "location": "query", "required": true},
+		},
+	})
+	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
+	cmd := NewCmdServiceMethod(f, imSpec(), method, "list", "messages", nil)
+	cmd.SetArgs([]string{"--dry-run"})
+
+	err := cmd.Execute()
+	var ve *errs.ValidationError
+	if !errors.As(err, &ve) {
+		t.Fatalf("expected *errs.ValidationError, got %T: %v", err, err)
+	}
+	if !strings.Contains(ve.Hint, `--params '{"format": "<value>"}'`) {
+		t.Errorf("hint %q should carry the --params form", ve.Hint)
+	}
+	if strings.Contains(ve.Hint, "set --format") {
+		t.Errorf("hint %q must not steer to the colliding --format flag", ve.Hint)
+	}
+}

cmd/service/paramhelp.go

@@ -0,0 +1,162 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Help rendering for generated param flags. fieldFacts is the single list of
+// agent-relevant facts a param exposes; every help surface (the typed flag's
+// usage line, the params-only --params addendum) renders that one list, so the
+// surfaces cannot drift over which facts exist. Values come from the
+// meta.Field accessors, so nothing here depends on internal/schema.
+
+package service
+
+import (
+	"fmt"
+	"regexp"
+	"strconv"
+	"strings"
+
+	"github.com/larksuite/cli/internal/meta"
+	"github.com/larksuite/cli/internal/util"
+)
+
+// fieldFacts returns a param field's facts in display order, each as a compact
+// one-line clause: the sanitized description, the allowed enum values (with
+// meanings), the min/max constraint, and the API default. This is the ONE
+// place that decides what a param's help says — add a fact here (e.g. a future
+// deprecation marker) and every surface shows it. Unabridged prose and
+// per-option detail stay in `lark-cli schema`.
+func fieldFacts(f meta.Field) []string {
+	var facts []string
+	if d := sanitizeFieldDesc(f.Description); d != "" {
+		facts = append(facts, d)
+	}
+	if opts := f.EnumOptions(); len(opts) > 0 {
+		facts = append(facts, "enum: "+formatEnumInline(opts))
+	}
+	if b := formatBoundsInline(f); b != "" {
+		facts = append(facts, b)
+	}
+	if s := literalStr(f.CoercedDefault()); s != "" {
+		facts = append(facts, "API default: "+s)
+	}
+	return facts
+}
+
+// paramFlagUsage renders the typed param flag's help line:
+//
+//	<param_name>, required|optional[. <fact>]...
+//
+// It leads with the canonical underscore param name (the key this flag
+// overrides in --params) and required/optional, then joins the field's facts
+// inline.
+func paramFlagUsage(f meta.Field) string {
+	req := "optional"
+	if f.Required {
+		req = "required"
+	}
+	parts := append([]string{fmt.Sprintf("%s, %s", f.Name, req)}, fieldFacts(f)...)
+	return strings.Join(parts, ". ") + "."
+}
+
+// paramExample picks a concrete sample for a params-only field's --help snippet:
+// its first allowed enum value, else its example, else a placeholder.
+func paramExample(f meta.Field) string {
+	if vals := enumStrings(f.EnumValues()); len(vals) > 0 {
+		return fmt.Sprintf("%q", vals[0])
+	}
+	if s := literalStr(f.CoercedExample()); s != "" {
+		return fmt.Sprintf("%q", s)
+	}
+	return `"<value>"`
+}
+
+var markdownLinkRe = regexp.MustCompile(`\[([^\]]*)\]\([^)]*\)`)
+
+// inlineClause compresses metadata prose into one help clause: markdown links
+// keep their text, the clause cuts at the first rune in stops, whitespace
+// collapses, trailing punctuation goes — sentence enders (the clause join adds
+// its own) and connectors a cut can strand, like a colon introducing a list the
+// newline cut dropped — and the result caps at max runes. The two policies
+// below differ only in where they cut and how much they keep.
+func inlineClause(s, stops string, max int) string {
+	if s == "" {
+		return ""
+	}
+	s = markdownLinkRe.ReplaceAllString(s, "$1")
+	// Backquotes must go: pflag's UnquoteUsage treats a backquoted word in a
+	// flag's usage string as the flag's metavar, so a description like wiki
+	// space_id's "可替换为`my_library`" would render the flag as
+	// "--space-id my_library" instead of "--space-id string".
+	s = strings.ReplaceAll(s, "`", "")
+	if i := strings.IndexAny(s, stops); i >= 0 {
+		s = s[:i]
+	}
+	s = strings.Join(strings.Fields(s), " ")
+	s = strings.TrimRight(s, "。.：:，,、")
+	return util.TruncateStrWithEllipsis(s, max)
+}
+
+// sanitizeOptionDesc is the enum-option policy: many values share one line, so
+// keep only the first clause (cut at 。 too) and stay ultra-compact.
+func sanitizeOptionDesc(s string) string { return inlineClause(s, "。；;\n\r", 40) }
+
+// sanitizeFieldDesc is the field-description policy: one line per field, so
+// keep full sentences and cut only at note separators (meta_data appends
+// bullet notes after ;/；) — the later sentence often carries the key
+// affordance, e.g. user_mailbox_id's `可以输入"me"`.
+func sanitizeFieldDesc(s string) string { return inlineClause(s, "；;\n\r", 60) }
+
+// formatEnumInline renders allowed values for the help line: "v=meaning" when
+// the value carries a (sanitized, truncated) description — so opaque numeric
+// enums like succeed_type read as "0=…|1=…|2=…" — else just "v". Full meanings
+// live in the envelope's enumDescriptions / `lark-cli schema`.
+func formatEnumInline(opts []meta.EnumOption) string {
+	items := make([]string, len(opts))
+	for i, o := range opts {
+		if d := sanitizeOptionDesc(o.Description); d != "" {
+			items[i] = fmt.Sprintf("%v=%s", o.Value, d)
+		} else {
+			items[i] = fmt.Sprintf("%v", o.Value)
+		}
+	}
+	return strings.Join(items, "|")
+}
+
+// formatBoundsInline renders the field's min/max constraint ("min: 1, max:
+// 100", or the single declared side), or "" when the field declares neither.
+// The vocabulary matches the envelope's minimum/maximum, so help and `lark-cli
+// schema` state the same constraint.
+func formatBoundsInline(f meta.Field) string {
+	min, max := f.MinBound(), f.MaxBound()
+	switch {
+	case min != nil && max != nil:
+		return fmt.Sprintf("min: %s, max: %s", formatBound(*min), formatBound(*max))
+	case min != nil:
+		return "min: " + formatBound(*min)
+	case max != nil:
+		return "max: " + formatBound(*max)
+	}
+	return ""
+}
+
+// formatBound renders a bound without a float artifact (100 not 100.000000).
+func formatBound(v float64) string {
+	return strconv.FormatFloat(v, 'f', -1, 64)
+}
+
+// literalStr renders a coerced literal (default/example) for flag help,
+// returning "" for a nil or empty value so the caller can omit the clause.
+func literalStr(v interface{}) string {
+	if v == nil {
+		return ""
+	}
+	return fmt.Sprintf("%v", v)
+}
+
+func enumStrings(enum []interface{}) []string {
+	out := make([]string, 0, len(enum))
+	for _, e := range enum {
+		out = append(out, fmt.Sprintf("%v", e))
+	}
+	return out
+}

cmd/service/sanitize_test.go

@@ -0,0 +1,61 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package service
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestSanitizeOptionDesc(t *testing.T) {
+	cases := map[string]string{
+		"":                         "",
+		"以 open_id 标识用户":           "以 open_id 标识用户",
+		"中文。English second clause": "中文",         // first clause only (。)
+		"head；tail":                "head",       // first clause (；)
+		"line one\nline two":       "line one",   // first clause (newline)
+		"  spaced   out  ":         "spaced out", // whitespace collapsed
+		"see [飞书后台](https://x/admin) 详情": "see 飞书后台 详情", // markdown link -> text, url dropped
+	}
+	for in, want := range cases {
+		if got := sanitizeOptionDesc(in); got != want {
+			t.Errorf("sanitizeOptionDesc(%q) = %q, want %q", in, got, want)
+		}
+	}
+
+	// Truncation: a long single clause is cut to 40 runes with an ellipsis,
+	// rune-safe (no split mid-character).
+	long := strings.Repeat("文", 60)
+	got := sanitizeOptionDesc(long)
+	if r := []rune(got); len(r) != 40 || !strings.HasSuffix(got, "...") {
+		t.Errorf("truncation = %q (%d runes), want 40 runes ending in ...", got, len(r))
+	}
+}
+
+func TestSanitizeFieldDesc_TrimsDanglingPunctuation(t *testing.T) {
+	// A clause cut can strand a connector (e.g. a colon introducing a list the
+	// newline cut drops, as in im.reactions.list's message_id); the help line
+	// joiner then renders "…获取方式：." — so dangling punctuation must go too.
+	cases := map[string]string{
+		"待查询的消息ID。ID 获取方式：\n- 调用接口获取": "待查询的消息ID。ID 获取方式",
+		"see the list below:\nitem":   "see the list below",
+		"逗号结尾，\n下一行":                  "逗号结尾",
+	}
+	for in, want := range cases {
+		if got := sanitizeFieldDesc(in); got != want {
+			t.Errorf("sanitizeFieldDesc(%q) = %q, want %q", in, got, want)
+		}
+	}
+}
+
+func TestSanitizeFieldDesc_StripsBackquotes(t *testing.T) {
+	// pflag's UnquoteUsage takes a backquoted word in a flag's usage string as
+	// the flag's metavar: wiki space_id's description rendered the flag as
+	// "--space-id my_library" instead of "--space-id string".
+	in := "[知识空间id](https://x/wiki)，如果查询我的文档库可替换为`my_library`"
+	want := "知识空间id，如果查询我的文档库可替换为my_library"
+	if got := sanitizeFieldDesc(in); got != want {
+		t.Errorf("sanitizeFieldDesc(%q) = %q, want %q", in, got, want)
+	}
+}

cmd/service/service.go

@@ -10,18 +10,20 @@ import (
 	"strings"
 
 	"github.com/larksuite/cli/errs"
+	"github.com/larksuite/cli/internal/apicatalog"
 	"github.com/larksuite/cli/internal/auth"
 	"github.com/larksuite/cli/internal/client"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/credential"
 	"github.com/larksuite/cli/internal/errclass"
+	"github.com/larksuite/cli/internal/meta"
 	"github.com/larksuite/cli/internal/output"
 	"github.com/larksuite/cli/internal/registry"
-	"github.com/larksuite/cli/internal/util"
 	"github.com/larksuite/cli/internal/validate"
 	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
 	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
 )
 
 // RegisterServiceCommands registers all service commands from from_meta specs.
@@ -30,85 +32,74 @@ func RegisterServiceCommands(parent *cobra.Command, f *cmdutil.Factory) {
 }
 
 func RegisterServiceCommandsWithContext(ctx context.Context, parent *cobra.Command, f *cmdutil.Factory) {
-	for _, project := range registry.ListFromMetaProjects() {
-		spec := registry.LoadFromMeta(project)
-		if spec == nil {
+	// Drive the service list from the same navigation catalog the method walk
+	// uses — RuntimeCatalog().Services() is the deterministic, sorted view of the
+	// merged metadata — so registration is catalog-sourced end to end. Kept as a
+	// per-service loop rather than a flat WalkMethods(nil) drive precisely so a
+	// service with no methods still gets its bare command (WalkMethods yields one
+	// ref per method, so empty services would vanish).
+	for _, svc := range registry.RuntimeCatalog().Services() {
+		if svc.Name == "" || svc.ServicePath == "" {
 			continue
 		}
-		specName := registry.GetStrFromMap(spec, "name")
-		servicePath := registry.GetStrFromMap(spec, "servicePath")
-		if specName == "" || servicePath == "" {
-			continue
-		}
-		resources, _ := spec["resources"].(map[string]interface{})
-		if resources == nil {
-			continue
-		}
-		registerServiceWithContext(ctx, parent, spec, resources, f)
+		registerServiceWithContext(ctx, parent, svc, f)
 	}
 }
 
-func registerService(parent *cobra.Command, spec map[string]interface{}, resources map[string]interface{}, f *cmdutil.Factory) {
-	registerServiceWithContext(context.Background(), parent, spec, resources, f)
+func registerService(parent *cobra.Command, svc meta.Service, f *cmdutil.Factory) {
+	registerServiceWithContext(context.Background(), parent, svc, f)
 }
 
-func registerServiceWithContext(ctx context.Context, parent *cobra.Command, spec map[string]interface{}, resources map[string]interface{}, f *cmdutil.Factory) {
-	specName := registry.GetStrFromMap(spec, "name")
-	specDesc := registry.GetServiceDescription(specName, "en")
-	if specDesc == "" {
-		specDesc = registry.GetStrFromMap(spec, "description")
-	}
+func registerServiceWithContext(ctx context.Context, parent *cobra.Command, svc meta.Service, f *cmdutil.Factory) {
+	svcCmd := ensureChildCommand(parent, svc.Name, serviceShort(svc))
 
-	// Find existing service command or create one
-	var svc *cobra.Command
+	// Build the service's subtree from the catalog's method walk
+	// (apicatalog.ServiceMethods recurses nested resources), so the command tree
+	// is sourced from the same navigation Module as schema/scope rather than a
+	// hand-rolled resource/method walk. Each ref's ResourcePath becomes the
+	// resource-command chain — one level for a flat dotted resource like
+	// "chat.members", deeper for genuinely nested resources. A service with no
+	// methods keeps its bare command (svcCmd is created above regardless).
+	for _, ref := range apicatalog.ServiceMethods(svc, nil) {
+		resCmd := svcCmd
+		for _, seg := range ref.ResourcePath {
+			resCmd = ensureChildCommand(resCmd, seg, seg+" operations")
+		}
+		resCmd.AddCommand(buildMethodCommand(ctx, f, newMethodCommandSpec(ref), nil, parent.PersistentFlags()))
+	}
+}
+
+// serviceShort is the service command's help summary: the localized description
+// from the registry, falling back to the metadata's own description.
+func serviceShort(svc meta.Service) string {
+	if d := registry.GetServiceDescription(svc.Name, "en"); d != "" {
+		return d
+	}
+	return svc.Description
+}
+
+// ensureChildCommand returns the child of parent named name, creating it (with
+// short) when absent — so re-registration merges into an existing command tree
+// instead of duplicating a level.
+func ensureChildCommand(parent *cobra.Command, name, short string) *cobra.Command {
 	for _, c := range parent.Commands() {
-		if c.Name() == specName {
-			svc = c
-			break
+		if c.Name() == name {
+			return c
 		}
 	}
-	if svc == nil {
-		svc = &cobra.Command{
-			Use:   specName,
-			Short: specDesc,
-		}
-		parent.AddCommand(svc)
-	}
-
-	for resName, resource := range resources {
-		resMap, _ := resource.(map[string]interface{})
-		if resMap == nil {
-			continue
-		}
-		registerResourceWithContext(ctx, svc, spec, resName, resMap, f)
-	}
-}
-
-func registerResourceWithContext(ctx context.Context, parent *cobra.Command, spec map[string]interface{}, name string, resource map[string]interface{}, f *cmdutil.Factory) {
-	res := &cobra.Command{
-		Use:   name,
-		Short: name + " operations",
-	}
-	parent.AddCommand(res)
-
-	methods, _ := resource["methods"].(map[string]interface{})
-	for methodName, method := range methods {
-		methodMap, _ := method.(map[string]interface{})
-		if methodMap == nil {
-			continue
-		}
-		registerMethodWithContext(ctx, res, spec, methodMap, methodName, name, f)
-	}
+	cmd := &cobra.Command{Use: name, Short: short}
+	parent.AddCommand(cmd)
+	return cmd
 }
 
 // ServiceMethodOptions holds all inputs for a dynamically registered service method command.
 type ServiceMethodOptions struct {
-	Factory    *cmdutil.Factory
-	Cmd        *cobra.Command
-	Ctx        context.Context
-	Spec       map[string]interface{}
-	Method     map[string]interface{}
-	SchemaPath string
+	Factory     *cmdutil.Factory
+	Cmd         *cobra.Command
+	Ctx         context.Context
+	ServicePath string
+	Method      meta.Method
+	SchemaPath  string
 
 	// Flags
 	Params     string
@@ -123,41 +114,113 @@ type ServiceMethodOptions struct {
 	DryRun     bool
 	File       string   // --file flag value
 	FileFields []string // auto-detected file field names from metadata
+
+	// binder owns the generated typed param flags — registration and the
+	// --params overlay — replacing the raw paramFlags side-channel.
+	binder *paramFlagBinder
 }
 
-// detectFileFields delegates to the shared cmdutil.DetectFileFields helper.
-func detectFileFields(method map[string]interface{}) []string {
-	return cmdutil.DetectFileFields(method)
-}
-
-func registerMethodWithContext(ctx context.Context, parent *cobra.Command, spec map[string]interface{}, method map[string]interface{}, name string, resName string, f *cmdutil.Factory) {
-	parent.AddCommand(NewCmdServiceMethodWithContext(ctx, f, spec, method, name, resName, nil))
+// detectFileFields returns the request-body file-upload field names.
+func detectFileFields(m meta.Method) []string {
+	files := m.Files()
+	if len(files) == 0 {
+		return nil
+	}
+	names := make([]string, len(files))
+	for i, f := range files {
+		names[i] = f.Name
+	}
+	return names
 }
 
 // NewCmdServiceMethod creates a command for a dynamically registered service method.
-func NewCmdServiceMethod(f *cmdutil.Factory, spec, method map[string]interface{}, name, resName string, runF func(*ServiceMethodOptions) error) *cobra.Command {
-	return NewCmdServiceMethodWithContext(context.Background(), f, spec, method, name, resName, runF)
+func NewCmdServiceMethod(f *cmdutil.Factory, svc meta.Service, m meta.Method, name, resName string, runF func(*ServiceMethodOptions) error) *cobra.Command {
+	return NewCmdServiceMethodWithContext(context.Background(), f, svc, m, name, resName, runF)
 }
 
-func NewCmdServiceMethodWithContext(ctx context.Context, f *cmdutil.Factory, spec, method map[string]interface{}, name, resName string, runF func(*ServiceMethodOptions) error) *cobra.Command {
-	desc := registry.GetStrFromMap(method, "description")
-	httpMethod := registry.GetStrFromMap(method, "httpMethod")
-	risk := registry.GetStrFromMap(method, "risk")
-	specName := registry.GetStrFromMap(spec, "name")
-	schemaPath := fmt.Sprintf("%s.%s.%s", specName, resName, name)
+// NewCmdServiceMethodWithContext builds the command for one service method from
+// its (service, resource, method) coordinates, deriving the methodCommandSpec
+// via an apicatalog.MethodRef so direct callers and the catalog-driven
+// registration assemble the command identically.
+func NewCmdServiceMethodWithContext(ctx context.Context, f *cmdutil.Factory, svc meta.Service, m meta.Method, name, resName string, runF func(*ServiceMethodOptions) error) *cobra.Command {
+	m.Name = name
+	ref := apicatalog.MethodRef{Service: svc, ResourcePath: []string{resName}, Method: m}
+	// No root in scope here; persistent-flag collisions don't apply to a
+	// standalone command, and local/standard-flag collisions are still caught.
+	return buildMethodCommand(ctx, f, newMethodCommandSpec(ref), runF, nil)
+}
 
+// methodCommandSpec is the static description of one generated service method
+// command, read off an apicatalog.MethodRef — the single place command
+// construction gets the method's facts (schema path, HTTP base path, risk,
+// identities, params, file fields, request-body support), so the cobra command
+// is assembled from a typed spec rather than recomputing paths/flags inline.
+type methodCommandSpec struct {
+	method      meta.Method
+	schemaPath  string       // "service.resource.method", for the --help hint
+	servicePath string       // service HTTP base path
+	risk        string       // RiskRead | RiskWrite | RiskHighRiskWrite
+	restricts   bool         // method declares accessTokens (identity-restricted)
+	identities  []string     // permitted --as values; empty when unrestricted
+	params      []meta.Field // path/query params -> typed flags
+	fileFields  []string     // request-body file-upload field names
+	// acceptsBody is whether the HTTP method allows a request body at all (so
+	// --data is offered as a raw escape hatch). declaresBody is whether the
+	// metadata documents body fields (data or file). They differ for e.g. a POST
+	// with no documented requestBody: --data still works, but help must not imply
+	// the API declares a body.
+	acceptsBody  bool
+	declaresBody bool
+	affordance   string // rendered hand-authored usage guidance (when-to-use, examples); "" if none
+}
+
+func newMethodCommandSpec(ref apicatalog.MethodRef) methodCommandSpec {
+	m := ref.Method
+	return methodCommandSpec{
+		method:       m,
+		schemaPath:   ref.SchemaPath(),
+		servicePath:  ref.Service.ServicePath,
+		risk:         m.Risk,
+		restricts:    m.RestrictsIdentity(),
+		identities:   m.Identities(),
+		params:       m.Params(),
+		fileFields:   detectFileFields(m),
+		acceptsBody:  methodTakesBody(m.HTTPMethod),
+		declaresBody: len(m.Data()) > 0 || len(m.Files()) > 0,
+		affordance:   renderAffordance(m),
+	}
+}
+
+// methodTakesBody reports whether the HTTP method allows a request body, i.e.
+// whether --data applies (as a raw escape hatch even when no body is declared).
+func methodTakesBody(httpMethod string) bool {
+	switch httpMethod {
+	case "POST", "PUT", "PATCH", "DELETE":
+		return true
+	}
+	return false
+}
+
+// buildMethodCommand assembles the cobra command for a service method from its
+// static spec: the standard flags, the conditional --data/--file/--yes flags,
+// the generated typed param flags (via paramFlagBinder), and the risk/identity
+// policy annotations.
+func buildMethodCommand(ctx context.Context, f *cmdutil.Factory, spec methodCommandSpec, runF func(*ServiceMethodOptions) error, reserved *pflag.FlagSet) *cobra.Command {
+	m := spec.method
 	opts := &ServiceMethodOptions{
-		Factory:    f,
-		Spec:       spec,
-		Method:     method,
-		SchemaPath: schemaPath,
+		Factory:     f,
+		ServicePath: spec.servicePath,
+		Method:      m,
+		SchemaPath:  spec.schemaPath,
+		FileFields:  spec.fileFields,
 	}
 	var asStr string
 
 	cmd := &cobra.Command{
-		Use:   name,
-		Short: desc,
-		Long:  fmt.Sprintf("%s\n\nView parameter definitions before calling:\n  lark-cli schema %s", desc, schemaPath),
+		Use:   m.Name,
+		Short: m.Description,
+		// Long is assembled below, once the binder knows which params got no
+		// typed flag.
 		RunE: func(cmd *cobra.Command, args []string) error {
 			opts.Cmd = cmd
 			opts.Ctx = cmd.Context()
@@ -169,10 +232,15 @@ func NewCmdServiceMethodWithContext(ctx context.Context, f *cmdutil.Factory, spe
 		},
 	}
 
-	cmd.Flags().StringVar(&opts.Params, "params", "", "URL/query parameters JSON (supports - for stdin, @file for file input)")
-	switch httpMethod {
-	case "POST", "PUT", "PATCH", "DELETE":
-		cmd.Flags().StringVar(&opts.Data, "data", "", "request body JSON (supports - for stdin, @file for file input)")
+	cmd.Flags().StringVar(&opts.Params, "params", "", "Raw URL/query params JSON. Supports - and @file.")
+	if spec.acceptsBody {
+		dataUsage := "JSON request body. Supports - and @file."
+		if !spec.declaresBody {
+			// POST/etc. with no documented body fields: --data is a raw escape
+			// hatch, not a declared body — say so rather than imply structure.
+			dataUsage = "Raw JSON request body (no documented fields; see schema). Supports - and @file."
+		}
+		cmd.Flags().StringVar(&opts.Data, "data", "", dataUsage)
 	}
 	cmdutil.AddAPIIdentityFlag(ctx, cmd, f, &asStr)
 	cmd.Flags().StringVarP(&opts.Output, "output", "o", "", "output file path for binary responses")
@@ -183,27 +251,61 @@ func NewCmdServiceMethodWithContext(ctx context.Context, f *cmdutil.Factory, spe
 	cmd.Flags().Bool("json", false, "shorthand for --format json")
 	cmd.Flags().StringVarP(&opts.JqExpr, "jq", "q", "", "jq expression to filter JSON output")
 	cmd.Flags().BoolVar(&opts.DryRun, "dry-run", false, "print request without executing")
-	if risk == "high-risk-write" {
+	if spec.risk == cmdutil.RiskHighRiskWrite {
 		cmd.Flags().Bool("yes", false, "confirm high-risk operation")
 	}
-
-	// Conditionally register --file for methods with file-type fields.
-	fileFields := detectFileFields(method)
-	opts.FileFields = fileFields
-	if len(fileFields) > 0 {
-		switch httpMethod {
-		case "POST", "PUT", "PATCH", "DELETE":
-			cmd.Flags().StringVar(&opts.File, "file", "", "file to upload ([field=]path, supports - for stdin)")
-		}
+	// --file only for body methods that actually declare file-type fields.
+	if len(spec.fileFields) > 0 && spec.acceptsBody {
+		cmd.Flags().StringVar(&opts.File, "file", "", "File upload [field=]path. Supports - and stdin.")
 	}
 	cmdutil.RegisterFlagCompletion(cmd, "format", func(_ *cobra.Command, _ []string, _ string) ([]string, cobra.ShellCompDirective) {
 		return []string{"json", "ndjson", "table", "csv"}, cobra.ShellCompDirectiveNoFileComp
 	})
 
-	cmdutil.SetTips(cmd, registry.GetStrSliceFromMap(method, "tips"))
-	cmdutil.SetRisk(cmd, risk)
-	if tokens, ok := method["accessTokens"].([]interface{}); ok && len(tokens) > 0 {
-		cmdutil.SetSupportedIdentities(cmd, cmdutil.AccessTokensToIdentities(tokens))
+	// Registered last so the collision guard sees the standard flags above.
+	opts.binder = newParamFlagBinder(cmd, spec.params, reserved)
+	// Single composition point for Long: description, affordance, schema
+	// pointer, and the binder's params-only addendum (params whose flag name is
+	// taken, reachable via --params only).
+	cmd.Long = methodLong(m.Description, spec.affordance, spec.schemaPath, opts.binder.paramsOnlyHelp())
+
+	// Group flags for the grouped --help renderer (typed param flags are grouped
+	// as API Parameters by the binder). tagFlagGroup is a no-op for flags not
+	// registered above (e.g. --data/--file/--yes only exist for some methods).
+	// --data sits under Request Body only when the metadata documents body
+	// fields; otherwise it's a raw escape hatch, grouped with --params so help
+	// doesn't imply a declared body the API doesn't have.
+	if fl := cmd.Flags().Lookup("data"); fl != nil {
+		if spec.declaresBody {
+			annotate(fl, flagGroupAnnotation, []string{groupBody})
+		} else {
+			annotate(fl, flagGroupAnnotation, []string{groupRaw})
+		}
+	}
+	tagFlagGroup(cmd.Flags(), "file", groupBody)
+	if fl := cmd.Flags().Lookup("params"); fl != nil {
+		annotate(fl, flagGroupAnnotation, []string{groupRaw})
+		// State the precedence rule where the agent reads it: --params is the
+		// base, typed flags override. Only meaningful when typed flags exist.
+		if len(spec.params) > 0 {
+			annotate(fl, flagNoteAnnotation, []string{
+				"Typed API parameter flags above are preferred.",
+				"If both are set, typed flags override matching keys in --params.",
+			})
+		}
+	}
+	for _, name := range []string{"as", "dry-run", "page-all", "page-limit", "page-delay", "yes"} {
+		tagFlagGroup(cmd.Flags(), name, groupExecution)
+	}
+	for _, name := range []string{"output", "format", "jq"} {
+		tagFlagGroup(cmd.Flags(), name, groupOutput)
+	}
+	applyGroupedUsage(cmd)
+
+	cmdutil.SetTips(cmd, m.Tips)
+	cmdutil.SetRisk(cmd, spec.risk)
+	if spec.restricts {
+		cmdutil.SetSupportedIdentities(cmd, spec.identities)
 	}
 
 	return cmd
@@ -218,8 +320,8 @@ func serviceMethodRun(opts *ServiceMethodOptions) error {
 	}
 
 	// Check if this API method supports the resolved identity.
-	if tokens, ok := opts.Method["accessTokens"].([]interface{}); ok && len(tokens) > 0 {
-		if err := f.CheckIdentity(opts.As, cmdutil.AccessTokensToIdentities(tokens)); err != nil {
+	if opts.Method.RestrictsIdentity() {
+		if err := f.CheckIdentity(opts.As, opts.Method.Identities()); err != nil {
 			return err
 		}
 	}
@@ -235,12 +337,10 @@ func serviceMethodRun(opts *ServiceMethodOptions) error {
 	if err != nil {
 		return err
 	}
-	// Identity info is now included in the JSON envelope; skip stderr printing.
-	// cmdutil.PrintIdentity(f.IOStreams.ErrOut, opts.As, config, f.IdentityAutoDetected)
+	// Identity is not printed to stderr here: it is part of the JSON envelope.
 
-	scopes, _ := opts.Method["scopes"].([]interface{})
 	if !opts.As.IsBot() {
-		if err := checkServiceScopes(opts.Ctx, f.Credential, opts.As, config, opts.Method, scopes); err != nil {
+		if err := checkServiceScopes(opts.Ctx, f.Credential, opts.As, config, opts.Method); err != nil {
 			return err
 		}
 	}
@@ -257,7 +357,7 @@ func serviceMethodRun(opts *ServiceMethodOptions) error {
 		return serviceDryRun(f, request, config, opts.Format)
 	}
 
-	if registry.GetStrFromMap(opts.Method, "risk") == "high-risk-write" {
+	if opts.Method.Risk == cmdutil.RiskHighRiskWrite {
 		if yes, _ := opts.Cmd.Flags().GetBool("yes"); !yes {
 			return cmdutil.RequireConfirmation(opts.SchemaPath)
 		}
@@ -302,7 +402,7 @@ func serviceMethodRun(opts *ServiceMethodOptions) error {
 }
 
 // checkServiceScopes pre-checks user scopes before making the API call.
-func checkServiceScopes(ctx context.Context, cred *credential.CredentialProvider, identity core.Identity, config *core.CliConfig, method map[string]interface{}, scopes []interface{}) error {
+func checkServiceScopes(ctx context.Context, cred *credential.CredentialProvider, identity core.Identity, config *core.CliConfig, method meta.Method) error {
 	if ctx.Err() != nil {
 		return ctx.Err()
 	}
@@ -311,23 +411,15 @@ func checkServiceScopes(ctx context.Context, cred *credential.CredentialProvider
 		return nil //nolint:nilerr // skip scope check when token resolution fails or has no scopes
 	}
 
-	requiredScopes, hasRequired := method["requiredScopes"].([]interface{})
-
-	if hasRequired && len(requiredScopes) > 0 {
+	if len(method.RequiredScopes) > 0 {
 		// Strict: ALL requiredScopes must be present
-		required := make([]string, 0, len(requiredScopes))
-		for _, s := range requiredScopes {
-			if str, ok := s.(string); ok {
-				required = append(required, str)
-			}
-		}
-		if missing := auth.MissingScopes(result.Scopes, required); len(missing) > 0 {
+		if missing := auth.MissingScopes(result.Scopes, method.RequiredScopes); len(missing) > 0 {
 			return newPreflightMissingScopeError(string(config.Brand), config.AppID, string(identity), missing)
 		}
 		return nil
 	}
 
-	if len(scopes) == 0 {
+	if len(method.Scopes) == 0 {
 		return nil
 	}
 
@@ -336,12 +428,12 @@ func checkServiceScopes(ctx context.Context, cred *credential.CredentialProvider
 	for _, s := range strings.Fields(result.Scopes) {
 		grantedSet[s] = true
 	}
-	for _, s := range scopes {
-		if str, ok := s.(string); ok && grantedSet[str] {
+	for _, s := range method.Scopes {
+		if grantedSet[s] {
 			return nil
 		}
 	}
-	recommended := registry.SelectRecommendedScope(scopes, "user")
+	recommended := registry.SelectRecommendedScopeFromStrings(method.Scopes, "user")
 	return newPreflightMissingScopeError(string(config.Brand), config.AppID, string(identity), []string{recommended})
 }
 
@@ -362,14 +454,44 @@ func newPreflightMissingScopeError(brand, appID, identity string, missing []stri
 		WithIdentity(identity)
 }
 
+// unusableParamValue reports whether a provided path/query parameter value
+// cannot form a usable request value: nil or an empty string. A key's presence
+// in params is the intent signal — a typed flag is overlaid only when
+// explicitly Changed, and a --params JSON key is deliberately written — so
+// false and 0 are real values and must not be conflated with "unset"
+// (reflect.IsZero would drop an explicit --with-deleted=false or --foo 0).
+// Only nil/"" stay treated as missing: that keeps the friendly pre-flight
+// error when a required param is fed an empty placeholder, and never emits a
+// declared param as an empty path segment or query value. Undeclared keys are
+// not judged by this rule — they pass through verbatim as the raw escape hatch.
+func unusableParamValue(v interface{}) bool {
+	if v == nil {
+		return true
+	}
+	s, ok := v.(string)
+	return ok && s == ""
+}
+
+// missingParamHint is the recovery hint for a missing required parameter. It
+// names both input paths — the typed flag when the binder registered one, and
+// the --params fallback — plus the schema pointer. A params-only field gets
+// only the --params form: a flag with its kebab name exists but belongs to
+// something else (e.g. the output --format), and the hint must not steer
+// there. Asking the binder, not cmd.Flags(), is what tells those apart.
+func missingParamHint(opts *ServiceMethodOptions, f meta.Field) string {
+	paramsForm := fmt.Sprintf("--params '{%q: \"<value>\"}'", f.Name)
+	if opts.binder.hasTypedFlag(f.Name) {
+		return fmt.Sprintf("set --%s <value> (or %s); see: lark-cli schema %s", f.FlagName(), paramsForm, opts.SchemaPath)
+	}
+	return fmt.Sprintf("set %s; see: lark-cli schema %s", paramsForm, opts.SchemaPath)
+}
+
 // buildServiceRequest parses flags, builds the URL with path/query params, and returns a RawApiRequest.
 // When dryRun is true and a file is provided, file reading is skipped and
 // FileUploadMeta is returned instead so the caller can render dry-run output.
 func buildServiceRequest(opts *ServiceMethodOptions) (client.RawApiRequest, *cmdutil.FileUploadMeta, error) {
-	spec := opts.Spec
 	method := opts.Method
-	schemaPath := opts.SchemaPath
-	httpMethod := registry.GetStrFromMap(method, "httpMethod")
+	httpMethod := method.HTTPMethod
 
 	// stdin is an io.Reader consumed at most once. Only one of --params/--data
 	// may use "-" (stdin); the conflict check below prevents silent data loss.
@@ -387,53 +509,55 @@ func buildServiceRequest(opts *ServiceMethodOptions) (client.RawApiRequest, *cmd
 	if err != nil {
 		return client.RawApiRequest{}, nil, err
 	}
+	opts.binder.overlay(opts.Cmd, params)
 
-	url := registry.GetStrFromMap(spec, "servicePath") + "/" + registry.GetStrFromMap(method, "path")
+	url := opts.ServicePath + "/" + method.Path
 
-	parameters, _ := method["parameters"].(map[string]interface{})
-	for name, param := range parameters {
-		p, _ := param.(map[string]interface{})
-		if registry.GetStrFromMap(p, "location") != "path" {
+	specs := method.Params()
+	for _, s := range specs {
+		if s.Location != "path" {
 			continue
 		}
-		val, ok := params[name]
-		if !ok || util.IsEmptyValue(val) {
+		val, ok := params[s.Name]
+		if !ok || unusableParamValue(val) {
 			return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
-				"missing required path parameter: %s", name).
-				WithHint("lark-cli schema %s", schemaPath).
-				WithParam(name)
+				"missing required path parameter: %s", s.Name).
+				WithHint("%s", missingParamHint(opts, s)).
+				WithParam(s.Name)
 		}
 		valStr := fmt.Sprintf("%v", val)
-		if err := validate.ResourceName(valStr, name); err != nil {
-			return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", err).WithParam(name).WithCause(err)
+		if err := validate.ResourceName(valStr, s.Name); err != nil {
+			return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument, "%s", err).WithParam(s.Name).WithCause(err)
 		}
-		url = strings.Replace(url, "{"+name+"}", validate.EncodePathSegment(valStr), 1)
-		delete(params, name)
+		url = strings.Replace(url, "{"+s.Name+"}", validate.EncodePathSegment(valStr), 1)
+		delete(params, s.Name)
 	}
 
 	queryParams := map[string]interface{}{}
-	for name, param := range parameters {
-		p, _ := param.(map[string]interface{})
-		if registry.GetStrFromMap(p, "location") != "query" {
+	for _, s := range specs {
+		if s.Location != "query" {
 			continue
 		}
-		value, exists := params[name]
-		required, _ := p["required"].(bool)
-		isPaginationParam := opts.PageAll && (name == "page_token" || name == "page_size")
-		if required && !isPaginationParam && (!exists || util.IsEmptyValue(value)) {
+		value, exists := params[s.Name]
+		isPaginationParam := opts.PageAll && (s.Name == "page_token" || s.Name == "page_size")
+		if s.Required && !isPaginationParam && (!exists || unusableParamValue(value)) {
 			return client.RawApiRequest{}, nil, errs.NewValidationError(errs.SubtypeInvalidArgument,
-				"missing required query parameter: %s", name).
-				WithHint("lark-cli schema %s", schemaPath).
-				WithParam(name)
+				"missing required query parameter: %s", s.Name).
+				WithHint("%s", missingParamHint(opts, s)).
+				WithParam(s.Name)
 		}
-		if exists && !util.IsEmptyValue(value) {
-			queryParams[name] = value
+		if exists && !unusableParamValue(value) {
+			queryParams[s.Name] = value
 		}
+		// This loop owns declared query params: consume the key so the
+		// passthrough below can't resurrect a value the gate dropped (an
+		// unusable "" would otherwise be sent as an empty query value).
+		delete(params, s.Name)
 	}
+	// Whatever remains is undeclared — the raw escape hatch for params the
+	// metadata doesn't (yet) describe; passed through verbatim, no filtering.
 	for name, value := range params {
-		if _, ok := queryParams[name]; !ok {
-			queryParams[name] = value
-		}
+		queryParams[name] = value
 	}
 
 	request := client.RawApiRequest{

cmd/service/service_risk_test.go

@@ -8,13 +8,14 @@ import (
 	"testing"
 
 	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/meta"
 )
 
 // highRiskDeleteMethod mirrors a simple DELETE API with a required path
-// parameter and risk metadata. The returned map is what service registration
-// reads; the test exercises --yes registration and the gate behavior.
-func highRiskDeleteMethod() map[string]interface{} {
-	return map[string]interface{}{
+// parameter and risk metadata. The test exercises --yes registration and the
+// gate behavior.
+func highRiskDeleteMethod() meta.Method {
+	return meta.FromMap(map[string]interface{}{
 		"path":       "files/{file_token}",
 		"httpMethod": "DELETE",
 		"risk":       "high-risk-write",
@@ -23,11 +24,11 @@ func highRiskDeleteMethod() map[string]interface{} {
 				"type": "string", "location": "path", "required": true,
 			},
 		},
-	}
+	})
 }
 
-func writeMethodNoRisk() map[string]interface{} {
-	return map[string]interface{}{
+func writeMethodNoRisk() meta.Method {
+	return meta.FromMap(map[string]interface{}{
 		"path":       "files/{file_token}",
 		"httpMethod": "DELETE",
 		"parameters": map[string]interface{}{
@@ -35,7 +36,7 @@ func writeMethodNoRisk() map[string]interface{} {
 				"type": "string", "location": "path", "required": true,
 			},
 		},
-	}
+	})
 }
 
 func TestServiceMethod_YesFlagRegisteredForHighRisk(t *testing.T) {

cmd/service/service_test.go

@@ -11,6 +11,7 @@ import (
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/httpmock"
+	"github.com/larksuite/cli/internal/meta"
 	"github.com/spf13/cobra"
 )
 
@@ -20,14 +21,14 @@ var testConfig = &core.CliConfig{
 	AppID: "test-app", AppSecret: "test-secret", Brand: core.BrandFeishu,
 }
 
-func driveSpec() map[string]interface{} {
-	return map[string]interface{}{
+func driveSpec() meta.Service {
+	return meta.ServiceFromMap(map[string]interface{}{
 		"name":        "drive",
 		"servicePath": "/open-apis/drive/v1",
-	}
+	})
 }
 
-func driveMethod(httpMethod string, params map[string]interface{}) map[string]interface{} {
+func driveMethod(httpMethod string, params map[string]interface{}) meta.Method {
 	m := map[string]interface{}{
 		"path":       "files/{file_token}/copy",
 		"httpMethod": httpMethod,
@@ -41,7 +42,7 @@ func driveMethod(httpMethod string, params map[string]interface{}) map[string]in
 			},
 		}
 	}
-	return m
+	return meta.FromMap(m)
 }
 
 // ── registerService ──
@@ -49,23 +50,23 @@ func driveMethod(httpMethod string, params map[string]interface{}) map[string]in
 func TestRegisterService(t *testing.T) {
 	parent := &cobra.Command{Use: "root"}
 	f := &cmdutil.Factory{}
-	spec := map[string]interface{}{
+	base := meta.ServiceFromMap(map[string]interface{}{
 		"name":        "base",
 		"description": "Base API",
 		"servicePath": "/open-apis/base/v3",
-	}
-	resources := map[string]interface{}{
-		"tables": map[string]interface{}{
-			"methods": map[string]interface{}{
-				"list": map[string]interface{}{
-					"description": "List tables",
-					"httpMethod":  "GET",
+		"resources": map[string]interface{}{
+			"tables": map[string]interface{}{
+				"methods": map[string]interface{}{
+					"list": map[string]interface{}{
+						"description": "List tables",
+						"httpMethod":  "GET",
+					},
 				},
 			},
 		},
-	}
+	})
 
-	registerService(parent, spec, resources, f)
+	registerService(parent, base, f)
 
 	// service command exists
 	svc, _, err := parent.Find([]string{"base"})
@@ -90,18 +91,18 @@ func TestRegisterService_MergesExistingCommand(t *testing.T) {
 	parent.AddCommand(existing)
 
 	f := &cmdutil.Factory{}
-	spec := map[string]interface{}{
+	svc := meta.ServiceFromMap(map[string]interface{}{
 		"name": "base", "description": "Base API", "servicePath": "/open-apis/base/v3",
-	}
-	resources := map[string]interface{}{
-		"tables": map[string]interface{}{
-			"methods": map[string]interface{}{
-				"list": map[string]interface{}{"description": "List", "httpMethod": "GET"},
+		"resources": map[string]interface{}{
+			"tables": map[string]interface{}{
+				"methods": map[string]interface{}{
+					"list": map[string]interface{}{"description": "List", "httpMethod": "GET"},
+				},
 			},
 		},
-	}
+	})
 
-	registerService(parent, spec, resources, f)
+	registerService(parent, svc, f)
 
 	// Should reuse existing, not duplicate
 	count := 0
@@ -143,7 +144,7 @@ func TestNewCmdServiceMethod_StrictModeHidesAsFlag(t *testing.T) {
 func TestNewCmdServiceMethod_GETHasNoDataFlag(t *testing.T) {
 	f := &cmdutil.Factory{}
 	cmd := NewCmdServiceMethod(f, driveSpec(),
-		map[string]interface{}{"description": "desc", "httpMethod": "GET"}, "list", "files", nil)
+		meta.FromMap(map[string]interface{}{"description": "desc", "httpMethod": "GET"}), "list", "files", nil)
 
 	if cmd.Flags().Lookup("data") != nil {
 		t.Error("GET method should not have --data flag")
@@ -159,7 +160,7 @@ func TestNewCmdServiceMethod_GETHasNoDataFlag(t *testing.T) {
 func TestNewCmdServiceMethod_POSTHasDataFlag(t *testing.T) {
 	f := &cmdutil.Factory{}
 	cmd := NewCmdServiceMethod(f, driveSpec(),
-		map[string]interface{}{"description": "desc", "httpMethod": "POST"}, "create", "files", nil)
+		meta.FromMap(map[string]interface{}{"description": "desc", "httpMethod": "POST"}), "create", "files", nil)
 
 	if cmd.Flags().Lookup("data") == nil {
 		t.Error("POST method should have --data flag")
@@ -171,7 +172,7 @@ func TestNewCmdServiceMethod_RunFCallback(t *testing.T) {
 
 	var captured *ServiceMethodOptions
 	cmd := NewCmdServiceMethod(f, driveSpec(),
-		map[string]interface{}{"description": "desc", "httpMethod": "GET"}, "list", "files",
+		meta.FromMap(map[string]interface{}{"description": "desc", "httpMethod": "GET"}), "list", "files",
 		func(opts *ServiceMethodOptions) error {
 			captured = opts
 			return nil
@@ -268,15 +269,15 @@ func TestServiceMethod_MissingPathParam(t *testing.T) {
 }
 
 func TestServiceMethod_MissingRequiredQueryParam(t *testing.T) {
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{
+	})
+	method := meta.FromMap(map[string]interface{}{
 		"path": "items", "httpMethod": "GET",
 		"parameters": map[string]interface{}{
 			"q": map[string]interface{}{"location": "query", "required": true},
 		},
-	}
+	})
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--params", `{}`, "--dry-run"})
@@ -291,15 +292,15 @@ func TestServiceMethod_MissingRequiredQueryParam(t *testing.T) {
 }
 
 func TestServiceMethod_PaginationParamSkippedWithPageAll(t *testing.T) {
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{
+	})
+	method := meta.FromMap(map[string]interface{}{
 		"path": "items", "httpMethod": "GET",
 		"parameters": map[string]interface{}{
 			"page_size": map[string]interface{}{"location": "query", "required": true},
 		},
-	}
+	})
 	f, stdout, _, _ := cmdutil.TestFactory(t, testConfig)
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--params", `{}`, "--page-all", "--dry-run"})
@@ -315,10 +316,10 @@ func TestServiceMethod_PaginationParamSkippedWithPageAll(t *testing.T) {
 
 func TestServiceMethod_InvalidParamsJSON(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET"}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET"})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--params", "{bad", "--dry-run"})
 
@@ -333,10 +334,10 @@ func TestServiceMethod_InvalidParamsJSON(t *testing.T) {
 
 func TestServiceMethod_InvalidDataJSON(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "POST", "parameters": map[string]interface{}{}}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "POST", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "create", "items", nil)
 	cmd.SetArgs([]string{"--data", "{bad", "--dry-run"})
 
@@ -351,10 +352,10 @@ func TestServiceMethod_InvalidDataJSON(t *testing.T) {
 
 func TestServiceMethod_ParamsAndDataBothStdinConflict(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "POST", "parameters": map[string]interface{}{}}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "POST", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "create", "items", nil)
 	cmd.SetArgs([]string{"--params", "-", "--data", "-", "--dry-run"})
 
@@ -369,10 +370,10 @@ func TestServiceMethod_ParamsAndDataBothStdinConflict(t *testing.T) {
 
 func TestServiceMethod_OutputAndPageAllConflict(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET"}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET"})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--page-all", "--output", "file.bin", "--as", "bot"})
 
@@ -398,8 +399,8 @@ func TestServiceMethod_BotMode_Success(t *testing.T) {
 		},
 	})
 
-	spec := map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}}
+	spec := meta.ServiceFromMap(map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--as", "bot"})
 
@@ -427,8 +428,8 @@ func TestServiceMethod_BotMode_PageAll_JSON(t *testing.T) {
 		},
 	})
 
-	spec := map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}}
+	spec := meta.ServiceFromMap(map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--as", "bot", "--page-all"})
 
@@ -450,8 +451,8 @@ func TestServiceMethod_UnknownFormat_Warning(t *testing.T) {
 		Body: map[string]interface{}{"code": 0, "msg": "ok", "data": map[string]interface{}{}},
 	})
 
-	spec := map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}}
+	spec := meta.ServiceFromMap(map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--as", "bot", "--format", "unknown"})
 
@@ -470,7 +471,7 @@ func TestNewCmdServiceMethod_JqFlag(t *testing.T) {
 
 	var captured *ServiceMethodOptions
 	cmd := NewCmdServiceMethod(f, driveSpec(),
-		map[string]interface{}{"description": "desc", "httpMethod": "GET"}, "list", "files",
+		meta.FromMap(map[string]interface{}{"description": "desc", "httpMethod": "GET"}), "list", "files",
 		func(opts *ServiceMethodOptions) error {
 			captured = opts
 			return nil
@@ -492,7 +493,7 @@ func TestNewCmdServiceMethod_JqShortForm(t *testing.T) {
 
 	var captured *ServiceMethodOptions
 	cmd := NewCmdServiceMethod(f, driveSpec(),
-		map[string]interface{}{"description": "desc", "httpMethod": "GET"}, "list", "files",
+		meta.FromMap(map[string]interface{}{"description": "desc", "httpMethod": "GET"}), "list", "files",
 		func(opts *ServiceMethodOptions) error {
 			captured = opts
 			return nil
@@ -508,10 +509,10 @@ func TestNewCmdServiceMethod_JqShortForm(t *testing.T) {
 
 func TestServiceMethod_JqAndOutputConflict(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET"}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET"})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--jq", ".data", "--output", "file.bin", "--as", "bot"})
 
@@ -542,8 +543,8 @@ func TestServiceMethod_JqFilter_AppliesExpression(t *testing.T) {
 		},
 	})
 
-	spec := map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}}
+	spec := meta.ServiceFromMap(map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--as", "bot", "--jq", ".data.items[].name"})
 
@@ -561,10 +562,10 @@ func TestServiceMethod_JqFilter_AppliesExpression(t *testing.T) {
 
 func TestServiceMethod_JqAndFormatConflict(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET"}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET"})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--jq", ".data", "--format", "ndjson", "--as", "bot"})
 
@@ -579,10 +580,10 @@ func TestServiceMethod_JqAndFormatConflict(t *testing.T) {
 
 func TestServiceMethod_JqInvalidExpression(t *testing.T) {
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	spec := map[string]interface{}{
+	spec := meta.ServiceFromMap(map[string]interface{}{
 		"name": "svc", "servicePath": "/open-apis/svc/v1",
-	}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET"}
+	})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET"})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--jq", "invalid[", "--as", "bot"})
 
@@ -611,8 +612,8 @@ func TestServiceMethod_PageAll_WithJq(t *testing.T) {
 		},
 	})
 
-	spec := map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"}
-	method := map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}}
+	spec := meta.ServiceFromMap(map[string]interface{}{"name": "svc", "servicePath": "/open-apis/svc/v1"})
+	method := meta.FromMap(map[string]interface{}{"path": "items", "httpMethod": "GET", "parameters": map[string]interface{}{}})
 	cmd := NewCmdServiceMethod(f, spec, method, "list", "items", nil)
 	cmd.SetArgs([]string{"--as", "bot", "--page-all", "--jq", ".data.items[].id"})
 
@@ -630,8 +631,8 @@ func TestServiceMethod_PageAll_WithJq(t *testing.T) {
 
 // ── file upload ──
 
-func imImageMethod() map[string]interface{} {
-	return map[string]interface{}{
+func imImageMethod() meta.Method {
+	return meta.FromMap(map[string]interface{}{
 		"path":       "images",
 		"httpMethod": "POST",
 		"requestBody": map[string]interface{}{
@@ -645,14 +646,14 @@ func imImageMethod() map[string]interface{} {
 			},
 		},
 		"accessTokens": []interface{}{"user", "tenant"},
-	}
+	})
 }
 
-func imSpec() map[string]interface{} {
-	return map[string]interface{}{
+func imSpec() meta.Service {
+	return meta.ServiceFromMap(map[string]interface{}{
 		"name":        "im",
 		"servicePath": "/open-apis/im/v1",
-	}
+	})
 }
 
 func TestServiceMethod_FileFlagRegistered(t *testing.T) {
@@ -684,7 +685,7 @@ func TestServiceMethod_FileFlagNotRegisteredForGET(t *testing.T) {
 		},
 	}
 	f, _, _, _ := cmdutil.TestFactory(t, testConfig)
-	cmd := NewCmdServiceMethod(f, imSpec(), getMethod, "get", "images", nil)
+	cmd := NewCmdServiceMethod(f, imSpec(), meta.FromMap(getMethod), "get", "images", nil)
 	flag := cmd.Flags().Lookup("file")
 	if flag != nil {
 		t.Fatal("expected --file flag NOT to be registered for GET method")
@@ -752,7 +753,7 @@ func TestDetectFileFields(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			got := detectFileFields(tt.method)
+			got := detectFileFields(meta.FromMap(tt.method))
 			if len(got) != len(tt.want) {
 				t.Errorf("detectFileFields() = %v, want %v", got, tt.want)
 				return
@@ -771,7 +772,7 @@ func TestServiceMethod_JsonFlag_Accepted(t *testing.T) {
 
 	var captured *ServiceMethodOptions
 	cmd := NewCmdServiceMethod(f, driveSpec(),
-		map[string]interface{}{"description": "desc", "httpMethod": "GET"}, "list", "files",
+		meta.FromMap(map[string]interface{}{"description": "desc", "httpMethod": "GET"}), "list", "files",
 		func(opts *ServiceMethodOptions) error {
 			captured = opts
 			return nil

go.mod

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

go.sum

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

internal/apicatalog/catalog.go

@@ -0,0 +1,396 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package apicatalog is the single navigation Module over the API metadata. It
+// owns every "which services/resources/methods exist and how does a path
+// resolve" question that was previously duplicated across cmd/schema,
+// cmd/service, internal/schema and internal/registry. It depends only on
+// internal/meta; registry is the source Adapter (EmbeddedCatalog/RuntimeCatalog),
+// so apicatalog never imports registry.
+package apicatalog
+
+import (
+	"sort"
+	"strings"
+
+	"github.com/larksuite/cli/internal/meta"
+)
+
+// Source records whether a catalog includes the remote overlay. It is carried
+// so callers (and tests) can assert determinism instead of guessing.
+type Source string
+
+const (
+	SourceEmbedded Source = "embedded" // compiled-in metadata only; deterministic
+	SourceRuntime  Source = "runtime"  // embedded + remote overlay
+)
+
+// MethodFilter optionally drops methods (e.g. by identity in strict mode).
+// A nil filter includes everything.
+type MethodFilter func(meta.Method) bool
+
+// Catalog is a navigation view over services with a name index. It owns its
+// ordering — New sorts by name — so WalkMethods/Resolve/Complete are
+// deterministic regardless of how the source adapter ordered its input.
+type Catalog struct {
+	source   Source
+	services []meta.Service
+	byName   map[string]meta.Service
+}
+
+// New builds a Catalog over the given services, owning its navigation order:
+// the slice is copied and sorted by name so callers may pass any order and the
+// ordering contract is not delegated to the adapter. The copy is shallow —
+// meta.Service values share their Resources maps, which are treated as
+// read-only.
+func New(source Source, services []meta.Service) Catalog {
+	sorted := append([]meta.Service(nil), services...)
+	sort.Slice(sorted, func(i, j int) bool { return sorted[i].Name < sorted[j].Name })
+	byName := make(map[string]meta.Service, len(sorted))
+	for _, s := range sorted {
+		byName[s.Name] = s
+	}
+	return Catalog{source: source, services: sorted, byName: byName}
+}
+
+// Source reports embedded vs runtime.
+func (c Catalog) Source() Source { return c.source }
+
+// Services returns the services in name order. Treat the result as read-only:
+// it is the Catalog's own ordered slice and its element Resources maps are
+// shared.
+func (c Catalog) Services() []meta.Service { return c.services }
+
+// Service looks up one service by name.
+func (c Catalog) Service(name string) (meta.Service, bool) {
+	s, ok := c.byName[name]
+	return s, ok
+}
+
+// Resolve maps a path (already split into segments) to a Target. An empty path
+// is TargetAll. Failures return a *ResolveError carrying the available
+// candidates so the command layer can render a hint.
+func (c Catalog) Resolve(parts []string) (Target, error) {
+	if len(parts) == 0 {
+		return Target{Kind: TargetAll}, nil
+	}
+	svc, ok := c.byName[parts[0]]
+	if !ok {
+		return Target{}, &ResolveError{Kind: ErrService, Subject: parts[0], Candidates: c.serviceNames()}
+	}
+	if len(parts) == 1 {
+		return Target{Kind: TargetService, Service: svc}, nil
+	}
+	res, path, remaining, ok := findResource(svc, parts[1:])
+	if !ok {
+		return Target{}, &ResolveError{
+			Kind:       ErrResource,
+			Subject:    svc.Name + "." + strings.Join(parts[1:], "."),
+			Candidates: resourceNames(svc),
+		}
+	}
+	resPath := strings.Join(path, ".")
+	if len(remaining) == 0 {
+		return Target{Kind: TargetResource, Service: svc, Resource: &ResourceRef{Service: svc, Resource: res, Path: path}}, nil
+	}
+	methodName := remaining[0]
+	m, ok := res.Method(methodName)
+	if !ok {
+		return Target{}, &ResolveError{
+			Kind:       ErrMethod,
+			Subject:    svc.Name + "." + resPath + "." + methodName,
+			Candidates: methodNames(res),
+		}
+	}
+	if len(remaining) > 1 {
+		// Method exists but trailing segments don't resolve — reject so a typo
+		// doesn't silently return this method's schema.
+		return Target{}, &ResolveError{
+			Kind:     ErrPath,
+			Subject:  svc.Name + "." + resPath + "." + strings.Join(remaining, "."),
+			Method:   methodName,
+			Trailing: strings.Join(remaining[1:], "."),
+		}
+	}
+	return Target{Kind: TargetMethod, Service: svc, Method: &MethodRef{Service: svc, Resource: res, ResourcePath: path, Method: m}}, nil
+}
+
+// MethodRefs returns the method refs selected by a resolved Target, filtered:
+// TargetAll -> every method, TargetService / TargetResource -> that subtree,
+// TargetMethod -> the single method if it passes the filter (else empty). It
+// unifies WalkMethods/ServiceMethods/ResourceMethods so the command layer maps a
+// Target to refs in one call instead of re-deciding the walker per Kind.
+func (c Catalog) MethodRefs(target Target, filter MethodFilter) []MethodRef {
+	switch target.Kind {
+	case TargetService:
+		return ServiceMethods(target.Service, filter)
+	case TargetResource:
+		return ResourceMethods(*target.Resource, filter)
+	case TargetMethod:
+		if filter != nil && !filter(target.Method.Method) {
+			return nil
+		}
+		return []MethodRef{*target.Method}
+	case TargetAll:
+		return c.WalkMethods(filter)
+	default:
+		// Unknown / zero-value Kind: return nothing rather than silently
+		// dumping every method (the safe direction for an invalid Target).
+		return nil
+	}
+}
+
+// WalkMethods returns one MethodRef per method across all services (optionally
+// filtered), recursing nested resources, in a deterministic order: services by
+// name, resources by name, methods by name.
+func (c Catalog) WalkMethods(filter MethodFilter) []MethodRef {
+	var out []MethodRef
+	for _, svc := range c.services {
+		out = append(out, ServiceMethods(svc, filter)...)
+	}
+	return out
+}
+
+// ServiceMethods returns the method refs of one service (filtered), recursing
+// nested resources, in deterministic resource/method name order.
+func ServiceMethods(svc meta.Service, filter MethodFilter) []MethodRef {
+	var out []MethodRef
+	walkResources(svc, svc.ResourceList(), nil, filter, &out)
+	return out
+}
+
+// ResourceMethods returns the method refs under one resource (filtered), using
+// the resource's resolved path as the base and recursing nested resources.
+func ResourceMethods(r ResourceRef, filter MethodFilter) []MethodRef {
+	var out []MethodRef
+	for _, m := range r.Resource.MethodList() {
+		if filter == nil || filter(m) {
+			out = append(out, MethodRef{Service: r.Service, Resource: r.Resource, ResourcePath: r.Path, Method: m})
+		}
+	}
+	walkResources(r.Service, r.Resource.SubResources(), r.Path, filter, &out)
+	return out
+}
+
+func walkResources(svc meta.Service, resources []meta.Resource, parentPath []string, filter MethodFilter, out *[]MethodRef) {
+	for _, res := range resources {
+		path := append(append([]string(nil), parentPath...), res.Name)
+		for _, m := range res.MethodList() {
+			if filter == nil || filter(m) {
+				*out = append(*out, MethodRef{Service: svc, Resource: res, ResourcePath: path, Method: m})
+			}
+		}
+		walkResources(svc, res.SubResources(), path, filter, out)
+	}
+}
+
+// Complete returns shell-completion candidates for the schema path argument,
+// supporting both the legacy single dotted arg ("im.reac") and the
+// space-separated form ("im reactions"). noSpace mirrors cobra's
+// ShellCompDirectiveNoSpace (so "service." / "service.resource." stay open for
+// the next segment). Filtering uses the caller's MethodFilter so strict-mode
+// unavailable methods are hidden.
+func (c Catalog) Complete(args []string, toComplete string, filter MethodFilter) (completions []string, noSpace bool) {
+	// Case 1: legacy single dotted arg — no resolved args yet.
+	if len(args) == 0 {
+		parts := strings.Split(toComplete, ".")
+		if len(parts) <= 1 {
+			for _, name := range c.serviceNames() {
+				if strings.HasPrefix(name, toComplete) {
+					completions = append(completions, name+".")
+				}
+			}
+			return completions, true
+		}
+		svc, ok := c.byName[parts[0]]
+		if !ok {
+			return nil, false
+		}
+		completions = c.completeDotted(svc, strings.Join(parts[1:], "."), filter)
+		allTrailingDot := len(completions) > 0
+		for _, comp := range completions {
+			if !strings.HasSuffix(comp, ".") {
+				allTrailingDot = false
+				break
+			}
+		}
+		return completions, allTrailingDot
+	}
+
+	// Case 2: space-separated form — args holds resolved segments.
+	svc, ok := c.byName[args[0]]
+	if !ok {
+		return nil, false
+	}
+	resource, _, _, ok := findResource(svc, args[1:])
+	if !ok {
+		// No resource matched yet — suggest top-level resources reachable in the
+		// current identity mode.
+		return completeChildren(svc.ResourceList(), nil, toComplete, filter), false
+	}
+	// Positioned in a resource — offer its methods and its sub-resources, so the
+	// next segment can drill deeper, symmetric to findResource's descent.
+	return completeChildren(resource.SubResources(), resource.MethodList(), toComplete, filter), false
+}
+
+// completeDotted suggests dotted completions for the text after the service
+// segment. It descends fully-typed "resource." segments (longest match per
+// level, so flat dotted keys like "chat.members" and genuinely nested resources
+// both resolve), then offers the reachable sub-resources (as "…name.") and the
+// methods (as "…name") of the level it lands in whose names extend the trailing
+// partial token. This descent is symmetric to findResource, so completion can
+// reach every method Resolve can.
+func (c Catalog) completeDotted(svc meta.Service, afterService string, filter MethodFilter) []string {
+	subs := svc.ResourceList()
+	base := svc.Name
+	rest := afterService
+	var here *meta.Resource // resource we're positioned in; nil at the service root
+	for {
+		matched, n, ok := longestResourceFollowedByDot(subs, rest)
+		if !ok {
+			break
+		}
+		base += "." + matched.Name
+		rest = rest[n:]
+		r := matched
+		here = &r
+		subs = matched.SubResources()
+	}
+
+	var out []string
+	for _, sub := range subs {
+		if strings.HasPrefix(sub.Name, rest) && resourceReachable(sub, filter) {
+			out = append(out, base+"."+sub.Name+".")
+		}
+	}
+	if here != nil {
+		for _, m := range here.MethodList() {
+			if (filter == nil || filter(m)) && strings.HasPrefix(m.Name, rest) {
+				out = append(out, base+"."+m.Name)
+			}
+		}
+	}
+	sort.Strings(out)
+	return out
+}
+
+// completeChildren returns the sorted next-segment candidates at one level: the
+// (filtered) methods and the reachable sub-resources whose names extend prefix.
+// Methods are terminal; sub-resources are bare names the caller drills into on
+// the next segment.
+func completeChildren(subResources []meta.Resource, methods []meta.Method, prefix string, filter MethodFilter) []string {
+	var out []string
+	for _, m := range methods {
+		if (filter == nil || filter(m)) && strings.HasPrefix(m.Name, prefix) {
+			out = append(out, m.Name)
+		}
+	}
+	for _, sub := range subResources {
+		if strings.HasPrefix(sub.Name, prefix) && resourceReachable(sub, filter) {
+			out = append(out, sub.Name)
+		}
+	}
+	sort.Strings(out)
+	return out
+}
+
+// longestResourceFollowedByDot finds the longest resource in resources whose
+// name is a fully-typed segment of text (text begins with "name."), returning
+// it, the byte length consumed (incl. the dot), and whether one matched.
+func longestResourceFollowedByDot(resources []meta.Resource, text string) (meta.Resource, int, bool) {
+	best := meta.Resource{}
+	bestLen := -1
+	for _, r := range resources {
+		if len(r.Name) > bestLen && strings.HasPrefix(text, r.Name+".") {
+			best = r
+			bestLen = len(r.Name)
+		}
+	}
+	if bestLen < 0 {
+		return meta.Resource{}, 0, false
+	}
+	return best, len(best.Name) + 1, true
+}
+
+// findResource resolves a resource path against a service, descending nested
+// resources. At each level it consumes the longest leading run of parts that
+// names a resource at that level, so both flat dotted keys ("chat.members")
+// and genuinely nested resources ("spaces" > "items") resolve. This descent is
+// symmetric to walkResources, which guarantees every path WalkMethods emits
+// resolves back (the round-trip contract). Returns the deepest matched resource
+// (Name injected), its path segments, the unconsumed remainder, and whether
+// anything matched.
+//
+// Descent is greedy and resource-first: the one ambiguous case is a resource
+// that has BOTH a method and a sub-resource of the same name — the sub-resource
+// wins and shadows the method, so Resolve can never reach that method. Real
+// metadata never collides the two, so this is theoretical.
+func findResource(svc meta.Service, parts []string) (res meta.Resource, path []string, remaining []string, ok bool) {
+	level := svc.Resources
+	remaining = parts
+	for len(remaining) > 0 {
+		matched, name, n := longestResourcePrefix(level, remaining)
+		if n == 0 {
+			break
+		}
+		matched.Name = name
+		res = matched
+		path = append(path, name)
+		remaining = remaining[n:]
+		level = matched.Resources
+		ok = true
+	}
+	return res, path, remaining, ok
+}
+
+// longestResourcePrefix finds the longest leading run of segs (joined by ".")
+// that names a resource in level, returning the resource, its dotted name, and
+// the number of segments consumed (0 if none match). Longest-first lets a flat
+// dotted key win over its single leading segment when present.
+func longestResourcePrefix(level map[string]meta.Resource, segs []string) (meta.Resource, string, int) {
+	for i := len(segs); i >= 1; i-- {
+		name := strings.Join(segs[:i], ".")
+		if r, ok := level[name]; ok {
+			return r, name, i
+		}
+	}
+	return meta.Resource{}, "", 0
+}
+
+// resourceReachable reports whether a resource exposes a method reachable under
+// the filter — directly or in any nested sub-resource (a nil filter accepts any
+// method). A resource whose methods are all filtered out but which contains a
+// reachable nested method is still offerable, so completion can drill into it.
+func resourceReachable(res meta.Resource, filter MethodFilter) bool {
+	for _, m := range res.MethodList() {
+		if filter == nil || filter(m) {
+			return true
+		}
+	}
+	for _, sub := range res.SubResources() {
+		if resourceReachable(sub, filter) {
+			return true
+		}
+	}
+	return false
+}
+
+func (c Catalog) serviceNames() []string {
+	names := make([]string, len(c.services))
+	for i, s := range c.services {
+		names[i] = s.Name
+	}
+	return names // c.services is already name-sorted
+}
+
+func resourceNames(svc meta.Service) []string { return sortedKeys(svc.Resources) }
+func methodNames(res meta.Resource) []string  { return sortedKeys(res.Methods) }
+
+func sortedKeys[V any](m map[string]V) []string {
+	keys := make([]string, 0, len(m))
+	for k := range m {
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+	return keys
+}

internal/apicatalog/catalog_test.go

@@ -0,0 +1,340 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package apicatalog_test
+
+import (
+	"errors"
+	"reflect"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/internal/apicatalog"
+	"github.com/larksuite/cli/internal/meta"
+)
+
+// testCatalog builds a small embedded catalog: services drive (no resources)
+// and im with a dotted resource (chat.members), a multi-method resource
+// (reactions, where list is user-only), and images.
+func testCatalog() apicatalog.Catalog {
+	im := meta.ServiceFromMap(map[string]interface{}{
+		"name": "im",
+		"resources": map[string]interface{}{
+			"chat.members": map[string]interface{}{
+				"methods": map[string]interface{}{"create": map[string]interface{}{}},
+			},
+			"reactions": map[string]interface{}{
+				"methods": map[string]interface{}{
+					"create": map[string]interface{}{},
+					"list":   map[string]interface{}{"accessTokens": []interface{}{"user"}},
+				},
+			},
+			"images": map[string]interface{}{
+				"methods": map[string]interface{}{"create": map[string]interface{}{}},
+			},
+		},
+	})
+	drive := meta.ServiceFromMap(map[string]interface{}{"name": "drive"})
+	return apicatalog.New(apicatalog.SourceEmbedded, []meta.Service{drive, im}) // already name-sorted
+}
+
+func TestNew_PreservesOrderAndLookup(t *testing.T) {
+	c := testCatalog()
+	if c.Source() != apicatalog.SourceEmbedded {
+		t.Fatalf("source = %q", c.Source())
+	}
+	names := []string{}
+	for _, s := range c.Services() {
+		names = append(names, s.Name)
+	}
+	if !reflect.DeepEqual(names, []string{"drive", "im"}) {
+		t.Errorf("Services order = %v, want [drive im]", names)
+	}
+	if _, ok := c.Service("im"); !ok {
+		t.Error("Service(im) not found")
+	}
+	if _, ok := c.Service("nope"); ok {
+		t.Error("Service(nope) should not be found")
+	}
+}
+
+// TestNew_SortsAndIsolatesInput pins the ordering contract New owns: it sorts
+// arbitrary input by service name and shallow-copies the slice so later caller
+// mutation can't reorder the Catalog.
+func TestNew_SortsAndIsolatesInput(t *testing.T) {
+	in := []meta.Service{
+		meta.ServiceFromMap(map[string]interface{}{"name": "zeta"}),
+		meta.ServiceFromMap(map[string]interface{}{"name": "alpha"}),
+	}
+	c := apicatalog.New(apicatalog.SourceEmbedded, in)
+
+	names := func() []string {
+		var out []string
+		for _, s := range c.Services() {
+			out = append(out, s.Name)
+		}
+		return out
+	}
+	if got := names(); !reflect.DeepEqual(got, []string{"alpha", "zeta"}) {
+		t.Errorf("New did not sort unsorted input: %v", got)
+	}
+
+	// Mutating the caller's slice afterward must not reorder the Catalog.
+	in[0] = meta.ServiceFromMap(map[string]interface{}{"name": "MUTATED"})
+	if got := names(); !reflect.DeepEqual(got, []string{"alpha", "zeta"}) {
+		t.Errorf("Catalog order changed after caller mutated its input slice: %v", got)
+	}
+}
+
+func TestWalkMethods_AllAndFiltered(t *testing.T) {
+	c := testCatalog()
+
+	all := c.WalkMethods(nil)
+	got := map[string]bool{}
+	for _, r := range all {
+		got[r.SchemaPath()] = true
+	}
+	want := []string{
+		"im.chat.members.create",
+		"im.images.create",
+		"im.reactions.create",
+		"im.reactions.list",
+	}
+	if len(all) != len(want) {
+		t.Fatalf("WalkMethods(nil) = %d refs, want %d (%v)", len(all), len(want), got)
+	}
+	for _, w := range want {
+		if !got[w] {
+			t.Errorf("WalkMethods(nil) missing %q", w)
+		}
+	}
+
+	// Deterministic order: services by name, resources by name, methods by name.
+	var order []string
+	for _, r := range all {
+		order = append(order, r.SchemaPath())
+	}
+	if !reflect.DeepEqual(order, want) {
+		t.Errorf("WalkMethods order = %v, want %v", order, want)
+	}
+
+	// Filter to bot-only ("tenant"): reactions.list (user-only) drops; methods
+	// with no accessTokens are permissive and stay.
+	botOnly := func(m meta.Method) bool {
+		if m.AccessTokens == nil {
+			return true
+		}
+		for _, tok := range m.AccessTokens {
+			if tok == "tenant" {
+				return true
+			}
+		}
+		return false
+	}
+	filtered := c.WalkMethods(botOnly)
+	for _, r := range filtered {
+		if r.SchemaPath() == "im.reactions.list" {
+			t.Error("filtered walk should drop user-only im.reactions.list")
+		}
+	}
+	if len(filtered) != len(all)-1 {
+		t.Errorf("filtered walk = %d, want %d", len(filtered), len(all)-1)
+	}
+}
+
+func TestMethodRef_Paths_DottedResourceStaysOneSegment(t *testing.T) {
+	c := testCatalog()
+	target, err := c.Resolve([]string{"im", "chat.members", "create"})
+	if err != nil {
+		t.Fatalf("resolve: %v", err)
+	}
+	if target.Kind != apicatalog.TargetMethod {
+		t.Fatalf("kind = %v", target.Kind)
+	}
+	m := target.Method
+	if m.SchemaPath() != "im.chat.members.create" {
+		t.Errorf("SchemaPath = %q", m.SchemaPath())
+	}
+	if !reflect.DeepEqual(m.CommandPath(), []string{"im", "chat.members", "create"}) {
+		t.Errorf("CommandPath = %v", m.CommandPath())
+	}
+	if m.ResourceName() != "chat.members" {
+		t.Errorf("ResourceName = %q, want chat.members (one segment)", m.ResourceName())
+	}
+	if m.Method.Name != "create" {
+		t.Errorf("Method.Name not injected: %q", m.Method.Name)
+	}
+}
+
+func TestResolve_DottedAndSplitFormsEquivalent(t *testing.T) {
+	c := testCatalog()
+	// schema.ParsePath splits both "im.chat.members.create" and
+	// "im chat.members create" into segments; findResource's longest-prefix
+	// must resolve the dotted resource either way.
+	a, errA := c.Resolve([]string{"im", "chat", "members", "create"}) // fully split
+	b, errB := c.Resolve([]string{"im", "chat.members", "create"})    // resource as one segment
+	if errA != nil || errB != nil {
+		t.Fatalf("errA=%v errB=%v", errA, errB)
+	}
+	if a.Method.SchemaPath() != b.Method.SchemaPath() || a.Method.SchemaPath() != "im.chat.members.create" {
+		t.Errorf("forms diverged: %q vs %q", a.Method.SchemaPath(), b.Method.SchemaPath())
+	}
+}
+
+func TestResolve_Targets(t *testing.T) {
+	c := testCatalog()
+	if tg, _ := c.Resolve(nil); tg.Kind != apicatalog.TargetAll {
+		t.Errorf("empty -> %v, want all", tg.Kind)
+	}
+	if tg, _ := c.Resolve([]string{"im"}); tg.Kind != apicatalog.TargetService || tg.Service.Name != "im" {
+		t.Errorf("[im] -> %v/%q", tg.Kind, tg.Service.Name)
+	}
+	if tg, _ := c.Resolve([]string{"im", "reactions"}); tg.Kind != apicatalog.TargetResource || tg.Resource.SchemaPath() != "im.reactions" {
+		t.Errorf("[im reactions] -> %v", tg.Kind)
+	}
+}
+
+func TestResolve_Errors(t *testing.T) {
+	c := testCatalog()
+	cases := []struct {
+		parts []string
+		kind  apicatalog.ResolveErrorKind
+	}{
+		{[]string{"nope"}, apicatalog.ErrService},
+		{[]string{"im", "nope"}, apicatalog.ErrResource},
+		{[]string{"im", "reactions", "nope"}, apicatalog.ErrMethod},
+		{[]string{"im", "reactions", "list", "extra"}, apicatalog.ErrPath},
+	}
+	for _, tc := range cases {
+		_, err := c.Resolve(tc.parts)
+		var re *apicatalog.ResolveError
+		if !errors.As(err, &re) {
+			t.Errorf("%v -> err %v, want *ResolveError", tc.parts, err)
+			continue
+		}
+		if re.Kind != tc.kind {
+			t.Errorf("%v -> kind %q, want %q", tc.parts, re.Kind, tc.kind)
+		}
+		if tc.kind != apicatalog.ErrPath && len(re.Candidates) == 0 {
+			t.Errorf("%v -> expected candidates", tc.parts)
+		}
+	}
+}
+
+// nestedCatalog adds a genuinely nested resource (spaces > items) on top of a
+// flat dotted resource (chat.members), so the round-trip contract is exercised
+// for real nesting — not just flat dotted keys.
+func nestedCatalog() apicatalog.Catalog {
+	im := meta.ServiceFromMap(map[string]interface{}{
+		"name": "im",
+		"resources": map[string]interface{}{
+			"chat.members": map[string]interface{}{
+				"methods": map[string]interface{}{"create": map[string]interface{}{}},
+			},
+			"spaces": map[string]interface{}{
+				"methods": map[string]interface{}{"create": map[string]interface{}{}},
+				"resources": map[string]interface{}{
+					"items": map[string]interface{}{
+						"methods": map[string]interface{}{"get": map[string]interface{}{}},
+					},
+				},
+			},
+		},
+	})
+	return apicatalog.New(apicatalog.SourceEmbedded, []meta.Service{im})
+}
+
+// TestResolve_WalkMethodsRoundTrip is the core catalog contract: every method
+// WalkMethods emits must Resolve back to the same method — both from its dotted
+// SchemaPath (fully split) and from its CommandPath (resource as one segment).
+// This pins findResource's nested-resource descent symmetric to walkResources,
+// so "traversable" implies "resolvable".
+func TestResolve_WalkMethodsRoundTrip(t *testing.T) {
+	for _, c := range []apicatalog.Catalog{testCatalog(), nestedCatalog()} {
+		for _, ref := range c.WalkMethods(nil) {
+			want := ref.SchemaPath()
+			for _, parts := range [][]string{
+				strings.Split(want, "."), // fully-split dotted form
+				ref.CommandPath(),        // command form (resource stays one segment)
+			} {
+				tg, err := c.Resolve(parts)
+				if err != nil {
+					t.Errorf("round-trip %v: %v", parts, err)
+					continue
+				}
+				if tg.Kind != apicatalog.TargetMethod {
+					t.Errorf("round-trip %v: kind=%v, want method", parts, tg.Kind)
+					continue
+				}
+				if tg.Method.SchemaPath() != want {
+					t.Errorf("round-trip %v: resolved to %q, want %q", parts, tg.Method.SchemaPath(), want)
+				}
+			}
+		}
+	}
+}
+
+// TestComplete_Nested pins completion closure for genuinely nested resources:
+// both the dotted and space forms must reach a nested method, symmetric to
+// Resolve (findResource descends, so completion must too).
+func TestComplete_Nested(t *testing.T) {
+	c := nestedCatalog()
+
+	// dotted: under a resource, offer its methods AND its sub-resources
+	if comps, ns := c.Complete(nil, "im.spaces.", nil); !reflect.DeepEqual(comps, []string{"im.spaces.create", "im.spaces.items."}) || ns {
+		t.Errorf("Complete([], im.spaces.) = %v noSpace=%v, want [im.spaces.create im.spaces.items.] false", comps, ns)
+	}
+	// dotted: drill into the nested sub-resource's method
+	if comps, ns := c.Complete(nil, "im.spaces.items.", nil); !reflect.DeepEqual(comps, []string{"im.spaces.items.get"}) || ns {
+		t.Errorf("Complete([], im.spaces.items.) = %v noSpace=%v, want [im.spaces.items.get] false", comps, ns)
+	}
+	// dotted: partial sub-resource name -> the sub-resource (NoSpace, more to type)
+	if comps, ns := c.Complete(nil, "im.spaces.it", nil); !reflect.DeepEqual(comps, []string{"im.spaces.items."}) || !ns {
+		t.Errorf("Complete([], im.spaces.it) = %v noSpace=%v, want [im.spaces.items.] true", comps, ns)
+	}
+	// space form: under a resource, offer methods AND sub-resources
+	if comps, _ := c.Complete([]string{"im", "spaces"}, "", nil); !reflect.DeepEqual(comps, []string{"create", "items"}) {
+		t.Errorf("Complete([im spaces], '') = %v, want [create items]", comps)
+	}
+	// space form: drill into the nested sub-resource's methods
+	if comps, _ := c.Complete([]string{"im", "spaces", "items"}, "", nil); !reflect.DeepEqual(comps, []string{"get"}) {
+		t.Errorf("Complete([im spaces items], '') = %v, want [get]", comps)
+	}
+}
+
+func TestComplete(t *testing.T) {
+	c := testCatalog()
+
+	// dotted: service prefix -> "im." (NoSpace)
+	if comps, ns := c.Complete(nil, "i", nil); !reflect.DeepEqual(comps, []string{"im."}) || !ns {
+		t.Errorf("Complete([], i) = %v noSpace=%v", comps, ns)
+	}
+	// dotted: resource prefix -> "im.reactions." (NoSpace)
+	if comps, _ := c.Complete(nil, "im.rea", nil); !reflect.DeepEqual(comps, []string{"im.reactions."}) {
+		t.Errorf("Complete([], im.rea) = %v", comps)
+	}
+	// space form: resource candidates under im (deterministic order)
+	comps, ns := c.Complete([]string{"im"}, "", nil)
+	if !reflect.DeepEqual(comps, []string{"chat.members", "images", "reactions"}) || ns {
+		t.Errorf("Complete([im], '') = %v noSpace=%v", comps, ns)
+	}
+	// space form: method candidates under reactions
+	if comps, _ := c.Complete([]string{"im", "reactions"}, "", nil); !reflect.DeepEqual(comps, []string{"create", "list"}) {
+		t.Errorf("Complete([im reactions], '') = %v", comps)
+	}
+	// filter applied: bot-only hides user-only list
+	botOnly := func(m meta.Method) bool {
+		if m.AccessTokens == nil {
+			return true
+		}
+		for _, tok := range m.AccessTokens {
+			if tok == "tenant" {
+				return true
+			}
+		}
+		return false
+	}
+	if comps, _ := c.Complete([]string{"im", "reactions"}, "", botOnly); !reflect.DeepEqual(comps, []string{"create"}) {
+		t.Errorf("Complete with bot filter = %v, want [create]", comps)
+	}
+}

internal/apicatalog/methodref.go

@@ -0,0 +1,75 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package apicatalog
+
+import (
+	"strings"
+
+	"github.com/larksuite/cli/internal/meta"
+)
+
+// TargetKind classifies what a schema/command path resolves to.
+type TargetKind string
+
+const (
+	TargetAll      TargetKind = "all"      // empty path: every method
+	TargetService  TargetKind = "service"  // <service>
+	TargetResource TargetKind = "resource" // <service> <resource...>
+	TargetMethod   TargetKind = "method"   // <service> <resource...> <method>
+)
+
+// Target is the result of Catalog.Resolve. Resource and Method are populated
+// only for TargetResource and TargetMethod respectively.
+type Target struct {
+	Kind     TargetKind
+	Service  meta.Service
+	Resource *ResourceRef
+	Method   *MethodRef
+}
+
+// ResourceRef identifies one resource within a service. Path holds the resource
+// path segments (one element for the common flat dotted resource like
+// "chat.members"; multiple for genuinely nested resources).
+type ResourceRef struct {
+	Service  meta.Service
+	Resource meta.Resource
+	Path     []string
+}
+
+// MethodRef identifies one method, carrying the full navigation context so the
+// command path and schema path can be derived without re-walking the catalog.
+type MethodRef struct {
+	Service      meta.Service
+	Resource     meta.Resource
+	ResourcePath []string
+	Method       meta.Method
+}
+
+// SchemaPath is the dotted "service.resource" identifier.
+func (r ResourceRef) SchemaPath() string {
+	return r.Service.Name + "." + strings.Join(r.Path, ".")
+}
+
+// ServiceName returns the owning service name.
+func (r MethodRef) ServiceName() string { return r.Service.Name }
+
+// ResourceName is the dotted resource path, e.g. "chat.members".
+func (r MethodRef) ResourceName() string { return strings.Join(r.ResourcePath, ".") }
+
+// MethodName returns the method's own name.
+func (r MethodRef) MethodName() string { return r.Method.Name }
+
+// SchemaPath is the dotted "service.resource.method" identifier, e.g.
+// "im.chat.members.create".
+func (r MethodRef) SchemaPath() string {
+	return r.Service.Name + "." + strings.Join(r.ResourcePath, ".") + "." + r.Method.Name
+}
+
+// CommandPath is the CLI argv segments, e.g. ["im", "chat.members", "create"].
+func (r MethodRef) CommandPath() []string {
+	out := make([]string, 0, len(r.ResourcePath)+2)
+	out = append(out, r.Service.Name)
+	out = append(out, r.ResourcePath...)
+	return append(out, r.Method.Name)
+}

internal/apicatalog/path.go

@@ -0,0 +1,31 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package apicatalog
+
+import "strings"
+
+// ParsePath normalizes positional command arguments into the path segments
+// Resolve consumes. It accepts two equivalent forms:
+//
+//	im.messages.reply  -> single arg, split on "."
+//	im messages reply  -> multiple args, used as-is
+//
+// "im chat.members bots" as a single quoted arg is NOT supported; quote
+// arguments individually if your shell needs it. A resource keeps its internal
+// dots when passed as one segment (e.g. "chat.members"); findResource's
+// longest-prefix descent resolves both the split and the one-segment forms to
+// the same target. Returns nil for zero args (bare invocation -> TargetAll).
+func ParsePath(args []string) []string {
+	switch len(args) {
+	case 0:
+		return nil
+	case 1:
+		if strings.Contains(args[0], ".") {
+			return strings.Split(args[0], ".")
+		}
+		return []string{args[0]}
+	default:
+		return args
+	}
+}

internal/apicatalog/path_test.go

@@ -1,11 +1,13 @@
 // Copyright (c) 2026 Lark Technologies Pte. Ltd.
 // SPDX-License-Identifier: MIT
 
-package schema
+package apicatalog_test
 
 import (
 	"reflect"
 	"testing"
+
+	"github.com/larksuite/cli/internal/apicatalog"
 )
 
 func TestParsePath(t *testing.T) {
@@ -25,7 +27,7 @@ func TestParsePath(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			got := ParsePath(tt.args)
+			got := apicatalog.ParsePath(tt.args)
 			if !reflect.DeepEqual(got, tt.want) {
 				t.Errorf("ParsePath(%v) = %v, want %v", tt.args, got, tt.want)
 			}

internal/apicatalog/resolveerror.go

@@ -0,0 +1,30 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package apicatalog
+
+// ResolveErrorKind classifies a Resolve failure so the command layer can render
+// the right hint without re-deriving what was being looked up.
+type ResolveErrorKind string
+
+const (
+	ErrService  ResolveErrorKind = "service"
+	ErrResource ResolveErrorKind = "resource"
+	ErrMethod   ResolveErrorKind = "method"
+	ErrPath     ResolveErrorKind = "path" // method exists but trailing segments don't resolve
+)
+
+// ResolveError is returned by Catalog.Resolve. Subject is the dotted thing that
+// failed to resolve; Candidates lists the available names at that level (nil for
+// ErrPath, which instead carries the matched Method and the unresolved Trailing).
+type ResolveError struct {
+	Kind       ResolveErrorKind
+	Subject    string
+	Candidates []string
+	Method     string
+	Trailing   string
+}
+
+func (e *ResolveError) Error() string {
+	return "unknown " + string(e.Kind) + ": " + e.Subject
+}

internal/cmdutil/factory_default.go

@@ -100,9 +100,19 @@ func safeRedirectPolicy(req *http.Request, via []*http.Request) error {
 	return nil
 }
 
+// warnIfProxied is a test seam for the proxy-warning gate. Production wires it
+// to transport.WarnIfProxied; tests swap in a spy to count invocations. It is
+// needed because the real function is guarded by an internal sync.Once, so
+// calling it directly would only fire on the first test (see
+// factory_proxy_warn_test.go). The terminal check is the IOStreams
+// .StderrIsTerminal field, which tests set directly.
+var warnIfProxied = transport.WarnIfProxied
+
 func cachedHttpClientFunc(f *Factory) func() (*http.Client, error) {
 	return sync.OnceValues(func() (*http.Client, error) {
-		transport.WarnIfProxied(f.IOStreams.ErrOut)
+		if f.IOStreams.StderrIsTerminal {
+			warnIfProxied(f.IOStreams.ErrOut)
+		}
 
 		var rt http.RoundTripper = transport.Shared()
 		rt = &RetryTransport{Base: rt}
@@ -129,7 +139,9 @@ func cachedLarkClientFunc(f *Factory) func() (*lark.Client, error) {
 			lark.WithLogLevel(larkcore.LogLevelError),
 			lark.WithHeaders(BaseSecurityHeaders()),
 		}
-		transport.WarnIfProxied(f.IOStreams.ErrOut)
+		if f.IOStreams.StderrIsTerminal {
+			warnIfProxied(f.IOStreams.ErrOut)
+		}
 		opts = append(opts, lark.WithHttpClient(&http.Client{
 			Transport:     buildSDKTransport(),
 			CheckRedirect: safeRedirectPolicy,

internal/cmdutil/factory_proxy_warn_test.go

@@ -0,0 +1,85 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package cmdutil
+
+import (
+	"io"
+	"testing"
+
+	_ "github.com/larksuite/cli/extension/credential/env" // registers the env-backed account provider
+	"github.com/larksuite/cli/internal/envvars"
+)
+
+// installProxyWarnSpy replaces warnIfProxied with a counter for one test and
+// restores it on cleanup. Returns a pointer to the call count so the caller can
+// assert how many times the warning fired. The terminal state is controlled via
+// the IOStreams.StderrIsTerminal field, not a seam.
+func installProxyWarnSpy(t *testing.T) *int {
+	t.Helper()
+	prevWarn := warnIfProxied
+	t.Cleanup(func() { warnIfProxied = prevWarn })
+	calls := 0
+	warnIfProxied = func(io.Writer) { calls++ }
+	return &calls
+}
+
+var proxyWarnGateCases = []struct {
+	name     string
+	terminal bool
+	want     int
+}{
+	{"terminal stderr warns once", true, 1},
+	{"non-terminal stderr stays silent", false, 0},
+}
+
+// TestCachedHttpClientFunc_ProxyWarnGate verifies the http-client init path
+// invokes WarnIfProxied only when stderr is an interactive terminal.
+func TestCachedHttpClientFunc_ProxyWarnGate(t *testing.T) {
+	for _, tc := range proxyWarnGateCases {
+		t.Run(tc.name, func(t *testing.T) {
+			calls := installProxyWarnSpy(t)
+
+			fn := cachedHttpClientFunc(&Factory{IOStreams: &IOStreams{
+				ErrOut: io.Discard, StderrIsTerminal: tc.terminal,
+			}})
+			if _, err := fn(); err != nil {
+				t.Fatalf("http client init: %v", err)
+			}
+
+			if *calls != tc.want {
+				t.Errorf("WarnIfProxied calls = %d, want %d", *calls, tc.want)
+			}
+		})
+	}
+}
+
+// TestCachedLarkClientFunc_ProxyWarnGate verifies the lark-client init path
+// invokes WarnIfProxied only when stderr is an interactive terminal. The gate
+// runs after ResolveAccount, so an env-backed credential is wired up to let
+// account resolution succeed without network or config files.
+func TestCachedLarkClientFunc_ProxyWarnGate(t *testing.T) {
+	for _, tc := range proxyWarnGateCases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Setenv(envvars.CliAppID, "env-app")
+			t.Setenv(envvars.CliAppSecret, "env-secret")
+			t.Setenv(envvars.CliDefaultAs, "")
+			t.Setenv(envvars.CliUserAccessToken, "")
+			t.Setenv(envvars.CliTenantAccessToken, "")
+			t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+			calls := installProxyWarnSpy(t)
+
+			// normalizeStreams copies the struct (out := *s), so the
+			// StderrIsTerminal field survives into f.IOStreams.
+			f := NewDefault(&IOStreams{ErrOut: io.Discard, StderrIsTerminal: tc.terminal}, InvocationContext{})
+			if _, err := cachedLarkClientFunc(f)(); err != nil {
+				t.Fatalf("lark client init: %v", err)
+			}
+
+			if *calls != tc.want {
+				t.Errorf("WarnIfProxied calls = %d, want %d", *calls, tc.want)
+			}
+		})
+	}
+}

internal/cmdutil/fileupload.go

@@ -12,23 +12,9 @@ import (
 
 	"github.com/larksuite/cli/extension/fileio"
 	"github.com/larksuite/cli/internal/output"
-	"github.com/larksuite/cli/internal/registry"
 	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
 )
 
-// DetectFileFields returns field names with type "file" in the method's requestBody.
-func DetectFileFields(method map[string]interface{}) []string {
-	rb, _ := method["requestBody"].(map[string]interface{})
-	var fields []string
-	for name, field := range rb {
-		f, _ := field.(map[string]interface{})
-		if registry.GetStrFromMap(f, "type") == "file" {
-			fields = append(fields, name)
-		}
-	}
-	return fields
-}
-
 // ParseFileFlag parses a --file flag value into its components.
 // The format is either "path" or "field=path". When no explicit "field="
 // prefix is present, defaultField is used as the field name.

internal/cmdutil/identity.go

@@ -10,22 +10,6 @@ import (
 	"github.com/larksuite/cli/internal/core"
 )
 
-// AccessTokensToIdentities converts from_meta accessTokens (e.g. ["tenant", "user"])
-// to CLI identity names (e.g. ["bot", "user"]).
-func AccessTokensToIdentities(tokens []interface{}) []string {
-	var identities []string
-	for _, t := range tokens {
-		if ts, ok := t.(string); ok {
-			if ts == "tenant" {
-				identities = append(identities, "bot")
-			} else {
-				identities = append(identities, ts)
-			}
-		}
-	}
-	return identities
-}
-
 // PrintIdentity outputs the current identity to stderr so callers (including AI agents)
 // can see which identity is being used for the API call.
 func PrintIdentity(w io.Writer, as core.Identity, config *core.CliConfig, autoDetected bool) {

internal/cmdutil/identity_test.go

@@ -11,54 +11,6 @@ import (
 	"github.com/larksuite/cli/internal/core"
 )
 
-func TestAccessTokensToIdentities(t *testing.T) {
-	tests := []struct {
-		name   string
-		tokens []interface{}
-		want   []string
-	}{
-		{
-			name:   "tenant becomes bot",
-			tokens: []interface{}{"tenant"},
-			want:   []string{"bot"},
-		},
-		{
-			name:   "user stays user",
-			tokens: []interface{}{"user"},
-			want:   []string{"user"},
-		},
-		{
-			name:   "tenant and user",
-			tokens: []interface{}{"tenant", "user"},
-			want:   []string{"bot", "user"},
-		},
-		{
-			name:   "empty list",
-			tokens: []interface{}{},
-			want:   nil,
-		},
-		{
-			name:   "non-string values skipped",
-			tokens: []interface{}{"tenant", 42, "user"},
-			want:   []string{"bot", "user"},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			got := AccessTokensToIdentities(tt.tokens)
-			if len(got) != len(tt.want) {
-				t.Fatalf("len: want %d, got %d (%v)", len(tt.want), len(got), got)
-			}
-			for i := range got {
-				if got[i] != tt.want[i] {
-					t.Errorf("[%d] want %s, got %s", i, tt.want[i], got[i])
-				}
-			}
-		})
-	}
-}
-
 func TestPrintIdentity_BotExplicit(t *testing.T) {
 	var buf bytes.Buffer
 	PrintIdentity(&buf, core.AsBot, nil, false)

internal/cmdutil/iostreams.go

@@ -18,17 +18,28 @@ type IOStreams struct {
 	Out        io.Writer
 	ErrOut     io.Writer
 	IsTerminal bool
+	// StderrIsTerminal reports whether ErrOut is an interactive terminal.
+	// Advisory warnings written to stderr (e.g. the proxy notice) gate on this
+	// so they stay out of non-interactive output (pipes, CI, agent runs).
+	// Computed once in NewIOStreams, mirroring IsTerminal; tests assign it
+	// directly like cmd/config/bind_test.go does for IsTerminal.
+	StderrIsTerminal bool
 }
 
 // NewIOStreams builds an IOStreams from arbitrary readers/writers.
-// IsTerminal is derived from in's underlying *os.File, if any; non-file
-// readers (bytes.Buffer, strings.Reader, …) yield IsTerminal=false.
+// IsTerminal / StderrIsTerminal are derived from in's / errOut's underlying
+// *os.File, if any; non-file streams (bytes.Buffer, strings.Reader, …) yield
+// false.
 func NewIOStreams(in io.Reader, out, errOut io.Writer) *IOStreams {
 	isTerminal := false
 	if f, ok := in.(*os.File); ok {
 		isTerminal = term.IsTerminal(int(f.Fd()))
 	}
-	return &IOStreams{In: in, Out: out, ErrOut: errOut, IsTerminal: isTerminal}
+	stderrIsTerminal := false
+	if f, ok := errOut.(*os.File); ok {
+		stderrIsTerminal = term.IsTerminal(int(f.Fd()))
+	}
+	return &IOStreams{In: in, Out: out, ErrOut: errOut, IsTerminal: isTerminal, StderrIsTerminal: stderrIsTerminal}
 }
 
 // SystemIO creates an IOStreams wired to the process's standard file descriptors.

internal/cmdutil/json.go

@@ -34,7 +34,9 @@ func ParseOptionalBody(httpMethod, data string, stdin io.Reader, fileIO fileio.F
 	return body, nil
 }
 
-// ParseJSONMap parses a JSON string into a map. Returns an empty map if input is empty.
+// ParseJSONMap parses a JSON string into a map. Returns an empty (never nil) map
+// for empty input or the JSON literal null, so callers can always overlay onto
+// the result without a nil-map panic.
 // Supports stdin (-), @file, @@-escape, and single-quote stripping via ResolveInput.
 func ParseJSONMap(input, label string, stdin io.Reader, fileIO fileio.FileIO) (map[string]any, error) {
 	resolved, err := ResolveInput(input, stdin, fileIO)
@@ -48,5 +50,10 @@ func ParseJSONMap(input, label string, stdin io.Reader, fileIO fileio.FileIO) (m
 	if err := json.Unmarshal([]byte(resolved), &result); err != nil {
 		return nil, output.ErrValidation("%s invalid format, expected JSON object", label)
 	}
+	if result == nil {
+		// `null` unmarshals into a nil map without error; normalize it so the
+		// returned map is always writable, matching the empty-input case.
+		return map[string]any{}, nil
+	}
 	return result, nil
 }

internal/cmdutil/json_test.go

@@ -47,6 +47,7 @@ func TestParseJSONMap(t *testing.T) {
 		wantErr bool
 	}{
 		{"empty input", "", "--params", 0, false},
+		{"json null", "null", "--params", 0, false},
 		{"valid json", `{"a":"1","b":"2"}`, "--params", 2, false},
 		{"invalid json", `{bad}`, "--params", 0, true},
 		{"json array", `[1,2]`, "--data", 0, true},
@@ -61,6 +62,12 @@ func TestParseJSONMap(t *testing.T) {
 			if !tt.wantErr && len(got) != tt.wantLen {
 				t.Errorf("ParseJSONMap() returned map with %d keys, want %d", len(got), tt.wantLen)
 			}
+			// A successful parse must yield a non-nil, writable map: callers
+			// overlay onto it (params[k]=v), so `null` — which unmarshals to a
+			// nil map without error — must normalize to {} like empty input.
+			if !tt.wantErr && got == nil {
+				t.Error("ParseJSONMap() = nil map on success, want non-nil")
+			}
 		})
 	}
 }

internal/cmdutil/risk.go

@@ -3,17 +3,20 @@
 
 package cmdutil
 
-import "github.com/spf13/cobra"
+import (
+	"github.com/larksuite/cli/internal/core"
+	"github.com/spf13/cobra"
+)
 
 const riskLevelAnnotationKey = "risk_level"
 
-// Risk level constants — the three-tier convention used across the CLI.
-// Use these in place of string literals so the typo radius is one place,
-// not every call site.
+// Risk level constants — aliases of the canonical core.Risk* values, re-exported
+// here so command code gets the risk vocabulary and the SetRisk/GetRisk helpers
+// from one package. core is the single source of truth.
 const (
-	RiskRead          = "read"
-	RiskWrite         = "write"
-	RiskHighRiskWrite = "high-risk-write"
+	RiskRead          = core.RiskRead
+	RiskWrite         = core.RiskWrite
+	RiskHighRiskWrite = core.RiskHighRiskWrite
 )
 
 // SetRisk stores a command's static risk level on cobra annotations so the

internal/core/risk.go

@@ -0,0 +1,15 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package core
+
+// Risk levels — the three-tier convention used across the CLI. They live here,
+// at the leaf, so the envelope renderer (internal/schema) and the command
+// toolkit (internal/cmdutil) share one vocabulary without a renderer depending
+// on command utilities. Framework confirmation gating acts only on
+// RiskHighRiskWrite.
+const (
+	RiskRead          = "read"
+	RiskWrite         = "write"
+	RiskHighRiskWrite = "high-risk-write"
+)

internal/meta/affordance.go

@@ -0,0 +1,44 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import "encoding/json"
+
+// Affordance is the hand-authored usage guidance overlaid on a method: when to
+// use it, when not to, prerequisites, few-shot examples, and related methods.
+// It is the single typed model of the affordance shape; the envelope renderer
+// and the command help both parse through ParsedAffordance so the vocabulary
+// is defined once. The JSON tags double as the envelope's wire shape.
+type Affordance struct {
+	UseWhen       []string         `json:"use_when,omitempty"`
+	DoNotUseWhen  []string         `json:"do_not_use_when,omitempty"`
+	Prerequisites []string         `json:"prerequisites,omitempty"`
+	Examples      []AffordanceCase `json:"examples,omitempty"`
+	Related       []string         `json:"related,omitempty"`
+}
+
+// AffordanceCase is one few-shot example: a one-line description and a
+// ready-to-run command.
+type AffordanceCase struct {
+	Description string `json:"description"`
+	Command     string `json:"command"`
+}
+
+// ParsedAffordance decodes the method's raw affordance overlay into the typed
+// Affordance. ok is false when the method carries no affordance, the JSON is
+// malformed, or every section is empty — so callers can treat "no guidance"
+// uniformly.
+func (m Method) ParsedAffordance() (Affordance, bool) {
+	if len(m.Affordance) == 0 {
+		return Affordance{}, false
+	}
+	var a Affordance
+	if json.Unmarshal(m.Affordance, &a) != nil {
+		return Affordance{}, false
+	}
+	if len(a.UseWhen) == 0 && len(a.DoNotUseWhen) == 0 && len(a.Prerequisites) == 0 && len(a.Examples) == 0 && len(a.Related) == 0 {
+		return Affordance{}, false
+	}
+	return a, true
+}

internal/meta/affordance_test.go

@@ -0,0 +1,56 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import (
+	"encoding/json"
+	"testing"
+)
+
+func TestMethod_ParsedAffordance(t *testing.T) {
+	// absent / empty / malformed all resolve to ok=false.
+	t.Run("nil affordance", func(t *testing.T) {
+		if _, ok := (Method{}).ParsedAffordance(); ok {
+			t.Error("ParsedAffordance on a method without affordance ok=true, want false")
+		}
+	})
+
+	notOK := map[string]string{
+		"empty payload":        ``,
+		"empty object":         `{}`,
+		"all empty arrays":     `{"use_when":[],"do_not_use_when":[],"prerequisites":[],"examples":[],"related":[]}`,
+		"malformed string":     `"not an object"`,
+		"malformed number":     `42`,
+		"nested type mismatch": `{"examples":"should be a list"}`,
+	}
+	for name, raw := range notOK {
+		t.Run(name, func(t *testing.T) {
+			if _, ok := (Method{Affordance: json.RawMessage(raw)}).ParsedAffordance(); ok {
+				t.Errorf("ParsedAffordance(%s) ok=true, want false", raw)
+			}
+		})
+	}
+
+	// Populated affordance parses with all fields.
+	raw := `{
+		"use_when": ["需要拿到当前用户的主日历 ID"],
+		"do_not_use_when": ["已知具体 calendar_id"],
+		"prerequisites": ["user 身份登录"],
+		"examples": [{"description":"获取主日历","command":"lark-cli calendar calendars primary"}],
+		"related": ["calendars.list"]
+	}`
+	a, ok := (Method{Affordance: json.RawMessage(raw)}).ParsedAffordance()
+	if !ok {
+		t.Fatal("ParsedAffordance ok=false, want populated")
+	}
+	if len(a.UseWhen) != 1 || a.UseWhen[0] != "需要拿到当前用户的主日历 ID" {
+		t.Errorf("UseWhen = %v", a.UseWhen)
+	}
+	if len(a.Examples) != 1 || a.Examples[0].Description != "获取主日历" || a.Examples[0].Command != "lark-cli calendar calendars primary" {
+		t.Errorf("Examples = %+v", a.Examples)
+	}
+	if len(a.Related) != 1 || a.Related[0] != "calendars.list" {
+		t.Errorf("Related = %v", a.Related)
+	}
+}

internal/meta/identity.go

@@ -0,0 +1,94 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import "sort"
+
+// Token is the metadata accessTokens vocabulary: which token kind a method
+// accepts. It is a distinct type so the two directions of the token<->identity
+// mapping below cannot be swapped silently — a bare string compiles on either
+// side of a string/string signature, a Token does not. The CLI identity
+// vocabulary ("bot"/"user") already has a home in internal/core (core.Identity);
+// meta is a leaf and must not import core, so the identity side stays a plain
+// string here and is typed at the core boundary.
+type Token string
+
+const (
+	TokenTenant Token = "tenant" // bot calls use tenant_access_token
+	TokenUser   Token = "user"
+)
+
+// IdentityForToken maps a metadata access token to the CLI identity (--as
+// value) that uses it: tenant -> "bot", user -> "user". ok is false for
+// unrecognized tokens. This is the single source of truth for the
+// token<->identity vocabulary; schema, registry and command code all go
+// through it instead of re-spelling the mapping.
+func IdentityForToken(token Token) (string, bool) {
+	switch token {
+	case TokenTenant:
+		return "bot", true
+	case TokenUser:
+		return "user", true
+	}
+	return "", false
+}
+
+// TokenForIdentity is the inverse of IdentityForToken: "bot" -> TokenTenant;
+// everything else (notably "user") maps to itself.
+func TokenForIdentity(identity string) Token {
+	if identity == "bot" {
+		return TokenTenant
+	}
+	return Token(identity)
+}
+
+// RestrictsIdentity reports whether the method limits which identities may call
+// it: true exactly when it declares one or more accessTokens. nil OR an empty
+// slice means unrestricted (any identity). This is the single rule that both
+// the strict-mode predicate (SupportsToken) and command identity gates use, so
+// nil and [] never diverge across schema/scope and execution.
+func (m Method) RestrictsIdentity() bool {
+	return len(m.AccessTokens) > 0
+}
+
+// SupportsToken reports whether this method is reachable with the given access
+// token (see TokenForIdentity). An unrestricted method (RestrictsIdentity ==
+// false, i.e. nil or empty accessTokens) is reachable by any token. This is
+// the single source of truth for the predicate; registry scope policy and
+// command identity checks build on it.
+func (m Method) SupportsToken(token Token) bool {
+	if !m.RestrictsIdentity() {
+		return true
+	}
+	for _, t := range m.AccessTokens {
+		if t == token {
+			return true
+		}
+	}
+	return false
+}
+
+// Identities returns the CLI identities (--as values) that can call this
+// method, derived from its metadata accessTokens: tenant -> "bot", user
+// stays "user"; unrecognized tokens are dropped; the result is deduped and
+// name-sorted. The slice is always non-nil so callers rendering it (e.g. the
+// envelope's access_tokens) emit [] rather than null.
+//
+// An empty result does NOT imply unrestricted — use RestrictsIdentity() for
+// that. Identities() lists only CLI-known identities, so a method restricted
+// solely to unrecognized tokens returns empty yet RestrictsIdentity() is true.
+func (m Method) Identities() []string {
+	seen := make(map[string]bool, len(m.AccessTokens))
+	for _, t := range m.AccessTokens {
+		if id, ok := IdentityForToken(t); ok {
+			seen[id] = true
+		}
+	}
+	out := make([]string, 0, len(seen))
+	for id := range seen {
+		out = append(out, id)
+	}
+	sort.Strings(out)
+	return out
+}

internal/meta/identity_test.go

@@ -0,0 +1,85 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestIdentityTokenBijection(t *testing.T) {
+	if got := TokenForIdentity("bot"); got != "tenant" {
+		t.Errorf("TokenForIdentity(bot) = %q, want tenant", got)
+	}
+	if got := TokenForIdentity("user"); got != "user" {
+		t.Errorf("TokenForIdentity(user) = %q, want user", got)
+	}
+	if id, ok := IdentityForToken("tenant"); id != "bot" || !ok {
+		t.Errorf("IdentityForToken(tenant) = %q,%v want bot,true", id, ok)
+	}
+	if id, ok := IdentityForToken("user"); id != "user" || !ok {
+		t.Errorf("IdentityForToken(user) = %q,%v want user,true", id, ok)
+	}
+	if _, ok := IdentityForToken("weird"); ok {
+		t.Error("IdentityForToken(weird) ok=true, want false")
+	}
+}
+
+func TestMethod_RestrictsIdentity(t *testing.T) {
+	// nil and empty both mean "unrestricted"; only a populated list restricts.
+	if (Method{}).RestrictsIdentity() {
+		t.Error("nil accessTokens must be unrestricted")
+	}
+	if (Method{AccessTokens: []Token{}}).RestrictsIdentity() {
+		t.Error("empty accessTokens must be unrestricted (same as nil)")
+	}
+	if !(Method{AccessTokens: []Token{"tenant"}}).RestrictsIdentity() {
+		t.Error("populated accessTokens must restrict identity")
+	}
+}
+
+func TestMethod_SupportsToken(t *testing.T) {
+	// unrestricted (nil OR empty) -> permissive for any token; the two must not
+	// diverge, else strict/scope and the command gate disagree.
+	for _, m := range []Method{{}, {AccessTokens: []Token{}}} {
+		if !m.SupportsToken("tenant") || !m.SupportsToken("user") {
+			t.Errorf("unrestricted method %#v should support any token", m.AccessTokens)
+		}
+	}
+	// restricted: only the declared tokens are reachable
+	m := Method{AccessTokens: []Token{"tenant"}}
+	if !m.SupportsToken("tenant") {
+		t.Error("tenant-declared method should support tenant")
+	}
+	if m.SupportsToken("user") {
+		t.Error("tenant-only method must NOT support user")
+	}
+}
+
+func TestMethod_Identities(t *testing.T) {
+	// tenant->bot, user stays; deduped + name-sorted (so order-independent);
+	// unrecognized dropped; absent tokens -> empty but NON-nil so the envelope
+	// renders [] not null.
+	tests := []struct {
+		name   string
+		tokens []Token
+		want   []string
+	}{
+		{"tenant only", []Token{"tenant"}, []string{"bot"}},
+		{"user only", []Token{"user"}, []string{"user"}},
+		{"tenant then user", []Token{"tenant", "user"}, []string{"bot", "user"}},
+		{"user then tenant", []Token{"user", "tenant"}, []string{"bot", "user"}},
+		{"deduped", []Token{"tenant", "tenant", "user"}, []string{"bot", "user"}},
+		{"empty", []Token{}, []string{}},
+		{"nil", nil, []string{}},
+		{"unknown skipped", []Token{"user", "admin"}, []string{"user"}},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := (Method{AccessTokens: tt.tokens}).Identities(); !reflect.DeepEqual(got, tt.want) {
+				t.Errorf("Identities(%v) = %#v, want %#v", tt.tokens, got, tt.want)
+			}
+		})
+	}
+}

internal/meta/meta.go

@@ -0,0 +1,215 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+// Package meta is the typed model of the API metadata registry and the single
+// place that parses it. The metadata is a fixed, regular vocabulary, so a plain
+// typed json.Unmarshal replaces hand-rolled map[string]interface{} walking. Map
+// key order is not preserved (Go maps are unordered); callers that need a
+// deterministic sequence get fields/methods/resources sorted by name via the
+// list accessors below.
+package meta
+
+import (
+	"encoding/json"
+	"sort"
+	"strings"
+)
+
+// Option is one enum option of a field. Value is `any` (not string) so a
+// metadata value that arrives as a JSON number — rather than the usual quoted
+// string — coerces like Field.Enum / EnumOption.Value instead of failing the
+// whole registry unmarshal and blanking the entire catalog. coerceLiteral
+// normalizes it to the field's declared type.
+type Option struct {
+	Value       any    `json:"value"`
+	Description string `json:"description"`
+}
+
+// Field is one parameter or body/response field. Name is the parent map key,
+// populated by the list accessors (not a JSON field). ref/annotations/enumName
+// exist in the metadata but are intentionally not modeled (unused downstream).
+type Field struct {
+	Name        string           `json:"-"`
+	Type        string           `json:"type"`
+	Location    string           `json:"location"` // "path" | "query"; empty for body/response
+	Required    bool             `json:"required"`
+	Description string           `json:"description"`
+	Default     any              `json:"default"`
+	Example     any              `json:"example"`
+	Min         string           `json:"min"`
+	Max         string           `json:"max"`
+	Enum        []any            `json:"enum"`
+	Options     []Option         `json:"options"`
+	Properties  map[string]Field `json:"properties"`
+}
+
+// FlagName is the kebab-case CLI flag for this field (chat_id -> chat-id).
+func (f Field) FlagName() string { return strings.ReplaceAll(f.Name, "_", "-") }
+
+// Children returns the field's nested properties sorted by name.
+func (f Field) Children() []Field { return fieldsOf(f.Properties, nil) }
+
+// Method is one API operation. Name is the parent map key. Affordance is kept
+// raw so this package stays free of envelope concerns.
+type Method struct {
+	Name           string           `json:"-"`
+	ID             string           `json:"id"`
+	Path           string           `json:"path"`
+	HTTPMethod     string           `json:"httpMethod"`
+	Description    string           `json:"description"`
+	Risk           string           `json:"risk"`
+	DocURL         string           `json:"docUrl"`
+	Danger         bool             `json:"danger"`
+	Tips           []string         `json:"tips"`
+	Scopes         []string         `json:"scopes"`
+	RequiredScopes []string         `json:"requiredScopes"`
+	AccessTokens   []Token          `json:"accessTokens"`
+	Affordance     json.RawMessage  `json:"affordance"`
+	Parameters     map[string]Field `json:"parameters"`
+	RequestBody    map[string]Field `json:"requestBody"`
+	ResponseBody   map[string]Field `json:"responseBody"`
+}
+
+// Params are the path/query parameters, sorted by name.
+func (m Method) Params() []Field {
+	return fieldsOf(m.Parameters, func(f Field) bool {
+		return f.Location == "path" || f.Location == "query"
+	})
+}
+
+// Data are the non-file request-body fields (--data JSON), sorted by name.
+func (m Method) Data() []Field {
+	return fieldsOf(m.RequestBody, func(f Field) bool { return f.Type != "file" })
+}
+
+// Files are the file-typed request-body fields (--file uploads), sorted by name.
+func (m Method) Files() []Field {
+	return fieldsOf(m.RequestBody, func(f Field) bool { return f.Type == "file" })
+}
+
+// Response are the response-body fields, sorted by name.
+func (m Method) Response() []Field { return fieldsOf(m.ResponseBody, nil) }
+
+// fieldsOf materializes a name->field map into a name-injected slice, optionally
+// filtered, sorted by name for deterministic output.
+func fieldsOf(byName map[string]Field, keep func(Field) bool) []Field {
+	out := make([]Field, 0, len(byName))
+	for name, f := range byName {
+		f.Name = name
+		if keep == nil || keep(f) {
+			out = append(out, f)
+		}
+	}
+	sort.Slice(out, func(i, j int) bool { return out[i].Name < out[j].Name })
+	return out
+}
+
+// Resource groups methods (and may nest sub-resources). Name is the parent key.
+type Resource struct {
+	Name      string              `json:"-"`
+	Methods   map[string]Method   `json:"methods"`
+	Resources map[string]Resource `json:"resources"`
+}
+
+// MethodList returns the resource's methods, name-injected and sorted by name.
+func (r Resource) MethodList() []Method {
+	out := make([]Method, 0, len(r.Methods))
+	for name, m := range r.Methods {
+		m.Name = name
+		out = append(out, m)
+	}
+	sort.Slice(out, func(i, j int) bool { return out[i].Name < out[j].Name })
+	return out
+}
+
+// Method looks up one method by name with Name injected, or false if absent.
+// Use this instead of indexing Methods directly so Name is never left empty.
+func (r Resource) Method(name string) (Method, bool) {
+	m, ok := r.Methods[name]
+	if !ok {
+		return Method{}, false
+	}
+	m.Name = name
+	return m, true
+}
+
+// SubResources returns nested resources, name-injected and sorted by name.
+func (r Resource) SubResources() []Resource { return resourcesOf(r.Resources) }
+
+// Service is one API service. Name is a real JSON field (services is an array).
+type Service struct {
+	Name        string              `json:"name"`
+	Version     string              `json:"version"`
+	Title       string              `json:"title"`
+	Description string              `json:"description"`
+	ServicePath string              `json:"servicePath"`
+	Resources   map[string]Resource `json:"resources"`
+}
+
+// ResourceList returns the service's top-level resources, name-injected and
+// sorted by name.
+func (s Service) ResourceList() []Resource { return resourcesOf(s.Resources) }
+
+// Resource looks up one (possibly dotted) resource by name with Name injected,
+// or false if absent. Use this instead of indexing Resources directly.
+func (s Service) Resource(name string) (Resource, bool) {
+	r, ok := s.Resources[name]
+	if !ok {
+		return Resource{}, false
+	}
+	r.Name = name
+	return r, true
+}
+
+func resourcesOf(byName map[string]Resource) []Resource {
+	out := make([]Resource, 0, len(byName))
+	for name, r := range byName {
+		r.Name = name
+		out = append(out, r)
+	}
+	sort.Slice(out, func(i, j int) bool { return out[i].Name < out[j].Name })
+	return out
+}
+
+// Registry is the top-level metadata document.
+type Registry struct {
+	Services []Service `json:"services"`
+	Version  string    `json:"version"`
+}
+
+// Parse decodes the metadata JSON into the typed Registry. Returns a zero
+// Registry for empty input.
+func Parse(data []byte) (Registry, error) {
+	if len(data) == 0 {
+		return Registry{}, nil
+	}
+	var reg Registry
+	if err := json.Unmarshal(data, &reg); err != nil {
+		return Registry{}, err
+	}
+	return reg, nil
+}
+
+// FromMap decodes a single method spec from its map form into a typed Method.
+// Convenience constructor for building typed values from map literals (tests).
+func FromMap(method map[string]interface{}) Method {
+	b, err := json.Marshal(method)
+	if err != nil {
+		return Method{}
+	}
+	var m Method
+	_ = json.Unmarshal(b, &m)
+	return m
+}
+
+// ServiceFromMap decodes a service spec from its map form into a typed Service.
+// Convenience constructor for building typed values from map literals (tests).
+func ServiceFromMap(svc map[string]interface{}) Service {
+	b, err := json.Marshal(svc)
+	if err != nil {
+		return Service{}
+	}
+	var s Service
+	_ = json.Unmarshal(b, &s)
+	return s
+}

internal/meta/meta_test.go

@@ -0,0 +1,140 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import (
+	"reflect"
+	"testing"
+)
+
+const sampleJSON = `{
+  "version": "1.0.0",
+  "services": [
+    {
+      "name": "im",
+      "servicePath": "/open-apis/im/v1",
+      "resources": {
+        "chat.members": {
+          "methods": {
+            "create": {
+              "httpMethod": "POST",
+              "risk": "high-risk-write",
+              "parameters": {
+                "member_id_type": {"type": "string", "location": "query", "options": [{"value": "open_id"}, {"value": "user_id"}]},
+                "chat_id":        {"type": "string", "location": "path", "required": true, "example": "oc_x"},
+                "x_header":       {"type": "string", "location": "header"}
+              },
+              "requestBody": {
+                "id_list": {"type": "list", "required": true},
+                "avatar":  {"type": "file"}
+              }
+            }
+          }
+        }
+      }
+    }
+  ]
+}`
+
+func loadSample(t *testing.T) (Resource, Method) {
+	t.Helper()
+	reg, err := Parse([]byte(sampleJSON))
+	if err != nil {
+		t.Fatalf("Parse: %v", err)
+	}
+	res := reg.Services[0].ResourceList()
+	if len(res) != 1 {
+		t.Fatalf("want 1 resource, got %d", len(res))
+	}
+	methods := res[0].MethodList()
+	if len(methods) != 1 {
+		t.Fatalf("want 1 method, got %d", len(methods))
+	}
+	return res[0], methods[0]
+}
+
+func TestParse_TypedAndNameInjected(t *testing.T) {
+	res, m := loadSample(t)
+	if res.Name != "chat.members" {
+		t.Errorf("resource name = %q, want chat.members", res.Name)
+	}
+	if m.Name != "create" || m.HTTPMethod != "POST" || m.Risk != "high-risk-write" {
+		t.Errorf("method = %+v", m)
+	}
+}
+
+func TestMethod_AccessorsSortedByName(t *testing.T) {
+	_, m := loadSample(t)
+
+	// Params: path/query only (header dropped), sorted by name.
+	var params []string
+	for _, f := range m.Params() {
+		params = append(params, f.Name)
+	}
+	if want := []string{"chat_id", "member_id_type"}; !reflect.DeepEqual(params, want) {
+		t.Errorf("Params() = %v, want %v (sorted, header dropped)", params, want)
+	}
+
+	if d := m.Data(); len(d) != 1 || d[0].Name != "id_list" {
+		t.Errorf("Data() = %+v, want [id_list]", d)
+	}
+	if f := m.Files(); len(f) != 1 || f[0].Name != "avatar" {
+		t.Errorf("Files() = %+v, want [avatar]", f)
+	}
+}
+
+func TestField_FlagNameAndOptions(t *testing.T) {
+	_, m := loadSample(t)
+	by := make(map[string]Field)
+	for _, f := range m.Params() {
+		by[f.Name] = f
+	}
+
+	if got := by["chat_id"].FlagName(); got != "chat-id" {
+		t.Errorf("FlagName = %q, want chat-id", got)
+	}
+	if !by["chat_id"].Required || by["chat_id"].Example != "oc_x" {
+		t.Errorf("chat_id required/example wrong: %+v", by["chat_id"])
+	}
+	opts := by["member_id_type"].Options
+	if len(opts) != 2 || opts[0].Value != "open_id" || opts[1].Value != "user_id" {
+		t.Errorf("member_id_type options = %+v", opts)
+	}
+}
+
+// TestParse_TolerantOptionValue guards against whole-catalog blanking: a single
+// options[].value that arrives as a JSON number (not the usual quoted string)
+// must NOT fail the entire registry unmarshal. Option.Value is `any`, so it
+// parses and coerces like Enum instead of returning an empty Registry.
+func TestParse_TolerantOptionValue(t *testing.T) {
+	data := []byte(`{"services":[{"name":"im","servicePath":"/x","resources":{
+		"chat":{"methods":{"create":{"parameters":{
+			"flag":{"type":"integer","location":"query","options":[{"value":0,"description":"off"},{"value":1,"description":"on"}]}
+		}}}}}}]}`)
+	reg, err := Parse(data)
+	if err != nil {
+		t.Fatalf("Parse failed on numeric option value (would blank the catalog): %v", err)
+	}
+	if len(reg.Services) != 1 {
+		t.Fatalf("expected 1 service, got %d (catalog blanked)", len(reg.Services))
+	}
+	// The numeric option coerces into the typed enum as sorted int64 (not
+	// float64): the integer field's canonical type drives normalization.
+	m, _ := reg.Services[0].Resource("chat")
+	method, _ := m.Method("create")
+	by := map[string]Field{}
+	for _, f := range method.Params() {
+		by[f.Name] = f
+	}
+	if got := by["flag"].EnumValues(); !reflect.DeepEqual(got, []any{int64(0), int64(1)}) {
+		t.Errorf("numeric-valued enum did not coerce to sorted int64: %#v", got)
+	}
+}
+
+func TestParse_Empty(t *testing.T) {
+	reg, err := Parse(nil)
+	if err != nil || len(reg.Services) != 0 {
+		t.Fatalf("Parse(nil) = %+v, %v", reg, err)
+	}
+}

internal/meta/normalize.go

@@ -0,0 +1,202 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import (
+	"fmt"
+	"sort"
+	"strconv"
+)
+
+// CanonicalType maps meta_data's non-standard type names to the standard
+// JSON-Schema/type vocabulary used downstream (envelope render, flag kinds):
+// "file" -> "string", "list" -> "array"; other types pass through unchanged.
+func (f Field) CanonicalType() string {
+	switch f.Type {
+	case "file":
+		return "string"
+	case "list":
+		return "array"
+	default:
+		return f.Type
+	}
+}
+
+// coerceLiteral converts a meta_data literal (default/enum/example) to the
+// field's canonical type. Literals may arrive as strings (meta_data's usual
+// form) OR already typed — a JSON number unmarshals to float64, a JSON bool to
+// bool — so both must be normalized to the SAME Go type the canonical type
+// implies (int64 for "integer", float64 for "number", bool for "boolean").
+// Otherwise enumLess, which type-asserts on that Go type, can't order the
+// values. Returns (value, true) on success, (nil, false) when the literal
+// cannot be represented in the declared type.
+func coerceLiteral(canonicalType string, raw any) (any, bool) {
+	switch canonicalType {
+	case "integer":
+		switch v := raw.(type) {
+		case string:
+			if n, err := strconv.ParseInt(v, 10, 64); err == nil {
+				return n, true
+			}
+		case float64: // JSON number; accept only when it's a whole value
+			if v == float64(int64(v)) {
+				return int64(v), true
+			}
+		case int64:
+			return v, true
+		case int:
+			return int64(v), true
+		}
+		return nil, false
+	case "number":
+		switch v := raw.(type) {
+		case string:
+			if n, err := strconv.ParseFloat(v, 64); err == nil {
+				return n, true
+			}
+		case float64:
+			return v, true
+		case int64:
+			return float64(v), true
+		case int:
+			return float64(v), true
+		}
+		return nil, false
+	case "boolean":
+		switch v := raw.(type) {
+		case string:
+			switch v {
+			case "true":
+				return true, true
+			case "false":
+				return false, true
+			}
+		case bool:
+			return v, true
+		}
+		return nil, false
+	default: // "string", "array", "" (objects), or unknown — pass through as-is
+		return raw, true
+	}
+}
+
+// enumLess orders two coerced enum values for the canonical type, so integer
+// enums end up [1 2 10] not lexicographic [1 10 2].
+func enumLess(canonicalType string, a, b any) bool {
+	switch canonicalType {
+	case "integer":
+		ai, _ := a.(int64)
+		bi, _ := b.(int64)
+		return ai < bi
+	case "number":
+		af, _ := a.(float64)
+		bf, _ := b.(float64)
+		return af < bf
+	case "boolean":
+		ab, _ := a.(bool)
+		bb, _ := b.(bool)
+		return !ab && bb
+	default:
+		as, _ := a.(string)
+		bs, _ := b.(string)
+		return as < bs
+	}
+}
+
+// EnumOption is one allowed value paired with its human description. The
+// description comes from options[].description and is empty for the bare `enum`
+// form (which carries no descriptions).
+type EnumOption struct {
+	Value       any
+	Description string
+}
+
+// EnumOptions returns the field's allowed values paired with their descriptions
+// — from enum, or from options when enum is absent — coerced to the canonical
+// type and ordered: numeric and boolean values are sorted; string values keep
+// source order (which can encode priority). Uncoercible literals are dropped.
+// Returns nil when the field declares no enum constraint.
+func (f Field) EnumOptions() []EnumOption {
+	ct := f.CanonicalType()
+	var out []EnumOption
+	switch {
+	case len(f.Enum) > 0:
+		for _, e := range f.Enum {
+			if v, ok := coerceLiteral(ct, e); ok {
+				out = append(out, EnumOption{Value: v})
+			}
+		}
+	case len(f.Options) > 0:
+		seen := make(map[string]bool)
+		for _, o := range f.Options {
+			key := fmt.Sprintf("%v", o.Value)
+			if seen[key] {
+				continue
+			}
+			seen[key] = true
+			if v, ok := coerceLiteral(ct, o.Value); ok {
+				out = append(out, EnumOption{Value: v, Description: o.Description})
+			}
+		}
+	}
+	if len(out) > 0 && ct != "string" && ct != "" {
+		sort.SliceStable(out, func(i, j int) bool { return enumLess(ct, out[i].Value, out[j].Value) })
+	}
+	return out
+}
+
+// EnumValues returns the field's allowed values — the value projection of
+// EnumOptions, in the same order. nil when the field declares no enum
+// constraint. (Kept as the values-only accessor for the envelope and flag
+// completion, which don't need descriptions.)
+func (f Field) EnumValues() []any {
+	opts := f.EnumOptions()
+	if len(opts) == 0 {
+		return nil
+	}
+	out := make([]any, len(opts))
+	for i, o := range opts {
+		out[i] = o.Value
+	}
+	return out
+}
+
+// CoercedDefault returns Default coerced to the canonical type, or nil when the
+// field has no default or the literal cannot be coerced.
+func (f Field) CoercedDefault() any { return f.coerce(f.Default) }
+
+// CoercedExample returns Example coerced to the canonical type, or nil when the
+// field has no example or the literal cannot be coerced.
+func (f Field) CoercedExample() any { return f.coerce(f.Example) }
+
+func (f Field) coerce(raw any) any {
+	if raw == nil {
+		return nil
+	}
+	if v, ok := coerceLiteral(f.CanonicalType(), raw); ok {
+		return v
+	}
+	return nil
+}
+
+// MinBound returns the field's min constraint as a number, or nil when absent
+// or unparseable. meta_data carries min/max as strings and does not say
+// whether they bound a value or a string's length; the accessors stay equally
+// agnostic, so every renderer (envelope minimum/maximum, flag help) presents
+// the same numbers without inventing a semantic the source doesn't declare.
+func (f Field) MinBound() *float64 { return parseBound(f.Min) }
+
+// MaxBound returns the field's max constraint as a number, or nil when absent
+// or unparseable. See MinBound.
+func (f Field) MaxBound() *float64 { return parseBound(f.Max) }
+
+func parseBound(s string) *float64 {
+	if s == "" {
+		return nil
+	}
+	if v, err := strconv.ParseFloat(s, 64); err == nil {
+		return &v
+	}
+	return nil
+}

internal/meta/normalize_test.go

@@ -0,0 +1,160 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package meta
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestField_CanonicalType(t *testing.T) {
+	cases := map[string]string{
+		"file":    "string", // meta_data's non-standard "file" is a string with binary format
+		"list":    "array",  // "list" is meta_data's spelling of a JSON array
+		"integer": "integer",
+		"boolean": "boolean",
+		"string":  "string",
+		"":        "",
+	}
+	for in, want := range cases {
+		if got := (Field{Type: in}).CanonicalType(); got != want {
+			t.Errorf("CanonicalType(%q) = %q, want %q", in, got, want)
+		}
+	}
+}
+
+func TestField_EnumValues(t *testing.T) {
+	// string enum keeps source order (order can encode priority)
+	if got := (Field{Type: "string", Enum: []any{"b", "a", "c"}}).EnumValues(); !reflect.DeepEqual(got, []any{"b", "a", "c"}) {
+		t.Errorf("string enum = %v, want source order [b a c]", got)
+	}
+	// integer enum: string-stored literals coerced to int64 and numerically sorted
+	if got := (Field{Type: "integer", Enum: []any{"10", "2", "1"}}).EnumValues(); !reflect.DeepEqual(got, []any{int64(1), int64(2), int64(10)}) {
+		t.Errorf("integer enum = %v, want [1 2 10] coerced+sorted", got)
+	}
+	// options used when enum absent, deduped, source order for strings
+	if got := (Field{Type: "string", Options: []Option{{Value: "x"}, {Value: "x"}, {Value: "y"}}}).EnumValues(); !reflect.DeepEqual(got, []any{"x", "y"}) {
+		t.Errorf("options enum = %v, want [x y] deduped", got)
+	}
+	// uncoercible literal dropped
+	if got := (Field{Type: "integer", Enum: []any{"1", "nope", "2"}}).EnumValues(); !reflect.DeepEqual(got, []any{int64(1), int64(2)}) {
+		t.Errorf("bad enum = %v, want [1 2] (nope dropped)", got)
+	}
+	// no enum/options -> nil
+	if got := (Field{Type: "string"}).EnumValues(); got != nil {
+		t.Errorf("empty enum = %v, want nil", got)
+	}
+}
+
+func TestField_EnumOptions(t *testing.T) {
+	// options carry descriptions, kept paired with their (string) value in source order
+	fo := Field{Type: "string", Options: []Option{
+		{Value: "open_id", Description: "以 open_id 标识"},
+		{Value: "open_id", Description: "dup ignored"}, // dedup keeps first
+		{Value: "user_id", Description: "以 user_id 标识"},
+	}}
+	got := fo.EnumOptions()
+	want := []EnumOption{
+		{Value: "open_id", Description: "以 open_id 标识"},
+		{Value: "user_id", Description: "以 user_id 标识"},
+	}
+	if !reflect.DeepEqual(got, want) {
+		t.Errorf("EnumOptions = %+v, want %+v", got, want)
+	}
+
+	// integer enum (bare form): values coerced + numerically sorted, no descriptions
+	fi := Field{Type: "integer", Enum: []any{"10", "2", "1"}}
+	gi := fi.EnumOptions()
+	if len(gi) != 3 || gi[0].Value != int64(1) || gi[2].Value != int64(10) || gi[0].Description != "" {
+		t.Errorf("EnumOptions(integer) = %+v, want [1 2 10] coerced+sorted, no desc", gi)
+	}
+
+	// EnumValues stays the value projection of EnumOptions (golden-critical)
+	if !reflect.DeepEqual(fo.EnumValues(), []any{"open_id", "user_id"}) {
+		t.Errorf("EnumValues diverged from EnumOptions values: %v", fo.EnumValues())
+	}
+	// unconstrained -> nil
+	if (Field{Type: "string"}).EnumOptions() != nil {
+		t.Error("EnumOptions should be nil when unconstrained")
+	}
+}
+
+func TestField_Enum_NumberAndBoolean(t *testing.T) {
+	// number: string-stored floats coerced to float64 and numerically sorted
+	if got := (Field{Type: "number", Enum: []any{"2.5", "1.5", "10"}}).EnumValues(); !reflect.DeepEqual(got, []any{1.5, 2.5, float64(10)}) {
+		t.Errorf("number enum = %v, want [1.5 2.5 10] coerced+sorted", got)
+	}
+	// number: uncoercible literal dropped
+	if got := (Field{Type: "number", Enum: []any{"1.5", "x", "2.5"}}).EnumValues(); !reflect.DeepEqual(got, []any{1.5, 2.5}) {
+		t.Errorf("number enum with bad value = %v, want [1.5 2.5]", got)
+	}
+	// boolean: true/false coerced and sorted (false before true); invalid dropped
+	if got := (Field{Type: "boolean", Enum: []any{"true", "maybe", "false"}}).EnumValues(); !reflect.DeepEqual(got, []any{false, true}) {
+		t.Errorf("boolean enum = %v, want [false true]", got)
+	}
+}
+
+func TestField_EnumOptions_NonStringValuesNormalized(t *testing.T) {
+	// JSON numbers/bools arrive already-typed (a number is float64, not a
+	// string) — e.g. options[].value: 0. They must still be normalized to the
+	// field's canonical type (int64 for "integer") and sorted numerically;
+	// leaving them as float64 both yields the wrong type and defeats enumLess,
+	// whose integer branch asserts int64 and would otherwise treat every value
+	// as zero (no sort).
+	if got := (Field{Type: "integer", Options: []Option{
+		{Value: float64(10)}, {Value: float64(2)}, {Value: float64(1)},
+	}}).EnumValues(); !reflect.DeepEqual(got, []any{int64(1), int64(2), int64(10)}) {
+		t.Errorf("integer options from float64 = %#v, want [int64(1) int64(2) int64(10)]", got)
+	}
+	// bare enum form, JSON numbers
+	if got := (Field{Type: "integer", Enum: []any{float64(3), float64(1), float64(2)}}).EnumValues(); !reflect.DeepEqual(got, []any{int64(1), int64(2), int64(3)}) {
+		t.Errorf("integer enum from float64 = %#v, want [int64(1) int64(2) int64(3)]", got)
+	}
+	// number field: a whole-valued float stays float64
+	if got := (Field{Type: "number", Enum: []any{float64(2), float64(1)}}).EnumValues(); !reflect.DeepEqual(got, []any{float64(1), float64(2)}) {
+		t.Errorf("number enum from float64 = %#v, want [float64(1) float64(2)]", got)
+	}
+	// boolean field: native bools coerce + sort (false < true)
+	if got := (Field{Type: "boolean", Enum: []any{true, false}}).EnumValues(); !reflect.DeepEqual(got, []any{false, true}) {
+		t.Errorf("boolean enum from bool = %#v, want [false true]", got)
+	}
+	// non-integral float under integer is uncoercible -> dropped (mirrors how "2.5" fails ParseInt)
+	if got := (Field{Type: "integer", Enum: []any{float64(1), float64(2.5), float64(3)}}).EnumValues(); !reflect.DeepEqual(got, []any{int64(1), int64(3)}) {
+		t.Errorf("integer enum with fractional float = %#v, want [int64(1) int64(3)]", got)
+	}
+}
+
+func TestField_CoercedDefaultAndExample(t *testing.T) {
+	if got := (Field{Type: "integer", Default: "5"}).CoercedDefault(); got != int64(5) {
+		t.Errorf("CoercedDefault integer = %v (%T), want int64(5)", got, got)
+	}
+	if got := (Field{Type: "integer", Default: "bad"}).CoercedDefault(); got != nil {
+		t.Errorf("CoercedDefault uncoercible = %v, want nil", got)
+	}
+	if got := (Field{Type: "string"}).CoercedDefault(); got != nil {
+		t.Errorf("CoercedDefault absent = %v, want nil", got)
+	}
+	if got := (Field{Type: "boolean", Example: "true"}).CoercedExample(); got != true {
+		t.Errorf("CoercedExample boolean = %v, want true", got)
+	}
+}
+
+func TestField_Bounds(t *testing.T) {
+	f := Field{Min: "1", Max: "100"}
+	if v := f.MinBound(); v == nil || *v != 1 {
+		t.Errorf("MinBound = %v, want 1", v)
+	}
+	if v := f.MaxBound(); v == nil || *v != 100 {
+		t.Errorf("MaxBound = %v, want 100", v)
+	}
+	if v := (Field{Min: "0.5"}).MinBound(); v == nil || *v != 0.5 {
+		t.Errorf("MinBound fractional = %v, want 0.5", v)
+	}
+	if v := (Field{}).MinBound(); v != nil {
+		t.Errorf("MinBound absent = %v, want nil", v)
+	}
+	if v := (Field{Max: "not_a_number"}).MaxBound(); v != nil {
+		t.Errorf("MaxBound unparseable = %v, want nil", v)
+	}
+}

internal/registry/catalog.go

@@ -0,0 +1,20 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package registry
+
+import "github.com/larksuite/cli/internal/apicatalog"
+
+// EmbeddedCatalog returns a navigation catalog over the embedded (overlay-free)
+// metadata — deterministic across machines, for `lark-cli schema`, golden tests
+// and schema lint.
+func EmbeddedCatalog() apicatalog.Catalog {
+	return apicatalog.New(apicatalog.SourceEmbedded, EmbeddedServicesTyped())
+}
+
+// RuntimeCatalog returns a navigation catalog over the merged (embedded + remote
+// overlay) metadata — for service command registration and scope discovery,
+// where overlay methods must be reachable.
+func RuntimeCatalog() apicatalog.Catalog {
+	return apicatalog.New(apicatalog.SourceRuntime, ServicesTyped())
+}

internal/registry/helpers.go

@@ -3,54 +3,17 @@
 
 package registry
 
-// GetStrFromMap extracts a string value from map[string]interface{}.
-func GetStrFromMap(m map[string]interface{}, key string) string {
-	if m == nil {
-		return ""
-	}
-	if v, ok := m[key]; ok {
-		if s, ok := v.(string); ok {
-			return s
-		}
-	}
-	return ""
-}
+import "github.com/larksuite/cli/internal/meta"
 
-// GetStrSliceFromMap extracts a []string value from map[string]interface{}.
-// Returns nil if the key is missing or the value is not a string slice.
-func GetStrSliceFromMap(m map[string]interface{}, key string) []string {
-	if m == nil {
-		return nil
-	}
-	raw, ok := m[key].([]interface{})
-	if !ok {
-		return nil
-	}
-	result := make([]string, 0, len(raw))
-	for _, v := range raw {
-		if s, ok := v.(string); ok {
-			result = append(result, s)
-		}
-	}
-	if len(result) == 0 {
-		return nil
-	}
-	return result
-}
-
-// DeclaredScopesForMethod returns the scopes declared by a method's
-// from_meta entry for the given identity. Prefers the explicit
-// `requiredScopes` field when present; otherwise returns the single
-// recommended scope from `scopes` (or the first scope as a final fallback).
-// Returns nil when the method has no scope information.
-func DeclaredScopesForMethod(method map[string]interface{}, identity string) []string {
-	if method == nil {
-		return nil
-	}
-	if requiredRaw, ok := method["requiredScopes"].([]interface{}); ok && len(requiredRaw) > 0 {
-		out := make([]string, 0, len(requiredRaw))
-		for _, v := range requiredRaw {
-			if s, ok := v.(string); ok && s != "" {
+// DeclaredScopesForMethod returns the scopes declared by a method for the given
+// identity. Prefers the explicit `requiredScopes` field when present; otherwise
+// returns the single recommended scope from `scopes` (or the first scope as a
+// final fallback). Returns nil when the method has no scope information.
+func DeclaredScopesForMethod(m meta.Method, identity string) []string {
+	if len(m.RequiredScopes) > 0 {
+		out := make([]string, 0, len(m.RequiredScopes))
+		for _, s := range m.RequiredScopes {
+			if s != "" {
 				out = append(out, s)
 			}
 		}
@@ -58,54 +21,11 @@ func DeclaredScopesForMethod(method map[string]interface{}, identity string) []s
 			return out
 		}
 	}
-	rawScopes, _ := method["scopes"].([]interface{})
-	if len(rawScopes) == 0 {
+	if len(m.Scopes) == 0 {
 		return nil
 	}
-	recommended := SelectRecommendedScope(rawScopes, identity)
-	if recommended == "" {
-		for _, raw := range rawScopes {
-			if s, ok := raw.(string); ok && s != "" {
-				recommended = s
-				break
-			}
-		}
+	if recommended := SelectRecommendedScopeFromStrings(m.Scopes, identity); recommended != "" {
+		return []string{recommended}
 	}
-	if recommended == "" {
-		return nil
-	}
-	return []string{recommended}
-}
-
-// SelectRecommendedScope selects the known scope with the highest priority score
-// (higher = more recommended / least privilege).
-// Scopes not in the priority table are skipped to avoid recommending invalid/unknown scopes.
-func SelectRecommendedScope(scopes []interface{}, identity string) string {
-	priorities := LoadScopePriorities()
-	bestScore := -1
-	bestScope := ""
-	for _, s := range scopes {
-		str, ok := s.(string)
-		if !ok {
-			continue
-		}
-		score, exists := priorities[str]
-		if !exists {
-			continue // skip unknown scopes
-		}
-		if score > bestScore {
-			bestScore = score
-			bestScope = str
-		}
-	}
-	if bestScope != "" {
-		return bestScope
-	}
-	// Fallback: if no scope is in the priority table, return the first one.
-	if len(scopes) > 0 {
-		if s, ok := scopes[0].(string); ok {
-			return s
-		}
-	}
-	return ""
+	return nil
 }

internal/registry/loader.go

@@ -14,6 +14,7 @@ import (
 	"sync"
 
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/meta"
 )
 
 //go:embed scope_priorities.json scope_overrides.json
@@ -22,68 +23,42 @@ var registryFS embed.FS
 // embeddedMetaJSON is set by loader_embedded.go when meta_data.json is compiled in.
 var embeddedMetaJSON []byte
 
-// EmbeddedMetaJSON returns the raw embedded meta_data.json bytes for callers
-// that need to parse key order or other JSON-level structure not exposed by
-// LoadFromMeta (which loses map insertion order).
-func EmbeddedMetaJSON() []byte {
-	return embeddedMetaJSON
-}
-
 var (
-	embeddedServicesMap  map[string]map[string]interface{} // service name -> spec
-	embeddedServiceNames []string                          // sorted
-	embeddedParseOnce    sync.Once
+	embeddedServices       []meta.Service          // parsed once, sorted by name (no overlay)
+	embeddedServicesByName map[string]meta.Service // same, keyed by name
+	embeddedVersion        string                  // version from embedded meta_data.json
+	embeddedParseOnce      sync.Once
 )
 
-// parseEmbeddedServices parses embeddedMetaJSON into a service name → spec map
-// without touching mergedServices. Safe to call multiple times (sync.Once).
-func parseEmbeddedServices() {
+// parseEmbedded decodes the embedded meta_data.json into the typed model exactly
+// once. It is the single parse of the embedded bytes: both the overlay-free
+// envelope path (EmbeddedServicesTyped) and the merged command/scope path
+// (loadEmbeddedIntoMerged) build from this result, so the JSON is never parsed
+// twice and no map round-trip is needed downstream.
+func parseEmbedded() {
 	embeddedParseOnce.Do(func() {
-		embeddedServicesMap = make(map[string]map[string]interface{})
-		if len(embeddedMetaJSON) == 0 {
-			return
+		reg, _ := meta.Parse(embeddedMetaJSON)
+		embeddedVersion = reg.Version
+		embeddedServices = reg.Services
+		sort.Slice(embeddedServices, func(i, j int) bool { return embeddedServices[i].Name < embeddedServices[j].Name })
+		embeddedServicesByName = make(map[string]meta.Service, len(embeddedServices))
+		for _, svc := range embeddedServices {
+			embeddedServicesByName[svc.Name] = svc
 		}
-		var wrapper struct {
-			Services []map[string]interface{} `json:"services"`
-		}
-		if err := json.Unmarshal(embeddedMetaJSON, &wrapper); err != nil {
-			return
-		}
-		for _, svc := range wrapper.Services {
-			name, _ := svc["name"].(string)
-			if name == "" {
-				continue
-			}
-			embeddedServicesMap[name] = svc
-		}
-		embeddedServiceNames = make([]string, 0, len(embeddedServicesMap))
-		for name := range embeddedServicesMap {
-			embeddedServiceNames = append(embeddedServiceNames, name)
-		}
-		sort.Strings(embeddedServiceNames)
 	})
 }
 
-// EmbeddedSpec returns the embedded spec for one service, or nil if unknown.
-// Bypasses remote overlay — used for deterministic envelope output.
-func EmbeddedSpec(serviceName string) map[string]interface{} {
-	parseEmbeddedServices()
-	return embeddedServicesMap[serviceName]
-}
-
-// EmbeddedServiceNames returns sorted embedded service names (no overlay).
-// Returns a defensive copy — callers must not mutate the package-level slice.
-func EmbeddedServiceNames() []string {
-	parseEmbeddedServices()
-	out := make([]string, len(embeddedServiceNames))
-	copy(out, embeddedServiceNames)
-	return out
+// EmbeddedServicesTyped returns the embedded services (no remote overlay) as the
+// typed meta model, sorted by name. This is the overlay-free parse boundary the
+// schema envelope builds from — deterministic across machines.
+func EmbeddedServicesTyped() []meta.Service {
+	parseEmbedded()
+	return embeddedServices
 }
 
 var (
-	mergedServices    = make(map[string]map[string]interface{}) // project name → parsed spec
-	mergedProjectList []string                                  // sorted project names
-	embeddedVersion   string                                    // version from embedded meta_data.json
+	mergedServices    = make(map[string]meta.Service) // project name → typed service (embedded + overlay)
+	mergedProjectList []string                        // sorted project names
 	initOnce          sync.Once
 )
 
@@ -106,8 +81,8 @@ func InitWithBrand(brand core.LarkBrand) {
 		// 2. Remote overlay
 		if remoteEnabled() && cacheWritable() {
 			// Check if brand changed since last cache
-			meta, metaErr := loadCacheMeta()
-			brandChanged := metaErr == nil && meta.Brand != "" && meta.Brand != string(brand)
+			cm, metaErr := loadCacheMeta()
+			brandChanged := metaErr == nil && cm.Brand != "" && cm.Brand != string(brand)
 
 			if !brandChanged {
 				if cached, err := loadCachedMerged(); err == nil {
@@ -117,7 +92,7 @@ func InitWithBrand(brand core.LarkBrand) {
 			if len(mergedServices) == 0 || brandChanged {
 				// No data at all or brand changed — must sync fetch
 				doSyncFetch()
-			} else if shouldRefresh(meta) || metaErr != nil {
+			} else if shouldRefresh(cm) || metaErr != nil {
 				// Have embedded/cached data; refresh in background if TTL expired or first run
 				triggerBackgroundRefresh()
 			}
@@ -127,18 +102,13 @@ func InitWithBrand(brand core.LarkBrand) {
 	})
 }
 
-// loadEmbeddedIntoMerged parses the embedded meta_data.json and populates
-// mergedServices. No-op if meta_data.json is not compiled in.
+// loadEmbeddedIntoMerged seeds mergedServices from the embedded typed services
+// (the same parse EmbeddedServicesTyped uses). No-op if no services compiled in.
 func loadEmbeddedIntoMerged() {
-	if len(embeddedMetaJSON) == 0 {
-		return
+	parseEmbedded()
+	for name, svc := range embeddedServicesByName {
+		mergedServices[name] = svc
 	}
-	var reg MergedRegistry
-	if err := json.Unmarshal(embeddedMetaJSON, &reg); err != nil {
-		return
-	}
-	embeddedVersion = reg.Version
-	overlayMergedServices(&reg)
 }
 
 // rebuildProjectList rebuilds the sorted list of project names from mergedServices.
@@ -150,83 +120,32 @@ func rebuildProjectList() {
 	sort.Strings(mergedProjectList)
 }
 
-var cachedAllScopes map[string][]string
+var (
+	servicesTyped     []meta.Service
+	servicesTypedOnce sync.Once
+)
 
-// CollectAllScopesFromMeta collects all unique scopes from from_meta/*.json
-// for the given identity ("user" or "tenant"). Results are deduplicated and sorted.
-func CollectAllScopesFromMeta(identity string) []string {
-	if cachedAllScopes == nil {
-		cachedAllScopes = make(map[string][]string)
-	}
-	if cached, ok := cachedAllScopes[identity]; ok {
-		return cached
-	}
-
-	scopeSet := make(map[string]bool)
-	for _, project := range ListFromMetaProjects() {
-		spec := LoadFromMeta(project)
-		if spec == nil {
-			continue
+// ServicesTyped returns the merged registry (embedded + remote overlay) as typed
+// meta.Services, sorted by name. The merged store is already typed, so this just
+// projects it into a sorted slice — no map round-trip. This is the typed entry
+// the command tree and scope computation build from.
+func ServicesTyped() []meta.Service {
+	servicesTypedOnce.Do(func() {
+		Init()
+		servicesTyped = make([]meta.Service, 0, len(mergedProjectList))
+		for _, name := range mergedProjectList {
+			servicesTyped = append(servicesTyped, mergedServices[name])
 		}
-		resources, ok := spec["resources"].(map[string]interface{})
-		if !ok {
-			continue
-		}
-		for _, resSpec := range resources {
-			resMap, ok := resSpec.(map[string]interface{})
-			if !ok {
-				continue
-			}
-			methods, ok := resMap["methods"].(map[string]interface{})
-			if !ok {
-				continue
-			}
-			for _, methodSpec := range methods {
-				methodMap, ok := methodSpec.(map[string]interface{})
-				if !ok {
-					continue
-				}
-				// Check if method supports the requested identity
-				if tokens, ok := methodMap["accessTokens"].([]interface{}); ok {
-					supported := false
-					for _, t := range tokens {
-						if ts, ok := t.(string); ok && ts == IdentityToAccessToken(identity) {
-							supported = true
-							break
-						}
-					}
-					if !supported {
-						continue
-					}
-				}
-				// Collect scopes
-				scopes, ok := methodMap["scopes"].([]interface{})
-				if !ok {
-					continue
-				}
-				for _, s := range scopes {
-					if str, ok := s.(string); ok {
-						scopeSet[str] = true
-					}
-				}
-			}
-		}
-	}
-
-	result := make([]string, 0, len(scopeSet))
-	for s := range scopeSet {
-		result = append(result, s)
-	}
-	sort.Strings(result)
-	cachedAllScopes[identity] = result
-	return result
+	})
+	return servicesTyped
 }
 
-// LoadFromMeta loads a service schema by project name.
-// It returns data from the merged registry (embedded + cached remote overlay).
-func LoadFromMeta(project string) map[string]interface{} {
+// ServiceTyped returns one merged service (embedded + overlay) by name, or false
+// if unknown.
+func ServiceTyped(name string) (meta.Service, bool) {
 	Init()
-	return mergedServices[project]
+	svc, ok := mergedServices[name]
+	return svc, ok
 }
 
 // ListFromMetaProjects lists available service project names (sorted).

internal/registry/registry_test.go

@@ -65,8 +65,7 @@ func TestSelectRecommendedScope_PicksHighestScore(t *testing.T) {
 	}
 	t.Logf("%s=%d, %s=%d", scopeA, scoreA, scopeB, scoreB)
 
-	scopes := []interface{}{scopeB, scopeA}
-	result := SelectRecommendedScope(scopes, "user")
+	result := bestScope([]string{scopeB, scopeA}, priorities)
 
 	// Should pick the higher-scored one (higher = more recommended)
 	if scoreA > scoreB {
@@ -81,11 +80,11 @@ func TestSelectRecommendedScope_PicksHighestScore(t *testing.T) {
 }
 
 func TestSelectRecommendedScope_FallbackToFirst(t *testing.T) {
-	scopes := []interface{}{
+	scopes := []string{
 		"zzz_unknown:scope:a",
 		"zzz_unknown:scope:b",
 	}
-	result := SelectRecommendedScope(scopes, "user")
+	result := bestScope(scopes, LoadScopePriorities())
 	// All unknown scopes get DefaultScopeScore; first one with that score wins
 	if result != "zzz_unknown:scope:a" {
 		t.Errorf("expected zzz_unknown:scope:a, got %s", result)
@@ -93,13 +92,10 @@ func TestSelectRecommendedScope_FallbackToFirst(t *testing.T) {
 }
 
 func TestSelectRecommendedScope_Empty(t *testing.T) {
-	result := SelectRecommendedScope(nil, "user")
-	if result != "" {
+	if result := bestScope(nil, LoadScopePriorities()); result != "" {
 		t.Errorf("expected empty string, got %s", result)
 	}
-
-	result = SelectRecommendedScope([]interface{}{}, "user")
-	if result != "" {
+	if result := bestScope([]string{}, LoadScopePriorities()); result != "" {
 		t.Errorf("expected empty string, got %s", result)
 	}
 }
@@ -318,27 +314,6 @@ func TestFilterAutoApproveScopes_Empty(t *testing.T) {
 
 // --- Helper functions ---
 
-func TestGetStrFromMap(t *testing.T) {
-	m := map[string]interface{}{
-		"key1": "value1",
-		"key2": 42,
-		"key3": nil,
-	}
-
-	if v := GetStrFromMap(m, "key1"); v != "value1" {
-		t.Errorf("expected value1, got %s", v)
-	}
-	if v := GetStrFromMap(m, "key2"); v != "" {
-		t.Errorf("expected empty for non-string value, got %s", v)
-	}
-	if v := GetStrFromMap(m, "missing"); v != "" {
-		t.Errorf("expected empty for missing key, got %s", v)
-	}
-	if v := GetStrFromMap(nil, "key"); v != "" {
-		t.Errorf("expected empty for nil map, got %s", v)
-	}
-}
-
 func TestGetRegistryDir(t *testing.T) {
 	dir := GetRegistryDir()
 	if dir == "" {

internal/registry/remote.go

@@ -17,6 +17,7 @@ import (
 
 	"github.com/larksuite/cli/internal/build"
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/meta"
 	"github.com/larksuite/cli/internal/transport"
 	"github.com/larksuite/cli/internal/validate"
 	"github.com/larksuite/cli/internal/vfs"
@@ -35,16 +36,20 @@ type CacheMeta struct {
 	Brand       string `json:"brand,omitempty"`
 }
 
-// MergedRegistry is the top-level structure of remote_meta.json.
+// MergedRegistry is the top-level structure of remote_meta.json. Services are
+// decoded straight into the typed meta model so embedded, cached, and remote
+// data share one representation (no map intermediary, no re-parse).
 type MergedRegistry struct {
-	Version  string                   `json:"version"`
-	Services []map[string]interface{} `json:"services"`
+	Version  string         `json:"version"`
+	Services []meta.Service `json:"services"`
 }
 
-// remoteResponse is the envelope returned by the remote API.
+// remoteResponse is the envelope returned by the remote API. Data stays raw:
+// the cache must store the server payload verbatim — a typed re-marshal would
+// strip every field the model doesn't (yet) declare and starve future binaries.
 type remoteResponse struct {
-	Msg  string         `json:"msg"`
-	Data MergedRegistry `json:"data"`
+	Msg  string          `json:"msg"`
+	Data json.RawMessage `json:"data"`
 }
 
 // configuredBrand is set by InitWithBrand and determines which API host to use.
@@ -125,22 +130,22 @@ func cacheWritable() bool {
 // --- cache I/O ---
 
 func loadCacheMeta() (CacheMeta, error) {
-	var meta CacheMeta
+	var cm CacheMeta
 	data, err := vfs.ReadFile(cacheMetaPath())
 	if err != nil {
-		return meta, err
+		return cm, err
 	}
-	if err = json.Unmarshal(data, &meta); err != nil {
-		return meta, err
+	if err = json.Unmarshal(data, &cm); err != nil {
+		return cm, err
 	}
-	return meta, nil
+	return cm, nil
 }
 
-func saveCacheMeta(meta CacheMeta) error {
+func saveCacheMeta(cm CacheMeta) error {
 	if err := vfs.MkdirAll(cacheDir(), 0700); err != nil {
 		return err
 	}
-	data, err := json.Marshal(meta)
+	data, err := json.Marshal(cm)
 	if err != nil {
 		return err
 	}
@@ -163,14 +168,16 @@ func loadCachedMerged() (*MergedRegistry, error) {
 	return &reg, nil
 }
 
-func saveCachedMerged(data []byte, meta CacheMeta) error {
+// saveCachedMerged writes the cache file. data must be the server's data
+// payload verbatim (see remoteResponse) — never a typed re-marshal.
+func saveCachedMerged(data []byte, cm CacheMeta) error {
 	if err := vfs.MkdirAll(cacheDir(), 0700); err != nil {
 		return err
 	}
 	if err := validate.AtomicWrite(cachePath(), data, 0644); err != nil {
 		return err
 	}
-	return saveCacheMeta(meta)
+	return saveCacheMeta(cm)
 }
 
 // --- HTTP fetch ---
@@ -211,18 +218,21 @@ func fetchRemoteMerged(localVersion string) (data []byte, reg *MergedRegistry, e
 		return nil, nil, fmt.Errorf("remote meta: unexpected msg %q", envelope.Msg)
 	}
 
-	// If data.Services is nil, the version is up-to-date (not modified)
-	if envelope.Data.Services == nil {
+	var parsed MergedRegistry
+	if len(envelope.Data) > 0 {
+		if err := json.Unmarshal(envelope.Data, &parsed); err != nil {
+			return nil, nil, fmt.Errorf("remote meta: parse data: %w", err)
+		}
+	}
+
+	// If data.services is nil, the version is up-to-date (not modified)
+	if parsed.Services == nil {
 		return nil, nil, nil
 	}
 
-	// Re-marshal just the data portion for caching
-	dataBytes, err := json.Marshal(envelope.Data)
-	if err != nil {
-		return nil, nil, err
-	}
-
-	return dataBytes, &envelope.Data, nil
+	// Cache the data portion verbatim (see remoteResponse for why not a typed
+	// re-marshal).
+	return envelope.Data, &parsed, nil
 }
 
 type httpError struct {
@@ -247,12 +257,12 @@ func doSyncFetch() {
 		})
 		return
 	}
-	meta := CacheMeta{
+	cm := CacheMeta{
 		LastCheckAt: time.Now().Unix(),
 		Version:     reg.Version,
 		Brand:       string(configuredBrand),
 	}
-	_ = saveCachedMerged(data, meta)
+	_ = saveCachedMerged(data, cm)
 	overlayMergedServices(reg)
 }
 
@@ -275,22 +285,22 @@ func triggerBackgroundRefresh() {
 
 func doBackgroundRefresh() {
 	defer func() { _ = recover() }()
-	meta, _ := loadCacheMeta()
-	version := meta.Version
+	cm, _ := loadCacheMeta()
+	version := cm.Version
 	if version == "" {
 		version = embeddedVersion
 	}
 	data, reg, err := fetchRemoteMerged(version)
 	if err != nil {
 		// On error, update last_check_at to avoid retrying every invocation
-		meta.LastCheckAt = time.Now().Unix()
-		_ = saveCacheMeta(meta)
+		cm.LastCheckAt = time.Now().Unix()
+		_ = saveCacheMeta(cm)
 		return
 	}
 	if reg == nil {
 		// Version unchanged — just update check time
-		meta.LastCheckAt = time.Now().Unix()
-		_ = saveCacheMeta(meta)
+		cm.LastCheckAt = time.Now().Unix()
+		_ = saveCacheMeta(cm)
 		return
 	}
 	newMeta := CacheMeta{
@@ -302,21 +312,20 @@ func doBackgroundRefresh() {
 }
 
 // shouldRefresh returns true if the cache TTL has expired.
-func shouldRefresh(meta CacheMeta) bool {
-	if meta.LastCheckAt == 0 {
+func shouldRefresh(cm CacheMeta) bool {
+	if cm.LastCheckAt == 0 {
 		return true
 	}
-	return time.Since(time.Unix(meta.LastCheckAt, 0)) > metaTTL()
+	return time.Since(time.Unix(cm.LastCheckAt, 0)) > metaTTL()
 }
 
 // overlayMergedServices merges remote services into the in-memory map.
 // Remote entries override embedded entries with the same name.
 func overlayMergedServices(reg *MergedRegistry) {
 	for _, svc := range reg.Services {
-		name, ok := svc["name"].(string)
-		if !ok || name == "" {
+		if svc.Name == "" {
 			continue
 		}
-		mergedServices[name] = svc
+		mergedServices[svc.Name] = svc
 	}
 }

internal/registry/remote_test.go

@@ -15,6 +15,7 @@ import (
 	"time"
 
 	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/meta"
 )
 
 // waitBackgroundRefresh blocks until any in-flight background refresh started by
@@ -30,7 +31,10 @@ func resetInit() {
 	// reads globals this function mutates (see CI race: TestComputeMinimumScopeSet → Tenant).
 	waitBackgroundRefresh()
 	initOnce = sync.Once{}
-	mergedServices = make(map[string]map[string]interface{})
+	embeddedParseOnce = sync.Once{}
+	servicesTypedOnce = sync.Once{}
+	servicesTyped = nil
+	mergedServices = make(map[string]meta.Service)
 	mergedProjectList = nil
 	embeddedVersion = ""
 	cachedAllScopes = nil
@@ -71,13 +75,12 @@ func hasEmbeddedServices() bool {
 func testRegistry(name string) MergedRegistry {
 	return MergedRegistry{
 		Version: "test-1.0",
-		Services: []map[string]interface{}{
+		Services: []meta.Service{
 			{
-				"name":        name,
-				"version":     "v1",
-				"title":       name + " API",
-				"servicePath": "/open-apis/" + name + "/v1",
-				"resources":   map[string]interface{}{},
+				Name:        name,
+				Version:     "v1",
+				Title:       name + " API",
+				ServicePath: "/open-apis/" + name + "/v1",
 			},
 		},
 	}
@@ -91,9 +94,10 @@ func testCacheJSON(name string) []byte {
 
 // testEnvelopeJSON returns the remote API envelope format: {"msg":"succeeded","data":{...}}.
 func testEnvelopeJSON(name string) []byte {
+	regData, _ := json.Marshal(testRegistry(name))
 	resp := remoteResponse{
 		Msg:  "succeeded",
-		Data: testRegistry(name),
+		Data: regData,
 	}
 	data, _ := json.Marshal(resp)
 	return data
@@ -131,7 +135,7 @@ func TestColdStart_NoEmbedded_SyncFetch(t *testing.T) {
 
 	Init()
 
-	if spec := LoadFromMeta("remote_calendar"); spec == nil {
+	if _, ok := ServiceTyped("remote_calendar"); !ok {
 		t.Fatal("expected remote_calendar from sync fetch")
 	}
 }
@@ -150,7 +154,7 @@ func TestRemoteOff_SkipsRemoteLogic(t *testing.T) {
 	Init()
 
 	// "fake_remote_svc" should not be loaded when remote is off
-	if spec := LoadFromMeta("fake_remote_svc"); spec != nil {
+	if _, ok := ServiceTyped("fake_remote_svc"); ok {
 		t.Error("expected fake_remote_svc to NOT be loaded when remote is off")
 	}
 }
@@ -181,12 +185,12 @@ func TestCacheHit_WithinTTL(t *testing.T) {
 	Init()
 
 	// custom_svc should be loaded from cache overlay
-	if spec := LoadFromMeta("custom_svc"); spec == nil {
+	if _, ok := ServiceTyped("custom_svc"); !ok {
 		t.Error("expected custom_svc from cache overlay")
 	}
 	// Embedded projects should still be present (if compiled in)
 	if hasEmbeddedServices() {
-		if spec := LoadFromMeta("calendar"); spec == nil {
+		if _, ok := ServiceTyped("calendar"); !ok {
 			t.Error("expected calendar from embedded data")
 		}
 	}
@@ -227,7 +231,7 @@ func TestNetworkError_SilentDegradation(t *testing.T) {
 	if len(projects) == 0 {
 		t.Fatal("expected projects after network error")
 	}
-	if spec := LoadFromMeta("cached_svc"); spec == nil {
+	if _, ok := ServiceTyped("cached_svc"); !ok {
 		t.Fatal("expected cached_svc after network error")
 	}
 
@@ -304,19 +308,19 @@ func TestMetaTTL(t *testing.T) {
 
 func TestOverlayMergedServices(t *testing.T) {
 	resetInit()
-	mergedServices = make(map[string]map[string]interface{})
-	mergedServices["existing"] = map[string]interface{}{"name": "existing", "version": "v1"}
+	mergedServices = make(map[string]meta.Service)
+	mergedServices["existing"] = meta.Service{Name: "existing", Version: "v1"}
 
 	reg := &MergedRegistry{
-		Services: []map[string]interface{}{
-			{"name": "existing", "version": "v2"},
-			{"name": "brand_new", "version": "v1"},
+		Services: []meta.Service{
+			{Name: "existing", Version: "v2"},
+			{Name: "brand_new", Version: "v1"},
 		},
 	}
 	overlayMergedServices(reg)
 
 	// existing should be overridden
-	if v := mergedServices["existing"]["version"].(string); v != "v2" {
+	if v := mergedServices["existing"].Version; v != "v2" {
 		t.Errorf("expected existing to be overridden to v2, got %s", v)
 	}
 	// brand_new should be added
@@ -333,18 +337,18 @@ func TestOverlayMergedServicesDoesNotPolluteFollowingInit(t *testing.T) {
 	const leakedExisting = "test_isolation_existing_sentinel"
 	const leakedOverlay = "test_isolation_overlay_sentinel"
 
-	mergedServices = map[string]map[string]interface{}{
-		leakedExisting: {"name": leakedExisting, "version": "v1"},
+	mergedServices = map[string]meta.Service{
+		leakedExisting: {Name: leakedExisting, Version: "v1"},
 	}
-	overlayMergedServices(&MergedRegistry{Services: []map[string]interface{}{{"name": leakedOverlay, "version": "v1"}}})
+	overlayMergedServices(&MergedRegistry{Services: []meta.Service{{Name: leakedOverlay, Version: "v1"}}})
 
 	resetInit()
 	Init()
 
-	if spec := LoadFromMeta(leakedExisting); spec != nil {
+	if _, ok := ServiceTyped(leakedExisting); ok {
 		t.Fatalf("polluted service %q survived resetInit", leakedExisting)
 	}
-	if spec := LoadFromMeta(leakedOverlay); spec != nil {
+	if _, ok := ServiceTyped(leakedOverlay); ok {
 		t.Fatalf("polluted service %q survived resetInit", leakedOverlay)
 	}
 }
@@ -372,6 +376,31 @@ func TestFetchRemoteMerged_200(t *testing.T) {
 	}
 }
 
+// TestFetchRemoteMerged_CacheBytesPreserveUnmodeledKeys pins that the bytes
+// destined for the on-disk cache are the server's data payload verbatim: a key
+// the typed model does not (yet) declare must survive (see remoteResponse for
+// why).
+func TestFetchRemoteMerged_CacheBytesPreserveUnmodeledKeys(t *testing.T) {
+	const payload = `{"msg":"succeeded","data":{"version":"test-1.0","services":[{"name":"svc","servicePath":"/open-apis/svc/v1","resources":{"items":{"methods":{"list":{"httpMethod":"GET","path":"items","parameters":{"status":{"type":"string","enumName":"StatusEnum"}}}}}}}]}}`
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(200)
+		w.Write([]byte(payload))
+	}))
+	defer ts.Close()
+	testMetaURL = ts.URL
+
+	data, reg, err := fetchRemoteMerged("")
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if reg == nil || len(reg.Services) != 1 || reg.Services[0].Name != "svc" {
+		t.Fatalf("typed registry not decoded, got %+v", reg)
+	}
+	if !strings.Contains(string(data), `"enumName":"StatusEnum"`) {
+		t.Errorf("cache payload dropped unmodeled key enumName, got:\n%s", data)
+	}
+}
+
 func TestFetchRemoteMerged_VersionMatch(t *testing.T) {
 	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(200)
@@ -484,7 +513,7 @@ func TestBrandSwitchInvalidatesCache(t *testing.T) {
 
 	// The old feishu_svc should NOT be loaded from stale cache
 	// The new lark_svc from sync fetch should be available
-	if spec := LoadFromMeta("lark_svc"); spec == nil {
+	if _, ok := ServiceTyped("lark_svc"); !ok {
 		t.Error("expected lark_svc after brand switch sync fetch")
 	}
 }

internal/registry/scope_hint.go

@@ -44,22 +44,12 @@ func ExtractRequiredScopes(detail interface{}) []string {
 	return scopes
 }
 
-// SelectRecommendedScopeFromStrings is a string-typed convenience wrapper
-// around SelectRecommendedScope. When no scope is recognized by the priority
-// table, it falls back to the first input scope so callers always have
-// something to surface to users.
-func SelectRecommendedScopeFromStrings(scopes []string, identity string) string {
-	if len(scopes) == 0 {
-		return ""
-	}
-	ifaces := make([]interface{}, len(scopes))
-	for i, s := range scopes {
-		ifaces[i] = s
-	}
-	if recommended := SelectRecommendedScope(ifaces, identity); recommended != "" {
-		return recommended
-	}
-	return scopes[0]
+// SelectRecommendedScopeFromStrings returns the highest-priority (least-privilege)
+// scope to surface to users, or "" for no scopes. Unknown scopes score
+// DefaultScopeScore, so an all-unknown list yields the first entry. Priority is
+// identity-independent; the parameter is kept for call-site clarity.
+func SelectRecommendedScopeFromStrings(scopes []string, _ string) string {
+	return bestScope(scopes, LoadScopePriorities())
 }
 
 // BuildConsoleScopeURL returns the developer-console "apply scope" URL for the

internal/registry/scopes.go

@@ -6,16 +6,63 @@ package registry
 import (
 	"sort"
 	"strings"
+
+	"github.com/larksuite/cli/internal/apicatalog"
+	"github.com/larksuite/cli/internal/core"
+	"github.com/larksuite/cli/internal/meta"
 )
 
-// IdentityToAccessToken maps the --identity flag value to the corresponding
-// accessTokens value used in from_meta JSON files. Bot identity uses
-// tenant_access_token, so "bot" maps to "tenant".
-func IdentityToAccessToken(identity string) string {
-	if identity == "bot" {
-		return "tenant"
+// methodsForProjects walks the runtime catalog once and returns the methods in
+// the given projects that are reachable by the identity. Catalog navigation is
+// owned by apicatalog; the collectors below only apply scope policy.
+func methodsForProjects(projects []string, identity string) []apicatalog.MethodRef {
+	want := make(map[string]bool, len(projects))
+	for _, p := range projects {
+		want[p] = true
 	}
-	return identity
+	wantToken := meta.TokenForIdentity(identity)
+	supported := func(m meta.Method) bool { return m.SupportsToken(wantToken) }
+	// Walk only the requested services (in catalog name order) instead of every
+	// service's methods then discarding the rest.
+	var out []apicatalog.MethodRef
+	for _, svc := range RuntimeCatalog().Services() {
+		if want[svc.Name] {
+			out = append(out, apicatalog.ServiceMethods(svc, supported)...)
+		}
+	}
+	return out
+}
+
+// bestScope returns the highest-priority scope from scopes (minimum privilege),
+// or "" when scopes is empty.
+func bestScope(scopes []string, priorities map[string]int) string {
+	best := ""
+	bestScore := -1
+	for _, s := range scopes {
+		score := DefaultScopeScore
+		if v, ok := priorities[s]; ok {
+			score = v
+		}
+		if score > bestScore {
+			bestScore = score
+			best = s
+		}
+	}
+	return best
+}
+
+// FilterForStrictMode returns a method filter enforcing the strict-mode forced
+// identity, or nil when strict mode is inactive (no filtering). The
+// token/identity vocabulary (meta.TokenForIdentity) and the "no accessTokens =
+// permissive" predicate (meta.Method.SupportsToken) both live in meta, so this
+// only composes them — schema completion/render and service commands never
+// re-derive identity semantics.
+func FilterForStrictMode(mode core.StrictMode) apicatalog.MethodFilter {
+	if !mode.IsActive() {
+		return nil
+	}
+	token := meta.TokenForIdentity(string(mode.ForcedIdentity()))
+	return func(m meta.Method) bool { return m.SupportsToken(token) }
 }
 
 // FilterScopes filters scopes by domain and permission level.
@@ -76,71 +123,45 @@ func FilterScopes(allScopes []string, domains []string, permissions []string) []
 	return result
 }
 
+var cachedAllScopes map[string][]string
+
+// CollectAllScopesFromMeta collects all unique scopes from from_meta/*.json
+// for the given identity ("user" or "tenant"). Results are deduplicated and sorted.
+func CollectAllScopesFromMeta(identity string) []string {
+	if cachedAllScopes == nil {
+		cachedAllScopes = make(map[string][]string)
+	}
+	if cached, ok := cachedAllScopes[identity]; ok {
+		return cached
+	}
+
+	wantToken := meta.TokenForIdentity(identity)
+	supported := func(m meta.Method) bool { return m.SupportsToken(wantToken) }
+	scopeSet := make(map[string]bool)
+	for _, ref := range RuntimeCatalog().WalkMethods(supported) {
+		for _, s := range ref.Method.Scopes {
+			scopeSet[s] = true
+		}
+	}
+
+	result := make([]string, 0, len(scopeSet))
+	for s := range scopeSet {
+		result = append(result, s)
+	}
+	sort.Strings(result)
+	cachedAllScopes[identity] = result
+	return result
+}
+
 // CollectScopesForProjects collects the recommended scope for each API method
 // in the specified from_meta projects. For each method, only the scope with
 // the highest priority score is selected.
 func CollectScopesForProjects(projects []string, identity string) []string {
 	priorities := LoadScopePriorities()
 	scopeSet := make(map[string]bool)
-	for _, project := range projects {
-		spec := LoadFromMeta(project)
-		if spec == nil {
-			continue
-		}
-		resources, ok := spec["resources"].(map[string]interface{})
-		if !ok {
-			continue
-		}
-		for _, resSpec := range resources {
-			resMap, ok := resSpec.(map[string]interface{})
-			if !ok {
-				continue
-			}
-			methods, ok := resMap["methods"].(map[string]interface{})
-			if !ok {
-				continue
-			}
-			for _, methodSpec := range methods {
-				methodMap, ok := methodSpec.(map[string]interface{})
-				if !ok {
-					continue
-				}
-				if tokens, ok := methodMap["accessTokens"].([]interface{}); ok {
-					supported := false
-					for _, t := range tokens {
-						if ts, ok := t.(string); ok && ts == IdentityToAccessToken(identity) {
-							supported = true
-							break
-						}
-					}
-					if !supported {
-						continue
-					}
-				}
-				scopes, ok := methodMap["scopes"].([]interface{})
-				if !ok || len(scopes) == 0 {
-					continue
-				}
-				bestScope := ""
-				bestScore := -1
-				for _, s := range scopes {
-					str, ok := s.(string)
-					if !ok {
-						continue
-					}
-					score := DefaultScopeScore
-					if v, exists := priorities[str]; exists {
-						score = v
-					}
-					if score > bestScore {
-						bestScore = score
-						bestScope = str
-					}
-				}
-				if bestScope != "" {
-					scopeSet[bestScope] = true
-				}
-			}
+	for _, ref := range methodsForProjects(projects, identity) {
+		if best := bestScope(ref.Method.Scopes, priorities); best != "" {
+			scopeSet[best] = true
 		}
 	}
 
@@ -165,78 +186,25 @@ func CollectScopesWithSources(projects []string, identity string) ([]string, map
 	scopeSet := make(map[string]bool)
 	sources := make(map[string]*ScopeSource)
 
-	for _, project := range projects {
-		spec := LoadFromMeta(project)
-		if spec == nil {
+	for _, ref := range methodsForProjects(projects, identity) {
+		m := ref.Method
+		best := bestScope(m.Scopes, priorities)
+		if best == "" {
 			continue
 		}
-		resources, ok := spec["resources"].(map[string]interface{})
-		if !ok {
-			continue
+		scopeSet[best] = true
+		if sources[best] == nil {
+			sources[best] = &ScopeSource{}
 		}
-		for resName, resSpec := range resources {
-			resMap, ok := resSpec.(map[string]interface{})
-			if !ok {
-				continue
-			}
-			methods, ok := resMap["methods"].(map[string]interface{})
-			if !ok {
-				continue
-			}
-			for methodName, methodSpec := range methods {
-				methodMap, ok := methodSpec.(map[string]interface{})
-				if !ok {
-					continue
-				}
-				if tokens, ok := methodMap["accessTokens"].([]interface{}); ok {
-					supported := false
-					for _, t := range tokens {
-						if ts, ok := t.(string); ok && ts == IdentityToAccessToken(identity) {
-							supported = true
-							break
-						}
-					}
-					if !supported {
-						continue
-					}
-				}
-				scopes, ok := methodMap["scopes"].([]interface{})
-				if !ok || len(scopes) == 0 {
-					continue
-				}
-				bestScope := ""
-				bestScore := -1
-				for _, s := range scopes {
-					str, ok := s.(string)
-					if !ok {
-						continue
-					}
-					score := DefaultScopeScore
-					if v, exists := priorities[str]; exists {
-						score = v
-					}
-					if score > bestScore {
-						bestScore = score
-						bestScope = str
-					}
-				}
-				if bestScope != "" {
-					scopeSet[bestScope] = true
-					if sources[bestScope] == nil {
-						sources[bestScope] = &ScopeSource{}
-					}
-					methodID := GetStrFromMap(methodMap, "id")
-					if methodID == "" {
-						methodID = project + "." + resName + "." + methodName
-					}
-					httpMethod := GetStrFromMap(methodMap, "httpMethod")
-					if httpMethod == "" {
-						httpMethod = "?"
-					}
-					sources[bestScope].APIs = append(sources[bestScope].APIs, httpMethod+" "+methodID)
-				}
-			}
+		methodID := m.ID
+		if methodID == "" {
+			methodID = ref.ServiceName() + "." + ref.ResourceName() + "." + ref.MethodName()
 		}
+		httpMethod := m.HTTPMethod
+		if httpMethod == "" {
+			httpMethod = "?"
+		}
+		sources[best].APIs = append(sources[best].APIs, httpMethod+" "+methodID)
 	}
 
 	// Sort API lists for stable output
@@ -267,92 +235,27 @@ type CommandEntry struct {
 //   - If the method has a "requiredScopes" field, all of those scopes are needed (conjunction).
 //   - Otherwise, only the highest-priority scope from "scopes" is shown (minimum privilege).
 func CollectCommandScopes(projects []string, identity string) []CommandEntry {
-	priorities := LoadScopePriorities()
 	var entries []CommandEntry
 
-	for _, project := range projects {
-		spec := LoadFromMeta(project)
-		if spec == nil {
+	for _, ref := range methodsForProjects(projects, identity) {
+		m := ref.Method
+		if len(m.Scopes) == 0 {
 			continue
 		}
-		resources, ok := spec["resources"].(map[string]interface{})
-		if !ok {
+
+		// Effective-scope policy (requiredScopes conjunction, else recommended)
+		// lives once in DeclaredScopesForMethod.
+		effectiveScopes := DeclaredScopesForMethod(m, identity)
+		if len(effectiveScopes) == 0 {
 			continue
 		}
-		for resName, resSpec := range resources {
-			resMap, ok := resSpec.(map[string]interface{})
-			if !ok {
-				continue
-			}
-			methods, ok := resMap["methods"].(map[string]interface{})
-			if !ok {
-				continue
-			}
-			for methodName, methodSpec := range methods {
-				methodMap, ok := methodSpec.(map[string]interface{})
-				if !ok {
-					continue
-				}
-				if tokens, ok := methodMap["accessTokens"].([]interface{}); ok {
-					supported := false
-					for _, t := range tokens {
-						if ts, ok := t.(string); ok && ts == IdentityToAccessToken(identity) {
-							supported = true
-							break
-						}
-					}
-					if !supported {
-						continue
-					}
-				}
-				rawScopes, ok := methodMap["scopes"].([]interface{})
-				if !ok || len(rawScopes) == 0 {
-					continue
-				}
 
-				// Check for requiredScopes (conjunction — all needed)
-				var effectiveScopes []string
-				if reqRaw, ok := methodMap["requiredScopes"].([]interface{}); ok && len(reqRaw) > 0 {
-					for _, s := range reqRaw {
-						if str, ok := s.(string); ok {
-							effectiveScopes = append(effectiveScopes, str)
-						}
-					}
-				} else {
-					// Pick the single best scope (minimum privilege)
-					bestScope := ""
-					bestScore := -1
-					for _, s := range rawScopes {
-						str, ok := s.(string)
-						if !ok {
-							continue
-						}
-						score := DefaultScopeScore
-						if v, exists := priorities[str]; exists {
-							score = v
-						}
-						if score > bestScore {
-							bestScore = score
-							bestScope = str
-						}
-					}
-					if bestScope != "" {
-						effectiveScopes = []string{bestScope}
-					}
-				}
-				if len(effectiveScopes) == 0 {
-					continue
-				}
-
-				httpMethod := GetStrFromMap(methodMap, "httpMethod")
-				entries = append(entries, CommandEntry{
-					Command:    resName + " " + methodName,
-					Type:       "api",
-					Scopes:     effectiveScopes,
-					HTTPMethod: httpMethod,
-				})
-			}
-		}
+		entries = append(entries, CommandEntry{
+			Command:    ref.ResourceName() + " " + ref.MethodName(),
+			Type:       "api",
+			Scopes:     effectiveScopes,
+			HTTPMethod: m.HTTPMethod,
+		})
 	}
 
 	sort.Slice(entries, func(i, j int) bool {

internal/schema/assembler_test.go

@@ -10,6 +10,8 @@ import (
 	"strings"
 	"testing"
 
+	"github.com/larksuite/cli/internal/apicatalog"
+	"github.com/larksuite/cli/internal/meta"
 	"github.com/larksuite/cli/internal/registry"
 )
 
@@ -35,58 +37,6 @@ func TestMain(m *testing.M) {
 	os.Exit(code)
 }
 
-func TestKeyOrderIndex_ImReactionsList(t *testing.T) {
-	// We only assert key-set membership, not absolute order — the upstream
-	// meta_data API does not guarantee a stable JSON key sequence across
-	// fetches, so hard-coding the order makes CI flaky. Order preservation
-	// from input to output is tested separately in TestBuildInputSchema_*.
-	order := lookupKeyOrder("im", []string{"reactions"}, "list")
-	if order == nil {
-		t.Fatal("expected key order for im.reactions.list, got nil")
-	}
-	wantParams := map[string]bool{
-		"message_id": true, "reaction_type": true, "page_token": true,
-		"page_size": true, "user_id_type": true,
-	}
-	if got, want := len(order.Parameters), len(wantParams); got != want {
-		t.Errorf("parameters count = %d, want %d (got %v)", got, want, order.Parameters)
-	}
-	for _, k := range order.Parameters {
-		if !wantParams[k] {
-			t.Errorf("unexpected parameter key %q", k)
-		}
-	}
-	// im.reactions.list 是 GET，没有 requestBody
-	if len(order.RequestBody) != 0 {
-		t.Errorf("expected empty RequestBody, got %v", order.RequestBody)
-	}
-}
-
-func TestKeyOrderIndex_ImImagesCreate(t *testing.T) {
-	// Membership-only assertion; see comment on TestKeyOrderIndex_ImReactionsList.
-	order := lookupKeyOrder("im", []string{"images"}, "create")
-	if order == nil {
-		t.Fatal("expected key order for im.images.create, got nil")
-	}
-	wantBody := map[string]bool{"image_type": true, "image": true}
-	if got, want := len(order.RequestBody), len(wantBody); got != want {
-		t.Errorf("requestBody count = %d, want %d (got %v)", got, want, order.RequestBody)
-	}
-	for _, k := range order.RequestBody {
-		if !wantBody[k] {
-			t.Errorf("unexpected requestBody key %q", k)
-		}
-	}
-}
-
-func TestKeyOrderIndex_UnknownPath(t *testing.T) {
-	// 远端缓存的命令（不在 embedded 内）查不到 key order，返回 nil 走字母序兜底
-	order := lookupKeyOrder("nonexistent_service", []string{"foo"}, "bar")
-	if order != nil {
-		t.Errorf("expected nil for unknown path, got %+v", order)
-	}
-}
-
 func TestConvertProperty_BasicTypes(t *testing.T) {
 	tests := []struct {
 		name     string
@@ -288,9 +238,6 @@ func TestConvertProperty_DescriptionDefaultExample(t *testing.T) {
 
 func TestBuildInputSchema_ReactionsList(t *testing.T) {
 	method := loadMethodFromRegistry(t, "im", []string{"reactions"}, "list")
-	mko := lookupKeyOrder("im", []string{"reactions"}, "list")
-	currentMethodOrder = mko
-	defer func() { currentMethodOrder = nil }()
 
 	is := buildInputSchema(method)
 
@@ -313,16 +260,13 @@ func TestBuildInputSchema_ReactionsList(t *testing.T) {
 	if !reflect.DeepEqual(params.Required, []string{"message_id"}) {
 		t.Errorf("params.Required = %v, want [message_id]", params.Required)
 	}
-	if !reflect.DeepEqual(params.Properties.Order, mko.Parameters) {
-		t.Errorf("params.properties order = %v, want (from key index) %v",
-			params.Properties.Order, mko.Parameters)
+	if want := []string{"message_id", "page_size", "page_token", "reaction_type", "user_id_type"}; !reflect.DeepEqual(params.Properties.Order, want) {
+		t.Errorf("params.properties order = %v, want %v (alphabetical)", params.Properties.Order, want)
 	}
 }
 
 func TestBuildInputSchema_ImagesCreate_FileAndBody(t *testing.T) {
 	method := loadMethodFromRegistry(t, "im", []string{"images"}, "create")
-	currentMethodOrder = lookupKeyOrder("im", []string{"images"}, "create")
-	defer func() { currentMethodOrder = nil }()
 
 	is := buildInputSchema(method)
 
@@ -382,10 +326,8 @@ func TestBuildInputSchema_HighRiskWriteInjectsYes(t *testing.T) {
 			},
 		},
 	}
-	currentMethodOrder = nil
-	defer func() { currentMethodOrder = nil }()
 
-	is := buildInputSchema(method)
+	is := buildInputSchema(meta.FromMap(method))
 
 	// yes lives at inputSchema.properties.yes (sibling of params/data)
 	yes, ok := is.Properties.Map["yes"]
@@ -413,9 +355,6 @@ func TestBuildInputSchema_HighRiskWriteInjectsYes(t *testing.T) {
 
 func TestBuildInputSchema_NoYesForReadRisk(t *testing.T) {
 	method := loadMethodFromRegistry(t, "im", []string{"reactions"}, "list")
-	mko := lookupKeyOrder("im", []string{"reactions"}, "list")
-	currentMethodOrder = mko
-	defer func() { currentMethodOrder = nil }()
 
 	is := buildInputSchema(method)
 	if _, ok := is.Properties.Map["yes"]; ok {
@@ -425,9 +364,6 @@ func TestBuildInputSchema_NoYesForReadRisk(t *testing.T) {
 
 func TestBuildOutputSchema_ReactionsList(t *testing.T) {
 	method := loadMethodFromRegistry(t, "im", []string{"reactions"}, "list")
-	mko := lookupKeyOrder("im", []string{"reactions"}, "list")
-	currentMethodOrder = mko
-	defer func() { currentMethodOrder = nil }()
 
 	os := buildOutputSchema(method)
 
@@ -450,31 +386,6 @@ func TestBuildOutputSchema_ReactionsList(t *testing.T) {
 	}
 }
 
-func TestConvertAccessTokens(t *testing.T) {
-	tests := []struct {
-		name  string
-		input []interface{}
-		want  []string
-	}{
-		{"tenant only", []interface{}{"tenant"}, []string{"bot"}},
-		{"user only", []interface{}{"user"}, []string{"user"}},
-		{"tenant then user", []interface{}{"tenant", "user"}, []string{"bot", "user"}},
-		{"user then tenant", []interface{}{"user", "tenant"}, []string{"bot", "user"}},
-		{"deduped", []interface{}{"tenant", "tenant", "user"}, []string{"bot", "user"}},
-		{"empty", []interface{}{}, []string{}},
-		{"nil", nil, []string{}},
-		{"unknown skipped", []interface{}{"user", "admin"}, []string{"user"}},
-	}
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			got := convertAccessTokens(tt.input)
-			if !reflect.DeepEqual(got, tt.want) {
-				t.Errorf("got %v, want %v", got, tt.want)
-			}
-		})
-	}
-}
-
 func TestBuildMeta_FullFields(t *testing.T) {
 	// Synthesized method to avoid runtime variance from remote-cache overlay
 	// (which strips `risk` from merged services). All other field semantics
@@ -489,7 +400,7 @@ func TestBuildMeta_FullFields(t *testing.T) {
 		"accessTokens": []interface{}{"tenant"},
 		"docUrl":       "https://open.feishu.cn/document/uAjLw4CM/ukTMukTMukTM/reference/im-v1/image/create",
 	}
-	m := buildMeta(method)
+	m := buildMeta(meta.FromMap(method))
 
 	if m.EnvelopeVersion != "1.0" {
 		t.Errorf("EnvelopeVersion = %q", m.EnvelopeVersion)
@@ -526,7 +437,7 @@ func TestBuildMeta_MissingRiskDefaultsToRead(t *testing.T) {
 		"accessTokens": []interface{}{"user"},
 		// no risk field
 	}
-	m := buildMeta(method)
+	m := buildMeta(meta.FromMap(method))
 	if m.Risk != "read" {
 		t.Errorf("Risk = %q, want \"read\" (default for missing risk)", m.Risk)
 	}
@@ -540,58 +451,26 @@ func TestBuildMeta_RequiredScopesPresent(t *testing.T) {
 	}
 }
 
-func TestParseAffordance_NilOrEmpty(t *testing.T) {
-	cases := []struct {
-		name string
-		raw  interface{}
-	}{
-		{"nil", nil},
-		{"empty object", map[string]interface{}{}},
-		{"all-five-empty-arrays", map[string]interface{}{
-			"use_when":        []interface{}{},
-			"do_not_use_when": []interface{}{},
-			"prerequisites":   []interface{}{},
-			"examples":        []interface{}{},
-			"related":         []interface{}{},
-		}},
-		{"malformed (string)", "not an object"},
-		{"malformed (number)", 42},
-		{"malformed (nested type mismatch)", map[string]interface{}{
-			"examples": "should be a list, not a string",
-		}},
+func TestConvert_EnumDescriptions(t *testing.T) {
+	// options carrying descriptions -> enum + parallel enumDescriptions
+	withDesc := Convert(meta.Field{Type: "string", Options: []meta.Option{
+		{Value: "open_id", Description: "A"},
+		{Value: "user_id", Description: "B"},
+	}})
+	if !reflect.DeepEqual(withDesc.Enum, []interface{}{"open_id", "user_id"}) {
+		t.Errorf("Enum = %v", withDesc.Enum)
 	}
-	for _, c := range cases {
-		t.Run(c.name, func(t *testing.T) {
-			if got := parseAffordance(c.raw); got != nil {
-				t.Errorf("parseAffordance(%v) = %+v, want nil", c.raw, got)
-			}
-		})
+	if !reflect.DeepEqual(withDesc.EnumDescriptions, []string{"A", "B"}) {
+		t.Errorf("EnumDescriptions = %v, want [A B] aligned with enum", withDesc.EnumDescriptions)
 	}
-}
 
-func TestParseAffordance_FullPopulated(t *testing.T) {
-	raw := map[string]interface{}{
-		"use_when":        []interface{}{"需要拿到当前用户的主日历 ID"},
-		"do_not_use_when": []interface{}{"已知具体某一个非主日历的 calendar_id"},
-		"prerequisites":   []interface{}{"user 身份登录"},
-		"examples": []interface{}{
-			map[string]interface{}{"description": "获取主日历", "command": "lark-cli calendar calendars primary"},
-		},
-		"related": []interface{}{"calendars.list"},
+	// bare enum form (no descriptions) -> enumDescriptions omitted (nil)
+	bare := Convert(meta.Field{Type: "string", Enum: []any{"x", "y"}})
+	if !reflect.DeepEqual(bare.Enum, []interface{}{"x", "y"}) {
+		t.Errorf("bare Enum = %v", bare.Enum)
 	}
-	a := parseAffordance(raw)
-	if a == nil {
-		t.Fatal("parseAffordance returned nil, want populated")
-	}
-	if len(a.UseWhen) != 1 || a.UseWhen[0] != "需要拿到当前用户的主日历 ID" {
-		t.Errorf("UseWhen = %v", a.UseWhen)
-	}
-	if len(a.Examples) != 1 || a.Examples[0].Description != "获取主日历" ||
-		a.Examples[0].Command != "lark-cli calendar calendars primary" {
-		t.Errorf("Examples = %+v", a.Examples)
-	}
-	if len(a.Related) != 1 || a.Related[0] != "calendars.list" {
-		t.Errorf("Related = %v", a.Related)
+	if bare.EnumDescriptions != nil {
+		t.Errorf("bare enum must have nil EnumDescriptions, got %v", bare.EnumDescriptions)
 	}
 }
 
@@ -604,7 +483,7 @@ func TestBuildMeta_AffordanceFromMethod(t *testing.T) {
 			"use_when": []interface{}{"trigger"},
 		},
 	}
-	m := buildMeta(method)
+	m := buildMeta(meta.FromMap(method))
 	if m.Affordance == nil {
 		t.Fatal("Affordance should be populated from method[\"affordance\"]")
 	}
@@ -620,7 +499,7 @@ func TestBuildMeta_MissingDocURLOmitted(t *testing.T) {
 		"risk":         "read",
 		// no docUrl
 	}
-	m := buildMeta(method)
+	m := buildMeta(meta.FromMap(method))
 	if m.DocURL != "" {
 		t.Errorf("DocURL = %q, want empty (will be omitempty)", m.DocURL)
 	}
@@ -634,8 +513,7 @@ func TestBuildMeta_MissingDocURLOmitted(t *testing.T) {
 func TestBuildOutputSchema_EmptyResponseBody(t *testing.T) {
 	// 装配器对空 responseBody 应生成 properties = {} （不 nil）
 	method := map[string]interface{}{}
-	currentMethodOrder = nil
-	os := buildOutputSchema(method)
+	os := buildOutputSchema(meta.FromMap(method))
 	if os.Type != "object" {
 		t.Errorf("Type = %q, want \"object\"", os.Type)
 	}
@@ -647,9 +525,16 @@ func TestBuildOutputSchema_EmptyResponseBody(t *testing.T) {
 	}
 }
 
+// synthEnvelope renders an envelope for a synthetic (service, resourcePath, method)
+// via the public ref entry, so these unit tests build the same MethodRef the
+// command layer feeds Envelope.
+func synthEnvelope(serviceName string, resourcePath []string, m meta.Method) Envelope {
+	return EnvelopeOf(apicatalog.MethodRef{Service: meta.Service{Name: serviceName}, ResourcePath: resourcePath, Method: m})
+}
+
 func TestAssembleEnvelope_ReactionsList_FullStructure(t *testing.T) {
 	method := loadMethodFromRegistry(t, "im", []string{"reactions"}, "list")
-	env := AssembleEnvelope("im", []string{"reactions"}, "list", method)
+	env := synthEnvelope("im", []string{"reactions"}, method)
 
 	if env.Name != "im reactions list" {
 		t.Errorf("Name = %q, want \"im reactions list\"", env.Name)
@@ -671,7 +556,7 @@ func TestAssembleEnvelope_NestedResource_NameJoinedWithSpaces(t *testing.T) {
 	// overlay strips `bots` from the loaded method map on this environment;
 	// the assertion is about name joining, not method specifics.
 	method := loadMethodFromRegistry(t, "im", []string{"chat.members"}, "create")
-	env := AssembleEnvelope("im", []string{"chat.members"}, "create", method)
+	env := synthEnvelope("im", []string{"chat.members"}, method)
 	// chat.members resourcePath stays as one element in the slice with a dot;
 	// name should split it to "im chat.members create" — we keep the dot as-is
 	// inside the resource segment to round-trip with completion logic.
@@ -683,8 +568,8 @@ func TestAssembleEnvelope_NestedResource_NameJoinedWithSpaces(t *testing.T) {
 func TestAssembleEnvelope_JSONIsStable(t *testing.T) {
 	// Assemble twice; JSON output must be byte-identical (determinism).
 	method := loadMethodFromRegistry(t, "im", []string{"reactions"}, "list")
-	a := AssembleEnvelope("im", []string{"reactions"}, "list", method)
-	b := AssembleEnvelope("im", []string{"reactions"}, "list", method)
+	a := synthEnvelope("im", []string{"reactions"}, method)
+	b := synthEnvelope("im", []string{"reactions"}, method)
 	ja, _ := json.MarshalIndent(a, "", "  ")
 	jb, _ := json.MarshalIndent(b, "", "  ")
 	if string(ja) != string(jb) {
@@ -693,8 +578,8 @@ func TestAssembleEnvelope_JSONIsStable(t *testing.T) {
 }
 
 func TestAssembleService_Im(t *testing.T) {
-	spec := registry.LoadFromMeta("im")
-	envs := AssembleService("im", spec, nil)
+	svc, _ := registry.ServiceTyped("im")
+	envs := Envelopes(apicatalog.ServiceMethods(svc, nil))
 	if len(envs) == 0 {
 		t.Fatal("expected non-empty envelopes for service im")
 	}
@@ -713,17 +598,16 @@ func TestAssembleService_Im(t *testing.T) {
 }
 
 func TestAssembleService_FilterByAccessToken(t *testing.T) {
-	spec := registry.LoadFromMeta("im")
+	svc, _ := registry.ServiceTyped("im")
 	// Filter to bot-only (--as bot, which corresponds to "tenant")
-	envs := AssembleService("im", spec, func(method map[string]interface{}) bool {
-		tokens, _ := method["accessTokens"].([]interface{})
-		for _, t := range tokens {
-			if s, _ := t.(string); s == "tenant" {
+	envs := Envelopes(apicatalog.ServiceMethods(svc, func(m meta.Method) bool {
+		for _, t := range m.AccessTokens {
+			if t == "tenant" {
 				return true
 			}
 		}
 		return false
-	})
+	}))
 	// Every envelope's _meta.access_tokens must contain "bot"
 	for _, e := range envs {
 		found := false
@@ -740,11 +624,11 @@ func TestAssembleService_FilterByAccessToken(t *testing.T) {
 }
 
 func TestAssembleAll_AtLeast193(t *testing.T) {
-	envs := AssembleAll(nil)
-	// Envelope assembly is overlay-independent (Task 17b): AssembleAll walks the
-	// embedded meta_data.json directly, so the count is stable across machines.
+	envs := Envelopes(registry.EmbeddedCatalog().WalkMethods(nil))
+	// Envelope assembly is overlay-independent: it walks the embedded
+	// meta_data.json directly, so the count is stable across machines.
 	if len(envs) < 193 {
-		t.Errorf("AssembleAll returned %d envelopes, expected >= 193", len(envs))
+		t.Errorf("envelope count = %d, expected >= 193", len(envs))
 	}
 	// Spot check: im reactions list should be present
 	found := false
@@ -759,24 +643,32 @@ func TestAssembleAll_AtLeast193(t *testing.T) {
 	}
 }
 
-// loadMethodFromRegistry is a test helper that pulls one method's spec from the
-// real embedded meta_data.json via the registry package.
-func loadMethodFromRegistry(t *testing.T, service string, resourcePath []string, methodName string) map[string]interface{} {
+// loadMethodFromRegistry is a test helper that pulls one method from the real
+// embedded meta_data.json via the registry's typed accessor, with Name set.
+func loadMethodFromRegistry(t *testing.T, service string, resourcePath []string, methodName string) meta.Method {
 	t.Helper()
-	spec := registry.LoadFromMeta(service)
-	if spec == nil {
+	svc, ok := registry.ServiceTyped(service)
+	if !ok {
 		t.Fatalf("service %q not found in registry", service)
 	}
-	resources, _ := spec["resources"].(map[string]interface{})
 	resKey := strings.Join(resourcePath, ".")
-	res, ok := resources[resKey].(map[string]interface{})
+	res, ok := svc.Resources[resKey]
 	if !ok {
 		t.Fatalf("resource %q.%s not found", service, resKey)
 	}
-	methods, _ := res["methods"].(map[string]interface{})
-	m, ok := methods[methodName].(map[string]interface{})
+	m, ok := res.Methods[methodName]
 	if !ok {
 		t.Fatalf("method %q.%s.%s not found", service, resKey, methodName)
 	}
+	m.Name = methodName
 	return m
 }
+
+// convertProperty is a test helper: it decodes a single field-spec map into a
+// meta.Field and renders its Property (the conversion the assembler does).
+func convertProperty(fieldMap map[string]interface{}, _ string) Property {
+	b, _ := json.Marshal(fieldMap)
+	var f meta.Field
+	_ = json.Unmarshal(b, &f)
+	return Convert(f)
+}

internal/schema/lint.go

@@ -7,7 +7,7 @@ import (
 	"errors"
 	"fmt"
 
-	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/internal/core"
 )
 
 var validJSONSchemaTypes = map[string]bool{
@@ -81,7 +81,7 @@ func lintEnvelope(env Envelope) []error {
 	}
 
 	// ---- L3: cross-field self-consistency ----
-	dangerExpected := env.Meta.Risk == cmdutil.RiskWrite || env.Meta.Risk == cmdutil.RiskHighRiskWrite
+	dangerExpected := env.Meta.Risk == core.RiskWrite || env.Meta.Risk == core.RiskHighRiskWrite
 	if env.Meta.Danger != dangerExpected {
 		errs = append(errs, fmt.Errorf("L3: _meta.danger=%v inconsistent with risk=%q", env.Meta.Danger, env.Meta.Risk))
 	}
@@ -92,7 +92,7 @@ func lintEnvelope(env Envelope) []error {
 	if env.InputSchema != nil && env.InputSchema.Properties != nil {
 		_, hasYes = env.InputSchema.Properties.Map["yes"]
 	}
-	wantYes := env.Meta.Risk == cmdutil.RiskHighRiskWrite
+	wantYes := env.Meta.Risk == core.RiskHighRiskWrite
 	if hasYes != wantYes {
 		errs = append(errs, fmt.Errorf("L3: inputSchema `yes` property=%v inconsistent with risk=%q", hasYes, env.Meta.Risk))
 	}
@@ -125,6 +125,9 @@ func walkForL2(props *OrderedProps, errs *[]error) {
 		if p.Minimum != nil && p.Maximum != nil && *p.Minimum >= *p.Maximum {
 			*errs = append(*errs, fmt.Errorf("L2: field %q minimum (%v) >= maximum (%v)", k, *p.Minimum, *p.Maximum))
 		}
+		if n := len(p.EnumDescriptions); n > 0 && n != len(p.Enum) {
+			*errs = append(*errs, fmt.Errorf("L2: field %q enumDescriptions length (%d) != enum length (%d)", k, n, len(p.Enum)))
+		}
 		if len(p.Required) > 0 && p.Properties != nil {
 			for _, r := range p.Required {
 				if _, ok := p.Properties.Map[r]; !ok {

internal/schema/lint_test.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 	"testing"
 
+	"github.com/larksuite/cli/internal/apicatalog"
 	"github.com/larksuite/cli/internal/registry"
 )
 
@@ -147,6 +148,18 @@ func TestLintEnvelope_L2_TypeChecks(t *testing.T) {
 			},
 			wantSub: "minimum",
 		},
+		{
+			name: "enumDescriptions length must match enum",
+			mutate: func(e *Envelope) {
+				e.InputSchema.Properties.Order = []string{"k"}
+				e.InputSchema.Properties.Map["k"] = Property{
+					Type:             "string",
+					Enum:             []interface{}{"a", "b", "c"},
+					EnumDescriptions: []string{"only one"}, // misaligned with 3 enum values
+				}
+			},
+			wantSub: "enumDescriptions",
+		},
 		{
 			// Regression guard: walkForL2 must recurse into the params/data
 			// sub-objects introduced by the 4-bucket inputSchema, not only the
@@ -334,9 +347,8 @@ func TestAllEnvelopesPass(t *testing.T) {
 	knownEnvelopes := map[string]bool{}
 	// Use embedded data only so the gate is deterministic across machines
 	// (matches Task 17b: envelope assembly is overlay-independent).
-	for _, svc := range registry.EmbeddedServiceNames() {
-		spec := registry.EmbeddedSpec(svc)
-		envs := AssembleService(svc, spec, nil)
+	for _, svc := range registry.EmbeddedServicesTyped() {
+		envs := Envelopes(apicatalog.ServiceMethods(svc, nil))
 		for _, env := range envs {
 			errs := lintEnvelope(env)
 			if len(errs) == 0 {
@@ -366,7 +378,7 @@ func TestAllEnvelopesPass(t *testing.T) {
 	}
 
 	// L4 coverage report (warn-only via t.Logf)
-	all := AssembleAll(nil)
+	all := Envelopes(registry.EmbeddedCatalog().WalkMethods(nil))
 	c := measureCoverage(all)
 	for metric, rate := range c {
 		baseline := coverageBaseline[metric]

internal/schema/path.go

@@ -1,30 +0,0 @@
-// Copyright (c) 2026 Lark Technologies Pte. Ltd.
-// SPDX-License-Identifier: MIT
-
-package schema
-
-import "strings"
-
-// ParsePath normalizes the positional arguments of `lark-cli schema` into a
-// slice of path segments. It accepts two equivalent forms:
-//
-//	lark-cli schema im.messages.reply  -> single arg, split on "."
-//	lark-cli schema im messages reply  -> multiple args, used as-is
-//	lark-cli schema "im chat.members bots" is NOT a supported form; quote
-//	arguments individually if your shell needs it. Nested resources keep their
-//	internal dots (e.g. "chat.members").
-//
-// Returns nil for zero args (bare invocation).
-func ParsePath(args []string) []string {
-	switch len(args) {
-	case 0:
-		return nil
-	case 1:
-		if strings.Contains(args[0], ".") {
-			return strings.Split(args[0], ".")
-		}
-		return []string{args[0]}
-	default:
-		return args
-	}
-}

internal/schema/types.go

@@ -8,6 +8,8 @@ import (
 	"encoding/json"
 	"fmt"
 	"sort"
+
+	"github.com/larksuite/cli/internal/meta"
 )
 
 // Envelope is the MCP Tool spec contract for a single API method command.
@@ -45,42 +47,32 @@ type Property struct {
 	Type        string        `json:"type,omitempty"`
 	Description string        `json:"description,omitempty"`
 	Enum        []interface{} `json:"enum,omitempty"`
-	Default     interface{}   `json:"default,omitempty"`
-	Example     interface{}   `json:"example,omitempty"`
-	Minimum     *float64      `json:"minimum,omitempty"`
-	Maximum     *float64      `json:"maximum,omitempty"`
-	Format      string        `json:"format,omitempty"`
-	Required    []string      `json:"required,omitempty"`
-	Properties  *OrderedProps `json:"properties,omitempty"`
-	Items       *Property     `json:"items,omitempty"`
+	// EnumDescriptions, when present, is parallel to Enum: the human meaning of
+	// each allowed value, in the same order. Omitted when no value carries a
+	// description. This is the widely-recognized JSON-Schema extension (VS Code,
+	// OpenAPI tooling) that lets an AI consumer learn what each enum value means
+	// without a second lookup.
+	EnumDescriptions []string      `json:"enumDescriptions,omitempty"`
+	Default          interface{}   `json:"default,omitempty"`
+	Example          interface{}   `json:"example,omitempty"`
+	Minimum          *float64      `json:"minimum,omitempty"`
+	Maximum          *float64      `json:"maximum,omitempty"`
+	Format           string        `json:"format,omitempty"`
+	Required         []string      `json:"required,omitempty"`
+	Properties       *OrderedProps `json:"properties,omitempty"`
+	Items            *Property     `json:"items,omitempty"`
 }
 
 // Meta is the Lark-specific extension namespace.
 type Meta struct {
-	EnvelopeVersion string      `json:"envelope_version"`
-	Scopes          []string    `json:"scopes"`
-	RequiredScopes  []string    `json:"required_scopes"`
-	AccessTokens    []string    `json:"access_tokens"`
-	Danger          bool        `json:"danger"`
-	Risk            string      `json:"risk"`
-	DocURL          string      `json:"doc_url,omitempty"`
-	Affordance      *Affordance `json:"affordance,omitempty"`
-}
-
-// Affordance is the hand-written overlay (PR-1 only defines the type, no YAML loaded).
-type Affordance struct {
-	UseWhen       []string         `json:"use_when,omitempty"`
-	DoNotUseWhen  []string         `json:"do_not_use_when,omitempty"`
-	Prerequisites []string         `json:"prerequisites,omitempty"`
-	Examples      []AffordanceCase `json:"examples,omitempty"`
-	Related       []string         `json:"related,omitempty"`
-}
-
-// AffordanceCase is one example entry: a one-line description plus a
-// ready-to-run lark-cli command string.
-type AffordanceCase struct {
-	Description string `json:"description"`
-	Command     string `json:"command"`
+	EnvelopeVersion string           `json:"envelope_version"`
+	Scopes          []string         `json:"scopes"`
+	RequiredScopes  []string         `json:"required_scopes"`
+	AccessTokens    []string         `json:"access_tokens"`
+	Danger          bool             `json:"danger"`
+	Risk            string           `json:"risk"`
+	DocURL          string           `json:"doc_url,omitempty"`
+	Affordance      *meta.Affordance `json:"affordance,omitempty"`
 }
 
 // OrderedProps is map[string]Property with preserved key order on MarshalJSON.
@@ -91,6 +83,20 @@ type OrderedProps struct {
 	Map   map[string]Property
 }
 
+// Set adds or replaces a property, recording first-seen keys in Order so JSON
+// output preserves insertion order. Re-setting an existing key updates its
+// value without reordering. Centralizing mutation here keeps Order and Map from
+// drifting out of sync.
+func (o *OrderedProps) Set(key string, p Property) {
+	if o.Map == nil {
+		o.Map = make(map[string]Property)
+	}
+	if _, exists := o.Map[key]; !exists {
+		o.Order = append(o.Order, key)
+	}
+	o.Map[key] = p
+}
+
 // MarshalJSON emits keys in Order, not alphabetical. If Order is empty but
 // Map has entries, fall back to alphabetical key order over Map so callers
 // that only populated Map (no explicit ordering) still see their fields.

internal/schema/types_test.go

@@ -8,6 +8,21 @@ import (
 	"testing"
 )
 
+func TestOrderedProps_Set(t *testing.T) {
+	op := &OrderedProps{}
+	op.Set("b", Property{Type: "string"})
+	op.Set("a", Property{Type: "integer"})
+	op.Set("b", Property{Type: "boolean"}) // re-set: updates value, keeps position
+
+	wantOrder := []string{"b", "a"}
+	if len(op.Order) != len(wantOrder) || op.Order[0] != "b" || op.Order[1] != "a" {
+		t.Errorf("Order = %v, want %v (insertion order, no duplicate on re-set)", op.Order, wantOrder)
+	}
+	if op.Map["b"].Type != "boolean" {
+		t.Errorf("re-set value = %q, want boolean", op.Map["b"].Type)
+	}
+}
+
 // OrderedProps 在测试里验证：MarshalJSON 按 Order 切片顺序输出 key，跳过 Go map 默认字母序。
 func TestOrderedProps_MarshalJSON_PreservesOrder(t *testing.T) {
 	op := &OrderedProps{

internal/transport/warn.go

@@ -10,8 +10,6 @@ import (
 	"os"
 	"strings"
 	"sync"
-
-	"github.com/larksuite/cli/internal/envvars"
 )
 
 // Proxy environment constants control shared transport proxy behavior.
@@ -79,16 +77,8 @@ func WarnIfProxied(w io.Writer) {
 		// Shared), so its warning and disable instructions take precedence.
 		// Emitting the env-proxy warning here would be misleading: it tells the
 		// user to set LARK_CLI_NO_PROXY=1, which does NOT disable the plugin proxy.
-		if addr, caPath, enabled := proxyPluginStatus(); enabled {
-			fmt.Fprintf(w, "[lark-cli] [WARN] proxy plugin enabled: all requests (including credentials) are forced through %s. To disable, set %s=false or remove %s.\n",
-				redactProxyURL(addr), envvars.CliProxyEnable, Path())
-			if strings.TrimSpace(caPath) != "" {
-				// A custom CA means upstream TLS can be intercepted/inspected by
-				// the proxy (MITM). Surface it so the operator is aware traffic
-				// (including Bearer tokens) is decryptable on this host.
-				fmt.Fprintf(w, "[lark-cli] [WARN] proxy plugin trusts a custom CA (%s); TLS to upstreams can be intercepted/inspected by this proxy.\n",
-					caPath)
-			}
+		if _, _, enabled := proxyPluginStatus(); enabled {
+			fmt.Fprintln(w, "[lark-cli] [WARN] proxy plugin enabled: all requests are forced through proxy.")
 			return
 		}
 		if os.Getenv(EnvNoProxy) != "" {

internal/transport/warn_test.go

@@ -8,8 +8,6 @@ import (
 	"strings"
 	"sync"
 	"testing"
-
-	"github.com/larksuite/cli/internal/envvars"
 )
 
 // TestDetectProxyEnv verifies proxy environment detection priority and empty-state behavior.
@@ -120,9 +118,9 @@ func TestWarnIfProxied_OnlyOnce(t *testing.T) {
 }
 
 // TestWarnIfProxied_ProxyPluginEnabled verifies that when proxy plugin mode is
-// enabled, the warning describes the plugin proxy and the correct disable method
-// (LARKSUITE_CLI_PROXY_ENABLE=false) instead of the misleading LARK_CLI_NO_PROXY
-// instruction — even when env proxy and LARK_CLI_NO_PROXY are also set.
+// enabled, the warning is a single concise line that does not leak the proxy
+// address or give the misleading LARK_CLI_NO_PROXY disable instruction — even
+// when env proxy and LARK_CLI_NO_PROXY are also set.
 func TestWarnIfProxied_ProxyPluginEnabled(t *testing.T) {
 	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
 	unsetProxyPluginEnv(t)
@@ -140,51 +138,24 @@ func TestWarnIfProxied_ProxyPluginEnabled(t *testing.T) {
 	WarnIfProxied(&buf)
 	out := buf.String()
 
-	if !strings.Contains(out, "127.0.0.1:3128") {
-		t.Errorf("warning should mention the plugin proxy address, got: %s", out)
+	if !strings.Contains(out, "proxy plugin enabled") {
+		t.Errorf("warning should announce proxy plugin enabled, got: %s", out)
 	}
-	if !strings.Contains(out, envvars.CliProxyEnable) {
-		t.Errorf("warning should mention %s as the disable method, got: %s", envvars.CliProxyEnable, out)
+	// Single line only — no address, CA, or disable hints.
+	if strings.Count(out, "\n") != 1 {
+		t.Errorf("warning must be a single line, got: %s", out)
+	}
+	if strings.Contains(out, "127.0.0.1:3128") || strings.Contains(out, "corp-proxy") {
+		t.Errorf("warning must not leak the proxy address, got: %s", out)
 	}
 	if strings.Contains(out, "Set "+EnvNoProxy+"=1") {
 		t.Errorf("warning must NOT give the misleading %s disable instruction when plugin is enabled, got: %s", EnvNoProxy, out)
 	}
-	// No custom CA configured -> no interception warning.
-	if strings.Contains(out, "custom CA") {
-		t.Errorf("warning should not mention a custom CA when none is configured, got: %s", out)
-	}
-}
-
-// TestWarnIfProxied_ProxyPluginCustomCAWarns verifies that when a custom CA is
-// trusted, the warning surfaces the TLS-interception capability.
-func TestWarnIfProxied_ProxyPluginCustomCAWarns(t *testing.T) {
-	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
-	unsetProxyPluginEnv(t)
-	proxyWarningOnce = sync.Once{}
-
-	old := proxyPluginStatus
-	proxyPluginStatus = func() (string, string, bool) {
-		return "http://127.0.0.1:3128", "/etc/lark/extra_ca.pem", true
-	}
-	t.Cleanup(func() { proxyPluginStatus = old })
-
-	var buf bytes.Buffer
-	WarnIfProxied(&buf)
-	out := buf.String()
-
-	if !strings.Contains(out, "custom CA") {
-		t.Errorf("warning should mention the custom CA, got: %s", out)
-	}
-	if !strings.Contains(out, "/etc/lark/extra_ca.pem") {
-		t.Errorf("warning should include the CA path, got: %s", out)
-	}
-	if !strings.Contains(out, "intercept") {
-		t.Errorf("warning should mention TLS interception, got: %s", out)
-	}
 }
 
 // TestWarnIfProxied_ProxyPluginEnabledRedactsCredentials verifies the plugin
-// warning never leaks credentials embedded in the configured proxy address.
+// warning never leaks credentials embedded in the configured proxy address —
+// the simplified message omits the address entirely.
 func TestWarnIfProxied_ProxyPluginEnabledRedactsCredentials(t *testing.T) {
 	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
 	unsetProxyPluginEnv(t)
@@ -204,9 +175,6 @@ func TestWarnIfProxied_ProxyPluginEnabledRedactsCredentials(t *testing.T) {
 	if strings.Contains(out, "user:") {
 		t.Errorf("plugin warning leaked username, got: %s", out)
 	}
-	if !strings.Contains(out, "***@127.0.0.1:3128") {
-		t.Errorf("plugin warning should contain redacted proxy URL, got: %s", out)
-	}
 }
 
 // TestRedactProxyURL verifies redaction of proxy credentials across supported formats.

internal/vfs/localfileio/path.go

@@ -60,12 +60,12 @@ func safePath(raw, flagName string) (string, error) {
 		return "", err
 	}
 
-	path := filepath.Clean(raw)
-
-	if filepath.IsAbs(path) {
+	if isAbsolutePath(raw) {
 		return "", fmt.Errorf("%s must be a relative path within the current directory, got %q (hint: cd to the target directory first, or use a relative path like ./filename)", flagName, raw)
 	}
 
+	path := filepath.Clean(raw)
+
 	cwd, err := vfs.Getwd()
 	if err != nil {
 		return "", fmt.Errorf("cannot determine working directory: %w", err)
@@ -114,6 +114,21 @@ func resolveNearestAncestor(path string) (string, error) {
 	}
 }
 
+func isAbsolutePath(path string) bool {
+	path = strings.TrimSpace(path)
+	if path == "" {
+		return false
+	}
+	if filepath.IsAbs(path) || strings.HasPrefix(path, "/") || strings.HasPrefix(path, `\`) {
+		return true
+	}
+	if len(path) >= 3 && path[1] == ':' && (path[2] == '/' || path[2] == '\\') {
+		drive := path[0]
+		return ('A' <= drive && drive <= 'Z') || ('a' <= drive && drive <= 'z')
+	}
+	return false
+}
+
 func isUnderDir(child, parent string) bool {
 	rel, err := filepath.Rel(parent, child)
 	if err != nil {

internal/vfs/localfileio/path_test.go

@@ -34,6 +34,10 @@ func TestSafeOutputPath_RejectsPathTraversalAndDangerousInput(t *testing.T) {
 		// ── GIVEN: absolute paths → THEN: rejected ──
 		{"absolute path unix", "/etc/passwd", true},
 		{"absolute path root", "/tmp/evil", true},
+		{"absolute path windows drive", `C:\Users\agent\secret.txt`, true},
+		{"absolute path windows drive slash", "C:/Users/agent/secret.txt", true},
+		{"absolute path windows rooted", `\Users\agent\secret.txt`, true},
+		{"absolute path windows unc", `\\server\share\secret.txt`, true},
 
 		// ── GIVEN: control characters in path → THEN: rejected ──
 		{"null byte", "file\x00.txt", true},
@@ -187,11 +191,23 @@ func TestSafeUploadPath_AllowsTempFileAbsolutePath(t *testing.T) {
 }
 
 func TestSafeUploadPath_RejectsNonTempAbsolutePath(t *testing.T) {
-	// GIVEN: an absolute path outside the temp directory
-	// WHEN / THEN: SafeUploadPath rejects it
-	_, err := SafeInputPath("/etc/passwd")
-	if err == nil {
-		t.Error("expected error for absolute non-temp path, got nil")
+	for _, tt := range []struct {
+		name  string
+		input string
+	}{
+		{"absolute path unix", "/etc/passwd"},
+		{"absolute path windows drive", `C:\Users\agent\secret.txt`},
+		{"absolute path windows drive slash", "C:/Users/agent/secret.txt"},
+		{"absolute path windows rooted", `\Users\agent\secret.txt`},
+		{"absolute path windows unc", `\\server\share\secret.txt`},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			// WHEN / THEN: SafeInputPath rejects absolute paths on every platform.
+			_, err := SafeInputPath(tt.input)
+			if err == nil {
+				t.Errorf("expected error for absolute path %q, got nil", tt.input)
+			}
+		})
 	}
 }
 

shortcuts/drive/drive_export.go

@@ -39,230 +39,296 @@ var DriveExport = common.Shortcut{
 		{Name: "overwrite", Type: "bool", Desc: "overwrite existing output file"},
 	},
 	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		return validateDriveExportSpec(driveExportSpec{
-			Token:         runtime.Str("token"),
-			DocType:       runtime.Str("doc-type"),
-			FileExtension: runtime.Str("file-extension"),
-			SubID:         runtime.Str("sub-id"),
-		})
+		return ValidateExport(exportParamsFromFlags(runtime))
 	},
 	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		spec := driveExportSpec{
-			Token:         runtime.Str("token"),
-			DocType:       runtime.Str("doc-type"),
-			FileExtension: runtime.Str("file-extension"),
-			SubID:         runtime.Str("sub-id"),
-		}
-		// Markdown export is a special case: docx markdown comes from the V2
-		// docs_ai fetch API directly instead of the Drive export task API.
-		if spec.FileExtension == "markdown" {
-			apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
-			dr := common.NewDryRunAPI().
-				Desc("2-step orchestration: fetch docx markdown -> write local file").
-				POST(apiPath).
-				Body(map[string]interface{}{
-					"format": "markdown",
-				}).
-				Set("output_dir", runtime.Str("output-dir"))
-			if name := strings.TrimSpace(runtime.Str("file-name")); name != "" {
-				dr.Set("file_name", ensureExportFileExtension(sanitizeExportFileName(name, spec.Token), spec.FileExtension))
-			}
-			return dr
-		}
+		return PlanExportDryRun(runtime, exportParamsFromFlags(runtime))
+	},
+	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		return RunExport(ctx, runtime, exportParamsFromFlags(runtime))
+	},
+}
 
-		body := map[string]interface{}{
-			"token":          spec.Token,
-			"type":           spec.DocType,
-			"file_extension": spec.FileExtension,
-		}
-		if strings.TrimSpace(spec.SubID) != "" {
-			body["sub_id"] = spec.SubID
-		}
+// ExportParams holds the user-facing inputs for an export flow, decoupled from
+// cobra flags so other command groups (e.g. sheets +workbook-export) can reuse
+// the drive export implementation. An empty OutputDir means "create the export
+// task and poll, but do not download" — callers that only need the ready file
+// token / status get it back without writing a local file.
+type ExportParams struct {
+	Token         string
+	DocType       string
+	FileExtension string
+	SubID         string
+	OutputDir     string
+	FileName      string
+	Overwrite     bool
+}
 
+func (p ExportParams) spec() driveExportSpec {
+	return driveExportSpec{
+		Token:         p.Token,
+		DocType:       p.DocType,
+		FileExtension: p.FileExtension,
+		SubID:         p.SubID,
+	}
+}
+
+// exportParamsFromFlags reads the standard drive +export flag set.
+func exportParamsFromFlags(runtime *common.RuntimeContext) ExportParams {
+	// drive +export always downloads; an empty --output-dir historically means
+	// the current directory (saveContentToOutputDir maps "" -> "."), so normalize
+	// it here to keep behavior identical and stay off the export-only ("" => skip
+	// download) path that only sheets +workbook-export uses.
+	outputDir := runtime.Str("output-dir")
+	if outputDir == "" {
+		outputDir = "."
+	}
+	return ExportParams{
+		Token:         runtime.Str("token"),
+		DocType:       runtime.Str("doc-type"),
+		FileExtension: runtime.Str("file-extension"),
+		SubID:         runtime.Str("sub-id"),
+		OutputDir:     outputDir,
+		FileName:      strings.TrimSpace(runtime.Str("file-name")),
+		Overwrite:     runtime.Bool("overwrite"),
+	}
+}
+
+// ValidateExport runs the CLI-level export constraint checks.
+func ValidateExport(p ExportParams) error {
+	return validateDriveExportSpec(p.spec())
+}
+
+// PlanExportDryRun builds the dry-run plan for an export without performing I/O.
+func PlanExportDryRun(runtime *common.RuntimeContext, p ExportParams) *common.DryRunAPI {
+	spec := p.spec()
+	// Markdown export is a special case: docx markdown comes from the V2
+	// docs_ai fetch API directly instead of the Drive export task API.
+	if spec.FileExtension == "markdown" {
+		apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
 		dr := common.NewDryRunAPI().
-			Desc("3-step orchestration: create export task -> limited polling -> download file").
-			POST("/open-apis/drive/v1/export_tasks").
-			Body(body).
-			Set("output_dir", runtime.Str("output-dir"))
-		if name := strings.TrimSpace(runtime.Str("file-name")); name != "" {
+			Desc("2-step orchestration: fetch docx markdown -> write local file").
+			POST(apiPath).
+			Body(map[string]interface{}{
+				"format": "markdown",
+			}).
+			Set("output_dir", p.OutputDir)
+		if name := strings.TrimSpace(p.FileName); name != "" {
 			dr.Set("file_name", ensureExportFileExtension(sanitizeExportFileName(name, spec.Token), spec.FileExtension))
 		}
 		return dr
-	},
-	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		spec := driveExportSpec{
-			Token:         runtime.Str("token"),
-			DocType:       runtime.Str("doc-type"),
-			FileExtension: runtime.Str("file-extension"),
-			SubID:         runtime.Str("sub-id"),
+	}
+
+	body := map[string]interface{}{
+		"token":          spec.Token,
+		"type":           spec.DocType,
+		"file_extension": spec.FileExtension,
+	}
+	if strings.TrimSpace(spec.SubID) != "" {
+		body["sub_id"] = spec.SubID
+	}
+
+	dr := common.NewDryRunAPI().
+		Desc("3-step orchestration: create export task -> limited polling -> download file").
+		POST("/open-apis/drive/v1/export_tasks").
+		Body(body).
+		Set("output_dir", p.OutputDir)
+	if name := strings.TrimSpace(p.FileName); name != "" {
+		dr.Set("file_name", ensureExportFileExtension(sanitizeExportFileName(name, spec.Token), spec.FileExtension))
+	}
+	return dr
+}
+
+// RunExport drives create export task -> bounded poll -> optional download. It
+// is the shared core behind both drive +export and sheets +workbook-export. An
+// empty p.OutputDir skips the download step and returns the ready file token.
+func RunExport(ctx context.Context, runtime *common.RuntimeContext, p ExportParams) error {
+	spec := p.spec()
+	outputDir := p.OutputDir
+	preferredFileName := strings.TrimSpace(p.FileName)
+	overwrite := p.Overwrite
+
+	// Markdown export bypasses the async export task and writes the fetched
+	// markdown content directly to disk. Uses the V2 docs_ai fetch API for
+	// higher-quality Lark-flavored Markdown output.
+	if spec.FileExtension == "markdown" {
+		fmt.Fprintf(runtime.IO().ErrOut, "Exporting docx as markdown: %s\n", common.MaskToken(spec.Token))
+		apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
+		data, err := runtime.CallAPITyped(
+			"POST",
+			apiPath,
+			nil,
+			map[string]interface{}{
+				"format": "markdown",
+			},
+		)
+		if err != nil {
+			return err
 		}
-		outputDir := runtime.Str("output-dir")
-		preferredFileName := strings.TrimSpace(runtime.Str("file-name"))
-		overwrite := runtime.Bool("overwrite")
 
-		// Markdown export bypasses the async export task and writes the fetched
-		// markdown content directly to disk. Uses the V2 docs_ai fetch API for
-		// higher-quality Lark-flavored Markdown output.
-		if spec.FileExtension == "markdown" {
-			fmt.Fprintf(runtime.IO().ErrOut, "Exporting docx as markdown: %s\n", common.MaskToken(spec.Token))
-			apiPath := fmt.Sprintf("/open-apis/docs_ai/v1/documents/%s/fetch", validate.EncodePathSegment(spec.Token))
-			data, err := runtime.CallAPITyped(
-				"POST",
-				apiPath,
-				nil,
-				map[string]interface{}{
-					"format": "markdown",
-				},
-			)
+		// Extract content from the V2 response: data.document.content
+		doc, ok := data["document"].(map[string]interface{})
+		if !ok {
+			return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document object")
+		}
+		content, ok := doc["content"].(string)
+		if !ok {
+			return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document.content")
+		}
+
+		fileName := preferredFileName
+		if fileName == "" {
+			// Prefer the remote title for the exported file name, but still fall
+			// back to the token if metadata is empty.
+			title, err := common.FetchDriveMetaTitle(runtime, spec.Token, spec.DocType)
 			if err != nil {
-				return err
+				fmt.Fprintf(runtime.IO().ErrOut, "Title lookup failed, using token as filename: %v\n", err)
+				title = spec.Token
 			}
+			fileName = title
+		}
+		fileName = ensureExportFileExtension(sanitizeExportFileName(fileName, spec.Token), spec.FileExtension)
+		savedPath, err := saveContentToOutputDir(runtime.FileIO(), outputDir, fileName, []byte(content), overwrite)
+		if err != nil {
+			return err
+		}
 
-			// Extract content from the V2 response: data.document.content
-			doc, ok := data["document"].(map[string]interface{})
-			if !ok {
-				return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document object")
+		runtime.Out(map[string]interface{}{
+			"token":          spec.Token,
+			"doc_type":       spec.DocType,
+			"file_extension": spec.FileExtension,
+			"file_name":      filepath.Base(savedPath),
+			"saved_path":     savedPath,
+			"size_bytes":     len(content),
+		}, nil)
+		return nil
+	}
+
+	ticket, err := createDriveExportTask(runtime, spec)
+	if err != nil {
+		return err
+	}
+	fmt.Fprintf(runtime.IO().ErrOut, "Created export task: %s\n", ticket)
+
+	var lastStatus driveExportStatus
+	var lastPollErr error
+	hasObservedStatus := false
+	// Keep the command responsive by polling for a bounded window. If the task
+	// is still running after that, return a resume command instead of blocking.
+	for attempt := 1; attempt <= driveExportPollAttempts; attempt++ {
+		if attempt > 1 {
+			select {
+			case <-ctx.Done():
+				return ctx.Err()
+			case <-time.After(driveExportPollInterval):
 			}
-			content, ok := doc["content"].(string)
-			if !ok {
-				return errs.NewInternalError(errs.SubtypeInvalidResponse, "invalid markdown fetch response: missing document.content")
+		}
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+
+		status, err := getDriveExportStatus(runtime, spec.Token, ticket)
+		if err != nil {
+			// Treat polling failures as transient so short-lived backend hiccups
+			// do not immediately fail an otherwise healthy export task.
+			lastPollErr = err
+			fmt.Fprintf(runtime.IO().ErrOut, "Export status attempt %d/%d failed: %v\n", attempt, driveExportPollAttempts, err)
+			continue
+		}
+		lastStatus = status
+		hasObservedStatus = true
+
+		if status.Ready() {
+			fmt.Fprintf(runtime.IO().ErrOut, "Export task completed: %s\n", common.MaskToken(status.FileToken))
+
+			// Export-only mode: caller wants the ready file token / metadata but
+			// no local download (e.g. sheets +workbook-export without an output
+			// path). Skip the download and return the status envelope.
+			if strings.TrimSpace(outputDir) == "" {
+				runtime.Out(map[string]interface{}{
+					"ticket":         ticket,
+					"token":          spec.Token,
+					"doc_type":       spec.DocType,
+					"file_extension": spec.FileExtension,
+					"file_token":     status.FileToken,
+					"file_name":      status.FileName,
+					"file_size":      status.FileSize,
+					"ready":          true,
+					"downloaded":     false,
+				}, nil)
+				return nil
 			}
 
 			fileName := preferredFileName
 			if fileName == "" {
-				// Prefer the remote title for the exported file name, but still fall
-				// back to the token if metadata is empty.
-				title, err := common.FetchDriveMetaTitle(runtime, spec.Token, spec.DocType)
-				if err != nil {
-					fmt.Fprintf(runtime.IO().ErrOut, "Title lookup failed, using token as filename: %v\n", err)
-					title = spec.Token
-				}
-				fileName = title
+				fileName = status.FileName
 			}
 			fileName = ensureExportFileExtension(sanitizeExportFileName(fileName, spec.Token), spec.FileExtension)
-			savedPath, err := saveContentToOutputDir(runtime.FileIO(), outputDir, fileName, []byte(content), overwrite)
+			out, err := downloadDriveExportFile(ctx, runtime, status.FileToken, outputDir, fileName, overwrite)
 			if err != nil {
-				return err
+				recoveryCommand := driveExportDownloadCommand(status.FileToken, fileName, outputDir, overwrite)
+				hint := fmt.Sprintf(
+					"the export artifact is already ready (ticket=%s, file_token=%s)\nretry download with: %s",
+					ticket,
+					status.FileToken,
+					recoveryCommand,
+				)
+				return appendDriveExportRecoveryHint(err, hint)
 			}
-
-			runtime.Out(map[string]interface{}{
-				"token":          spec.Token,
-				"doc_type":       spec.DocType,
-				"file_extension": spec.FileExtension,
-				"file_name":      filepath.Base(savedPath),
-				"saved_path":     savedPath,
-				"size_bytes":     len(content),
-			}, nil)
+			out["ticket"] = ticket
+			out["doc_type"] = spec.DocType
+			out["file_extension"] = spec.FileExtension
+			runtime.Out(out, nil)
 			return nil
 		}
 
-		ticket, err := createDriveExportTask(runtime, spec)
-		if err != nil {
-			return err
-		}
-		fmt.Fprintf(runtime.IO().ErrOut, "Created export task: %s\n", ticket)
-
-		var lastStatus driveExportStatus
-		var lastPollErr error
-		hasObservedStatus := false
-		// Keep the command responsive by polling for a bounded window. If the task
-		// is still running after that, return a resume command instead of blocking.
-		for attempt := 1; attempt <= driveExportPollAttempts; attempt++ {
-			if attempt > 1 {
-				select {
-				case <-ctx.Done():
-					return ctx.Err()
-				case <-time.After(driveExportPollInterval):
-				}
+		if status.Failed() {
+			msg := strings.TrimSpace(status.JobErrorMsg)
+			if msg == "" {
+				msg = status.StatusLabel()
 			}
-			if err := ctx.Err(); err != nil {
-				return err
-			}
-
-			status, err := getDriveExportStatus(runtime, spec.Token, ticket)
-			if err != nil {
-				// Treat polling failures as transient so short-lived backend hiccups
-				// do not immediately fail an otherwise healthy export task.
-				lastPollErr = err
-				fmt.Fprintf(runtime.IO().ErrOut, "Export status attempt %d/%d failed: %v\n", attempt, driveExportPollAttempts, err)
-				continue
-			}
-			lastStatus = status
-			hasObservedStatus = true
-
-			if status.Ready() {
-				fmt.Fprintf(runtime.IO().ErrOut, "Export task completed: %s\n", common.MaskToken(status.FileToken))
-				fileName := preferredFileName
-				if fileName == "" {
-					fileName = status.FileName
-				}
-				fileName = ensureExportFileExtension(sanitizeExportFileName(fileName, spec.Token), spec.FileExtension)
-				out, err := downloadDriveExportFile(ctx, runtime, status.FileToken, outputDir, fileName, overwrite)
-				if err != nil {
-					recoveryCommand := driveExportDownloadCommand(status.FileToken, fileName, outputDir, overwrite)
-					hint := fmt.Sprintf(
-						"the export artifact is already ready (ticket=%s, file_token=%s)\nretry download with: %s",
-						ticket,
-						status.FileToken,
-						recoveryCommand,
-					)
-					return appendDriveExportRecoveryHint(err, hint)
-				}
-				out["ticket"] = ticket
-				out["doc_type"] = spec.DocType
-				out["file_extension"] = spec.FileExtension
-				runtime.Out(out, nil)
-				return nil
-			}
-
-			if status.Failed() {
-				msg := strings.TrimSpace(status.JobErrorMsg)
-				if msg == "" {
-					msg = status.StatusLabel()
-				}
-				return errs.NewAPIError(errs.SubtypeServerError, "export task failed: %s (ticket=%s)", msg, ticket)
-			}
-
-			fmt.Fprintf(runtime.IO().ErrOut, "Export status %d/%d: %s\n", attempt, driveExportPollAttempts, status.StatusLabel())
+			return errs.NewAPIError(errs.SubtypeServerError, "export task failed: %s (ticket=%s)", msg, ticket)
 		}
 
-		nextCommand := driveExportTaskResultCommand(ticket, spec.Token)
-		if !hasObservedStatus && lastPollErr != nil {
-			hint := fmt.Sprintf(
-				"the export task was created but every status poll failed (ticket=%s)\nretry status lookup with: %s",
-				ticket,
-				nextCommand,
-			)
-			return appendDriveExportRecoveryHint(lastPollErr, hint)
-		}
+		fmt.Fprintf(runtime.IO().ErrOut, "Export status %d/%d: %s\n", attempt, driveExportPollAttempts, status.StatusLabel())
+	}
 
-		failed := false
-		var jobStatus interface{}
-		jobStatusLabel := "unknown"
-		if hasObservedStatus {
-			failed = lastStatus.Failed()
-			jobStatus = lastStatus.JobStatus
-			jobStatusLabel = lastStatus.StatusLabel()
-		}
-		// Return the last observed status so callers can resume from a known task
-		// state instead of losing all progress information on timeout.
-		result := map[string]interface{}{
-			"ticket":           ticket,
-			"token":            spec.Token,
-			"doc_type":         spec.DocType,
-			"file_extension":   spec.FileExtension,
-			"ready":            false,
-			"failed":           failed,
-			"job_status":       jobStatus,
-			"job_status_label": jobStatusLabel,
-			"timed_out":        true,
-			"next_command":     nextCommand,
-		}
-		if preferredFileName != "" {
-			result["file_name"] = ensureExportFileExtension(sanitizeExportFileName(preferredFileName, spec.Token), spec.FileExtension)
-		}
-		runtime.Out(result, nil)
-		fmt.Fprintf(runtime.IO().ErrOut, "Export task is still in progress. Continue with: %s\n", nextCommand)
-		return nil
-	},
+	nextCommand := driveExportTaskResultCommand(ticket, spec.Token)
+	if !hasObservedStatus && lastPollErr != nil {
+		hint := fmt.Sprintf(
+			"the export task was created but every status poll failed (ticket=%s)\nretry status lookup with: %s",
+			ticket,
+			nextCommand,
+		)
+		return appendDriveExportRecoveryHint(lastPollErr, hint)
+	}
+
+	failed := false
+	var jobStatus interface{}
+	jobStatusLabel := "unknown"
+	if hasObservedStatus {
+		failed = lastStatus.Failed()
+		jobStatus = lastStatus.JobStatus
+		jobStatusLabel = lastStatus.StatusLabel()
+	}
+	// Return the last observed status so callers can resume from a known task
+	// state instead of losing all progress information on timeout.
+	result := map[string]interface{}{
+		"ticket":           ticket,
+		"token":            spec.Token,
+		"doc_type":         spec.DocType,
+		"file_extension":   spec.FileExtension,
+		"ready":            false,
+		"failed":           failed,
+		"job_status":       jobStatus,
+		"job_status_label": jobStatusLabel,
+		"timed_out":        true,
+		"next_command":     nextCommand,
+	}
+	if preferredFileName != "" {
+		result["file_name"] = ensureExportFileExtension(sanitizeExportFileName(preferredFileName, spec.Token), spec.FileExtension)
+	}
+	runtime.Out(result, nil)
+	fmt.Fprintf(runtime.IO().ErrOut, "Export task is still in progress. Continue with: %s\n", nextCommand)
+	return nil
 }

shortcuts/drive/drive_export_test.go

@@ -488,6 +488,72 @@ func TestDriveExportAsyncSuccess(t *testing.T) {
 	}
 }
 
+// TestDriveExportEmptyOutputDirDownloadsToCwd guards the export refactor: an
+// explicit empty --output-dir must still download to the current directory
+// (normalized to "."), not trigger the export-only no-download path that the
+// shared RunExport core uses for sheets +workbook-export.
+func TestDriveExportEmptyOutputDirDownloadsToCwd(t *testing.T) {
+	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/drive/v1/export_tasks",
+		Body:   map[string]interface{}{"code": 0, "data": map[string]interface{}{"ticket": "tk_e"}},
+	})
+	reg.Register(&httpmock.Stub{
+		Method: "GET",
+		URL:    "/open-apis/drive/v1/export_tasks/tk_e",
+		Body: map[string]interface{}{"code": 0, "data": map[string]interface{}{
+			"result": map[string]interface{}{
+				"job_status": 0, "file_token": "box_e", "file_name": "report",
+				"file_extension": "pdf", "type": "docx", "file_size": 3,
+			},
+		}},
+	})
+	reg.Register(&httpmock.Stub{
+		Method:  "GET",
+		URL:     "/open-apis/drive/v1/export_tasks/file/box_e/download",
+		Status:  200,
+		RawBody: []byte("pdf"),
+		Headers: http.Header{
+			"Content-Type":        []string{"application/pdf"},
+			"Content-Disposition": []string{`attachment; filename="report.pdf"`},
+		},
+	})
+
+	tmpDir := t.TempDir()
+	withDriveWorkingDir(t, tmpDir)
+
+	prevAttempts, prevInterval := driveExportPollAttempts, driveExportPollInterval
+	driveExportPollAttempts, driveExportPollInterval = 1, 0
+	t.Cleanup(func() {
+		driveExportPollAttempts, driveExportPollInterval = prevAttempts, prevInterval
+	})
+
+	err := mountAndRunDrive(t, DriveExport, []string{
+		"+export",
+		"--token", "docx123",
+		"--doc-type", "docx",
+		"--file-extension", "pdf",
+		"--output-dir", "",
+		"--as", "bot",
+	}, f, stdout)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	// Empty --output-dir must still write to cwd, not skip the download.
+	data, err := os.ReadFile(filepath.Join(tmpDir, "report.pdf"))
+	if err != nil {
+		t.Fatalf("empty --output-dir should still download to cwd: %v", err)
+	}
+	if string(data) != "pdf" {
+		t.Fatalf("downloaded content = %q", string(data))
+	}
+	if strings.Contains(stdout.String(), `"downloaded": false`) {
+		t.Fatalf("export-only path must not trigger for drive +export: %s", stdout.String())
+	}
+}
+
 func TestDriveExportAsyncUsesProvidedFileName(t *testing.T) {
 	f, stdout, _, reg := cmdutil.TestFactory(t, driveTestConfig())
 	reg.Register(&httpmock.Stub{

shortcuts/drive/drive_import.go

@@ -34,128 +34,160 @@ var DriveImport = common.Shortcut{
 		{Name: "target-token", Desc: "existing token to import data into (only for type=bitable); when set, data is mounted into this bitable instead of creating a new one"},
 	},
 	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		return validateDriveImportSpec(driveImportSpec{
-			FilePath:    runtime.Str("file"),
-			DocType:     strings.ToLower(runtime.Str("type")),
-			FolderToken: runtime.Str("folder-token"),
-			Name:        runtime.Str("name"),
-			TargetToken: runtime.Str("target-token"),
-		})
+		return ValidateImport(importParamsFromFlags(runtime))
 	},
 	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
-		spec := driveImportSpec{
-			FilePath:    runtime.Str("file"),
-			DocType:     strings.ToLower(runtime.Str("type")),
-			FolderToken: runtime.Str("folder-token"),
-			Name:        runtime.Str("name"),
-			TargetToken: runtime.Str("target-token"),
-		}
-		fileSize, err := preflightDriveImportFile(runtime.FileIO(), &spec)
-		if err != nil {
-			return common.NewDryRunAPI().Set("error", err.Error())
-		}
-		if valErr := validateDriveImportSpec(spec); valErr != nil {
-			return common.NewDryRunAPI().Set("error", valErr.Error())
-		}
-
-		dry := common.NewDryRunAPI()
-		dry.Desc("Upload file (single-part or multipart) -> create import task -> poll status")
-
-		appendDriveImportUploadDryRun(dry, spec, fileSize)
-
-		dry.POST("/open-apis/drive/v1/import_tasks").
-			Desc("[2] Create import task").
-			Body(spec.CreateTaskBody("<file_token>"))
-
-		dry.GET("/open-apis/drive/v1/import_tasks/:ticket").
-			Desc("[3] Poll import task result").
-			Set("ticket", "<ticket>")
-		if runtime.IsBot() {
-			dry.Desc("After the import result returns the final cloud document target in bot mode, the CLI will also try to grant the current CLI user full_access (可管理权限) on it.")
-		}
-
-		return dry
+		return PlanImportDryRun(runtime, importParamsFromFlags(runtime))
 	},
 	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
-		spec := driveImportSpec{
-			FilePath:    runtime.Str("file"),
-			DocType:     strings.ToLower(runtime.Str("type")),
-			FolderToken: runtime.Str("folder-token"),
-			Name:        runtime.Str("name"),
-			TargetToken: runtime.Str("target-token"),
-		}
-		if _, err := preflightDriveImportFile(runtime.FileIO(), &spec); err != nil {
-			return err
-		}
-
-		// Step 1: Upload file as media
-		fileToken, uploadErr := uploadMediaForImport(ctx, runtime, spec.FilePath, spec.SourceFileName(), spec.DocType)
-		if uploadErr != nil {
-			return uploadErr
-		}
-
-		fmt.Fprintf(runtime.IO().ErrOut, "Creating import task for %s as %s...\n", spec.TargetFileName(), spec.DocType)
-
-		// Step 2: Create import task
-		ticket, err := createDriveImportTask(runtime, spec, fileToken)
-		if err != nil {
-			return err
-		}
-
-		// Step 3: Poll task
-		fmt.Fprintf(runtime.IO().ErrOut, "Polling import task %s...\n", ticket)
-
-		status, ready, err := pollDriveImportTask(runtime, ticket)
-		if err != nil {
-			return err
-		}
-
-		// Some intermediate responses omit the final type, so fall back to the
-		// requested type to keep the output shape stable.
-		resultType := status.DocType
-		if resultType == "" {
-			resultType = spec.DocType
-		}
-		out := map[string]interface{}{
-			"ticket":           ticket,
-			"type":             resultType,
-			"ready":            ready,
-			"job_status":       status.JobStatus,
-			"job_status_label": status.StatusLabel(),
-		}
-		if status.Token != "" {
-			out["token"] = status.Token
-		}
-		if statusURL := strings.TrimSpace(status.URL); statusURL != "" {
-			out["url"] = statusURL
-		} else if status.Token != "" {
-			if u := common.BuildResourceURL(runtime.Config.Brand, normalizeDriveImportKindForURL(resultType, spec.DocType), status.Token); u != "" {
-				out["url"] = u
-			}
-		}
-		if status.JobErrorMsg != "" {
-			out["job_error_msg"] = status.JobErrorMsg
-		}
-		if status.Extra != nil {
-			out["extra"] = status.Extra
-		}
-		if !ready {
-			nextCommand := driveImportTaskResultCommand(ticket)
-			fmt.Fprintf(runtime.IO().ErrOut, "Import task is still in progress. Continue with: %s\n", nextCommand)
-			out["timed_out"] = true
-			out["next_command"] = nextCommand
-		}
-		if ready {
-			if grant := common.AutoGrantCurrentUserDrivePermission(runtime, common.GetString(out, "token"), resultType); grant != nil {
-				out["permission_grant"] = grant
-			}
-		}
-
-		runtime.Out(out, nil)
-		return nil
+		return RunImport(ctx, runtime, importParamsFromFlags(runtime))
 	},
 }
 
+// ImportParams holds the user-facing inputs for an import flow, decoupled from
+// cobra flags so other command groups (e.g. sheets +workbook-import) can reuse
+// the drive import implementation without taking a dependency on a --type flag.
+type ImportParams struct {
+	File        string
+	DocType     string
+	FolderToken string
+	Name        string
+	TargetToken string
+}
+
+func (p ImportParams) spec() driveImportSpec {
+	return driveImportSpec{
+		FilePath:    p.File,
+		DocType:     strings.ToLower(p.DocType),
+		FolderToken: p.FolderToken,
+		Name:        p.Name,
+		TargetToken: p.TargetToken,
+	}
+}
+
+// importParamsFromFlags reads the standard drive +import flag set.
+func importParamsFromFlags(runtime *common.RuntimeContext) ImportParams {
+	return ImportParams{
+		File:        runtime.Str("file"),
+		DocType:     runtime.Str("type"),
+		FolderToken: runtime.Str("folder-token"),
+		Name:        runtime.Str("name"),
+		TargetToken: runtime.Str("target-token"),
+	}
+}
+
+// ValidateImport runs the CLI-level compatibility checks for an import.
+func ValidateImport(p ImportParams) error {
+	return validateDriveImportSpec(p.spec())
+}
+
+// PlanImportDryRun builds the dry-run plan (upload -> create task -> poll) for
+// an import without performing any network or file I/O beyond a local stat.
+func PlanImportDryRun(runtime *common.RuntimeContext, p ImportParams) *common.DryRunAPI {
+	spec := p.spec()
+	fileSize, err := preflightDriveImportFile(runtime.FileIO(), &spec)
+	if err != nil {
+		return common.NewDryRunAPI().Set("error", err.Error())
+	}
+	if valErr := validateDriveImportSpec(spec); valErr != nil {
+		return common.NewDryRunAPI().Set("error", valErr.Error())
+	}
+
+	dry := common.NewDryRunAPI()
+	dry.Desc("Upload file (single-part or multipart) -> create import task -> poll status")
+
+	appendDriveImportUploadDryRun(dry, spec, fileSize)
+
+	dry.POST("/open-apis/drive/v1/import_tasks").
+		Desc("[2] Create import task").
+		Body(spec.CreateTaskBody("<file_token>"))
+
+	dry.GET("/open-apis/drive/v1/import_tasks/:ticket").
+		Desc("[3] Poll import task result").
+		Set("ticket", "<ticket>")
+	if runtime.IsBot() {
+		dry.Desc("After the import result returns the final cloud document target in bot mode, the CLI will also try to grant the current CLI user full_access (可管理权限) on it.")
+	}
+
+	return dry
+}
+
+// RunImport executes the full import flow: upload media -> create import task ->
+// bounded poll, then writes the result envelope to the runtime output. It is
+// the shared core behind both drive +import and sheets +workbook-import.
+func RunImport(ctx context.Context, runtime *common.RuntimeContext, p ImportParams) error {
+	spec := p.spec()
+	if _, err := preflightDriveImportFile(runtime.FileIO(), &spec); err != nil {
+		return err
+	}
+
+	// Step 1: Upload file as media
+	fileToken, uploadErr := uploadMediaForImport(ctx, runtime, spec.FilePath, spec.SourceFileName(), spec.DocType)
+	if uploadErr != nil {
+		return uploadErr
+	}
+
+	fmt.Fprintf(runtime.IO().ErrOut, "Creating import task for %s as %s...\n", spec.TargetFileName(), spec.DocType)
+
+	// Step 2: Create import task
+	ticket, err := createDriveImportTask(runtime, spec, fileToken)
+	if err != nil {
+		return err
+	}
+
+	// Step 3: Poll task
+	fmt.Fprintf(runtime.IO().ErrOut, "Polling import task %s...\n", ticket)
+
+	status, ready, err := pollDriveImportTask(runtime, ticket)
+	if err != nil {
+		return err
+	}
+
+	// Some intermediate responses omit the final type, so fall back to the
+	// requested type to keep the output shape stable.
+	resultType := status.DocType
+	if resultType == "" {
+		resultType = spec.DocType
+	}
+	out := map[string]interface{}{
+		"ticket":           ticket,
+		"type":             resultType,
+		"ready":            ready,
+		"job_status":       status.JobStatus,
+		"job_status_label": status.StatusLabel(),
+	}
+	if status.Token != "" {
+		out["token"] = status.Token
+	}
+	if statusURL := strings.TrimSpace(status.URL); statusURL != "" {
+		out["url"] = statusURL
+	} else if status.Token != "" {
+		if u := common.BuildResourceURL(runtime.Config.Brand, normalizeDriveImportKindForURL(resultType, spec.DocType), status.Token); u != "" {
+			out["url"] = u
+		}
+	}
+	if status.JobErrorMsg != "" {
+		out["job_error_msg"] = status.JobErrorMsg
+	}
+	if status.Extra != nil {
+		out["extra"] = status.Extra
+	}
+	if !ready {
+		nextCommand := driveImportTaskResultCommand(ticket)
+		fmt.Fprintf(runtime.IO().ErrOut, "Import task is still in progress. Continue with: %s\n", nextCommand)
+		out["timed_out"] = true
+		out["next_command"] = nextCommand
+	}
+	if ready {
+		if grant := common.AutoGrantCurrentUserDrivePermission(runtime, common.GetString(out, "token"), resultType); grant != nil {
+			out["permission_grant"] = grant
+		}
+	}
+
+	runtime.Out(out, nil)
+	return nil
+}
+
 func preflightDriveImportFile(fio fileio.FileIO, spec *driveImportSpec) (int64, error) {
 	// Keep dry-run and execution aligned on path normalization, file existence,
 	// and format-specific size limits before planning the upload path.

shortcuts/sheets/batch_op_contract_test.go

@@ -177,6 +177,18 @@ func TestBatchOp_BodyMatchesStandalone(t *testing.T) {
 			args:     []string{"--sheet-id", "sh1", "--color", "#FF0000"},
 			subInput: `{"sheet-id":"sh1","color":"#FF0000"}`,
 		},
+		{
+			shortcut: "+sheet-show-gridline",
+			sc:       SheetShowGridline,
+			args:     []string{"--sheet-id", "sh1"},
+			subInput: `{"sheet-id":"sh1"}`,
+		},
+		{
+			shortcut: "+sheet-hide-gridline",
+			sc:       SheetHideGridline,
+			args:     []string{"--sheet-id", "sh1"},
+			subInput: `{"sheet-id":"sh1"}`,
+		},
 		{
 			shortcut: "+dropdown-set",
 			sc:       DropdownSet,

shortcuts/sheets/batch_op_dispatch.go

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

shortcuts/sheets/data/flag-defs.json

@@ -25,6 +25,119 @@
       }
     ]
   },
+  "+history-list": {
+    "risk": "read",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "cursor",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Pagination cursor; pass the previous page's next_cursor, omit for first page"
+      },
+      {
+        "name": "count",
+        "kind": "own",
+        "type": "int",
+        "required": "optional",
+        "desc": "History versions per page, default 20"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
+  "+history-revert": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "revision-id",
+        "kind": "own",
+        "type": "string",
+        "required": "required",
+        "desc": "Restore the whole spreadsheet to this version: a revision_id (minor id) from +history-list"
+      },
+      {
+        "name": "edit-time",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "The matching edit_time from the same +history-list entry; pass it to locate the version faster"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
+  "+history-revert-status": {
+    "risk": "read",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "task-id",
+        "kind": "own",
+        "type": "string",
+        "required": "required",
+        "desc": "Revert task id returned by +history-revert"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
   "+sheet-create": {
     "risk": "write",
     "flags": [
@@ -54,7 +167,7 @@
         "kind": "own",
         "type": "int",
         "required": "optional",
-        "desc": "Insert position; appended to the end when omitted",
+        "desc": "Insert position (0-based); appended to the end when omitted",
         "default": "-1"
       },
       {
@@ -413,6 +526,86 @@
       }
     ]
   },
+  "+sheet-hide-gridline": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "sheet-id",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet reference_id (XOR with `--sheet-name`)"
+      },
+      {
+        "name": "sheet-name",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet name (XOR with `--sheet-id`)"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
+  "+sheet-show-gridline": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "sheet-id",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet reference_id (XOR with `--sheet-name`)"
+      },
+      {
+        "name": "sheet-name",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Sheet name (XOR with `--sheet-id`)"
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
   "+workbook-create": {
     "risk": "write",
     "flags": [
@@ -431,27 +624,45 @@
         "desc": "Target folder token; placed at the drive root when omitted"
       },
       {
-        "name": "headers",
+        "name": "values",
         "kind": "own",
         "type": "string",
         "required": "optional",
-        "desc": "Header row as a JSON array: `[\"Col A\",\"Col B\"]`",
+        "desc": "Untyped initial data as one 2D JSON array (`[[\"alice\",95]]`); values are written as-is with their type auto-detected, through the same batched set_cell_range path as --sheets — pair with --styles for number formats, colors, merges, and row/col sizes",
         "input": [
           "file",
           "stdin"
         ]
       },
       {
-        "name": "values",
+        "name": "sheets",
         "kind": "own",
         "type": "string",
         "required": "optional",
-        "desc": "Initial data as a 2D JSON array: `[[\"alice\",95]]`",
+        "desc": "Typed table payload as JSON (same shape as `+table-put`): a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it from a DataFrame via `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. Mutually exclusive with --values and --dataframe. Creates the workbook, then writes typed type-faithful data (dates land as real dates, numbers keep precision).",
         "input": [
           "file",
           "stdin"
         ]
       },
+      {
+        "name": "styles",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Initial visual operations as JSON: top-level `{styles:[...]}`. Each item corresponds to one target sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. With --sheets, styles array length/order/name must match --sheets.sheets. With --values, pass exactly one styles item for the initial sheet (its name is ignored).",
+        "input": [
+          "file",
+          "stdin"
+        ]
+      },
+      {
+        "name": "dataframe",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Single-sheet typed table from one Arrow IPC file (Feather v2 — what `pandas.DataFrame.to_feather()` writes), mutually exclusive with --values and --sheets. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema; per-column `number_format` may be set via Arrow field metadata. Creates the workbook and fills its default sheet (`Sheet1` — adopted in place, no empty Sheet1 left behind). For multi-sheet or non-default placement, use `--sheets` instead."
+      },
       {
         "name": "dry-run",
         "kind": "system",
@@ -513,6 +724,32 @@
       }
     ]
   },
+  "+workbook-import": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "file",
+        "kind": "own",
+        "type": "string",
+        "required": "required",
+        "desc": "Local file path (.xlsx / .xls / .csv)"
+      },
+      {
+        "name": "folder-token",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Target folder token; imported to the cloud drive root when omitted"
+      },
+      {
+        "name": "name",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Imported spreadsheet name; defaults to the local file name without its extension"
+      }
+    ]
+  },
   "+sheet-info": {
     "risk": "read",
     "flags": [
@@ -1212,19 +1449,72 @@
         "desc": "Skip hidden rows and columns; default `false`"
       },
       {
-        "name": "rows-json",
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": "Print the request path and parameters without executing"
+      }
+    ]
+  },
+  "+table-get": {
+    "risk": "read",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token (XOR with `--url`)"
+      },
+      {
+        "name": "sheet-id",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Read only this sheet (by id); omit to read all sheets"
+      },
+      {
+        "name": "sheet-name",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Read only this sheet (by name); omit to read all sheets"
+      },
+      {
+        "name": "range",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "A1 range to read; omit to read each sheet current region"
+      },
+      {
+        "name": "no-header",
         "kind": "own",
         "type": "bool",
         "required": "optional",
-        "desc": "Return structured rows ({row_number, values:{col→cell}}) instead of CSV text; default false",
-        "default": "false"
+        "desc": "Treat the first row as data instead of a header (columns get positional names col1, col2, ...)"
+      },
+      {
+        "name": "dataframe-out",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Write the typed table as one Arrow IPC file (Feather v2) instead of the default JSON. Pass `@<path>` for a file or `-` for binary stdout (same convention as other binary I/O flags). Mirror of the input-side `--dataframe` on `+table-put` / `+workbook-create` — pandas users round-trip via `df = pd.read_feather(\"x.arrow\")` or `pd.read_feather(io.BytesIO(stdout))`. Single-sheet only: requires `--sheet-id` or `--sheet-name`; whole-workbook reads keep the default JSON path. Column types come from the typed read-back (string/number/date/bool); per-column `number_format` is preserved as Arrow field metadata so the Arrow file can round-trip straight back through `+table-put --dataframe`."
       },
       {
         "name": "dry-run",
         "kind": "system",
         "type": "bool",
         "required": "optional",
-        "desc": "Print the request path and parameters without executing"
+        "desc": ""
       }
     ]
   },
@@ -1849,7 +2139,7 @@
         "kind": "own",
         "type": "string",
         "required": "required",
-        "desc": "RFC 4180 CSV text; plain values only (no formulas / styles / comments)",
+        "desc": "RFC 4180 CSV text; values or formulas (a leading = is evaluated as a formula); no styles / comments / images (use +cells-set for those).",
         "input": [
           "file",
           "stdin"
@@ -1880,6 +2170,61 @@
       }
     ]
   },
+  "+table-put": {
+    "risk": "write",
+    "flags": [
+      {
+        "name": "url",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet URL to write into (XOR with `--spreadsheet-token`)"
+      },
+      {
+        "name": "spreadsheet-token",
+        "kind": "public",
+        "type": "string",
+        "required": "xor",
+        "desc": "Spreadsheet token to write into (XOR with `--url`)"
+      },
+      {
+        "name": "sheets",
+        "kind": "own",
+        "type": "string",
+        "required": "xor",
+        "desc": "Typed table payload (pandas-DataFrame-shaped) as JSON, XOR with `--dataframe`: a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it with `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. `dtypes` values are pandas dtype strings (`int64`, `float64`, `Int64`, `bool`, `boolean`, `datetime64[ns]`, `object`, ...); the writer maps them to internal string/number/date/bool — omit `dtypes` and a column writes as text (good for raw CSV-shaped data). `formats[col]` is an Excel number_format string (e.g. `#,##0.00`, `0.0%`, `yyyy-mm`); when absent, date columns default to `yyyy-mm-dd` and string columns to text format (`@`).",
+        "input": [
+          "file",
+          "stdin"
+        ]
+      },
+      {
+        "name": "dataframe",
+        "kind": "own",
+        "type": "string",
+        "required": "xor",
+        "desc": "Single-sheet typed table from one Arrow IPC file (a.k.a. Feather v2 — what `pandas.DataFrame.to_feather()` writes), XOR with `--sheets`. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema (int*/uint*/float* → number, date32/date64/timestamp → date, utf8/large_utf8 → string, bool → bool); per-column `number_format` may be set via Arrow field metadata (`pa.field(\"price\", pa.float64(), metadata={b\"number_format\": b\"$#,##0.00\"})`). Writes the sheet at default placement: name `Sheet1` (created when absent), overwrite from A1 with header. For a different sheet name, anchor, mode, or to write multiple sheets, use `--sheets` instead."
+      },
+      {
+        "name": "styles",
+        "kind": "own",
+        "type": "string",
+        "required": "optional",
+        "desc": "Visual operations applied after the typed write, as JSON: top-level `{styles:[...]}`. Each item corresponds to one written sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. The styles array length/order/name must match the written sheets: with --sheets, match --sheets.sheets; with --dataframe (single sheet named Sheet1), pass exactly one styles item with name `Sheet1`.",
+        "input": [
+          "file",
+          "stdin"
+        ]
+      },
+      {
+        "name": "dry-run",
+        "kind": "system",
+        "type": "bool",
+        "required": "optional",
+        "desc": ""
+      }
+    ]
+  },
   "+cells-clear": {
     "risk": "high-risk-write",
     "flags": [

shortcuts/sheets/data/flag-schemas.json

@@ -462,8 +462,21 @@
                       "type": "string"
                     },
                     "mention_type": {
-                      "description": "@提及类型编号（仅 type='mention' 时可选）",
-                      "type": "number"
+                      "description": "@提及类型编号（仅 type='mention' 时可选）。0 或不填=@用户；@文件时按类型取：1=文档 3=电子表格 8=多维表格 11=思维笔记 12=文件 15=旧版幻灯片 16=知识库 22=新版文档 30=幻灯片 38=画板",
+                      "type": "number",
+                      "enum": [
+                        0,
+                        1,
+                        3,
+                        8,
+                        11,
+                        12,
+                        15,
+                        16,
+                        22,
+                        30,
+                        38
+                      ]
                     },
                     "notify": {
                       "description": "是否发送通知（仅 type='mention' 时可选，默认 true）",
@@ -1730,11 +1743,12 @@
                             },
                             "aggregateType": {
                               "type": "string",
-                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效",
+                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效。count 只统计数值单元格；counta 统计所有非空单元格（含文本），按文本/分类列统计出现次数（如各类别的数量、频次分布）时用 counta。",
                               "enum": [
                                 "sum",
                                 "average",
                                 "count",
+                                "counta",
                                 "min",
                                 "max",
                                 "median"
@@ -1787,11 +1801,7 @@
               "data"
             ]
           }
-        },
-        "required": [
-          "position",
-          "size"
-        ]
+        }
       }
     },
     "+chart-update": {
@@ -2769,11 +2779,12 @@
                             },
                             "aggregateType": {
                               "type": "string",
-                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效",
+                              "description": "汇总方式，默认为'sum'，仅在 aggregate 为 true 时生效。count 只统计数值单元格；counta 统计所有非空单元格（含文本），按文本/分类列统计出现次数（如各类别的数量、频次分布）时用 counta。",
                               "enum": [
                                 "sum",
                                 "average",
                                 "count",
+                                "counta",
                                 "min",
                                 "max",
                                 "median"
@@ -2826,11 +2837,7 @@
               "data"
             ]
           }
-        },
-        "required": [
-          "position",
-          "size"
-        ]
+        }
       }
     },
     "+cond-format-create": {
@@ -6249,6 +6256,744 @@
           }
         }
       }
+    },
+    "+table-put": {
+      "sheets": {
+        "type": "array",
+        "minItems": 1,
+        "description": "一个或多个子表的 typed 数据，每个数组元素写入一张子表；支持多 DataFrame → 多子表一次写入。整体形状对齐 pandas `df.to_json(orient=\"split\")`：列名走 `columns`、二维取值走 `data`、每列的 pandas dtype 走 `dtypes`、可选的展示格式走 `formats`。一行式用法：`{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`。",
+        "items": {
+          "type": "object",
+          "required": [
+            "name",
+            "columns",
+            "data"
+          ],
+          "properties": {
+            "name": {
+              "type": "string",
+              "description": "目标子表名。按名匹配已有子表；不存在则新建该子表。同一次调用内子表名不可重复。"
+            },
+            "start_cell": {
+              "type": "string",
+              "default": "A1",
+              "description": "写入起点单元格（A1 记法，如 \"B2\"），默认 \"A1\"。mode=append 时忽略其行号、仅沿用其列。"
+            },
+            "mode": {
+              "type": "string",
+              "enum": [
+                "overwrite",
+                "append"
+              ],
+              "default": "overwrite",
+              "description": "overwrite（默认）：从 start_cell 起写「表头 + 数据」块；append：把数据追加到子表已有数据下方（默认不重复表头）。"
+            },
+            "header": {
+              "type": "boolean",
+              "description": "是否写一行列名表头。省略时按 mode 取默认:overwrite→true、append→false（避免在已有表头下重复）；显式给值可覆盖。"
+            },
+            "allow_overwrite": {
+              "type": "boolean",
+              "default": true,
+              "description": "为 false 时，若写入会落在非空单元格则拒写以保护原数据（返回 partial_success）。默认 true。"
+            },
+            "columns": {
+              "type": "array",
+              "minItems": 1,
+              "description": "列名字符串数组，顺序与 `data` 中每行取值一一对应。同一子表内列名不可重复。",
+              "items": {
+                "type": "string"
+              }
+            },
+            "data": {
+              "type": "array",
+              "description": "数据行；每行是一个数组，长度必须等于 `columns` 数。元素按 `dtypes` 推得的列类型取值（date 列写 ISO yyyy-mm-dd 字符串、number 列写数值、bool 列写布尔、其余写文本），null 表示空单元格。",
+              "items": {
+                "type": "array",
+                "items": {
+                  "type": [
+                    "string",
+                    "number",
+                    "boolean",
+                    "null"
+                  ],
+                  "description": "单元格值：date→ISO yyyy-mm-dd 字符串；number→数值（json.Number 精度保留）；bool→布尔；string→文本；null→空单元格。"
+                }
+              }
+            },
+            "dtypes": {
+              "type": "object",
+              "description": "可选。列名 → pandas dtype 字符串的映射；缺失项默认按 object（string + 文本格式 `@`）处理，所以省略整段时整张表按文本写入（导入 CSV-shaped 数据的最简形态）。dtype 解析规则：`int*` / `uint*` / `Int*` / `UInt*` / `float*` / `Float*` / `complex*` → number（精度保留），`bool` / `boolean` → bool，`datetime64[ns]` / 含时区的 `datetime64[ns, UTC]` 等 → date（默认 `yyyy-mm-dd` 格式），`object` / `string` / `category` / 未识别 → string + 文本格式 `@`（数字样字符串如「00123」不会塌缩成数字）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            },
+            "formats": {
+              "type": "object",
+              "description": "可选。列名 → Excel number_format 字符串的映射，覆盖 dtype 自带的默认格式（金额 `#,##0.00`、百分比 `0.0%`、自定义日期 `yyyy-mm` 等）。percent 列的数值尺度由调用方负责（0.0469 配 `0.00%` 显示 4.69%）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            }
+          }
+        }
+      },
+      "styles": {
+        "items": {
+          "properties": {
+            "cell_merges": {
+              "description": "单元格合并操作数组；range 使用 A1 单元格范围，merge_type 默认 all。",
+              "items": {
+                "properties": {
+                  "merge_type": {
+                    "enum": [
+                      "all",
+                      "rows",
+                      "columns"
+                    ],
+                    "type": "string"
+                  },
+                  "range": {
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "cell_styles": {
+              "description": "单元格样式操作数组；每项用 A1 单元格 range 指定范围，字段名与 +cells-set-style 对齐。",
+              "items": {
+                "properties": {
+                  "background_color": {
+                    "type": "string"
+                  },
+                  "border_styles": {
+                    "type": "object",
+                    "description": "边框配置，结构同 +cells-set-style --border-styles。",
+                    "properties": {
+                      "bottom": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "left": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "right": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "top": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      }
+                    }
+                  },
+                  "font_color": {
+                    "type": "string"
+                  },
+                  "font_line": {
+                    "enum": [
+                      "none",
+                      "underline",
+                      "line-through"
+                    ],
+                    "type": "string"
+                  },
+                  "font_size": {
+                    "type": "number"
+                  },
+                  "font_style": {
+                    "enum": [
+                      "normal",
+                      "italic"
+                    ],
+                    "type": "string"
+                  },
+                  "font_weight": {
+                    "enum": [
+                      "normal",
+                      "bold"
+                    ],
+                    "type": "string"
+                  },
+                  "horizontal_alignment": {
+                    "enum": [
+                      "left",
+                      "center",
+                      "right"
+                    ],
+                    "type": "string"
+                  },
+                  "number_format": {
+                    "type": "string"
+                  },
+                  "range": {
+                    "description": "A1 单元格范围，必须落在该子表本次写入区域内；例如 A1:B1、B2。",
+                    "type": "string"
+                  },
+                  "vertical_alignment": {
+                    "enum": [
+                      "top",
+                      "middle",
+                      "bottom"
+                    ],
+                    "type": "string"
+                  },
+                  "word_wrap": {
+                    "enum": [
+                      "overflow",
+                      "auto-wrap",
+                      "word-clip"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "col_sizes": {
+              "description": "列宽操作数组；range 使用列范围如 A:C，type 为 pixel/standard，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "name": {
+              "description": "子表名。--sheets 模式下必须与同位置 --sheets.sheets[].name 一致；--values 模式下建议写 Sheet1（其 name 会被忽略）。",
+              "type": "string"
+            },
+            "row_sizes": {
+              "description": "行高操作数组；range 使用行范围如 1:3，type 为 pixel/standard/auto，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard",
+                      "auto"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            }
+          },
+          "required": [
+            "name"
+          ],
+          "type": "object"
+        },
+        "type": "array"
+      }
+    },
+    "+workbook-create": {
+      "sheets": {
+        "type": "array",
+        "minItems": 1,
+        "description": "一个或多个子表的 typed 数据，每个数组元素写入一张子表；支持多 DataFrame → 多子表一次写入。整体形状对齐 pandas `df.to_json(orient=\"split\")`：列名走 `columns`、二维取值走 `data`、每列的 pandas dtype 走 `dtypes`、可选的展示格式走 `formats`。一行式用法：`{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`。",
+        "items": {
+          "type": "object",
+          "required": [
+            "name",
+            "columns",
+            "data"
+          ],
+          "properties": {
+            "name": {
+              "type": "string",
+              "description": "目标子表名。按名匹配已有子表；不存在则新建该子表。同一次调用内子表名不可重复。"
+            },
+            "start_cell": {
+              "type": "string",
+              "default": "A1",
+              "description": "写入起点单元格（A1 记法，如 \"B2\"），默认 \"A1\"。mode=append 时忽略其行号、仅沿用其列。"
+            },
+            "mode": {
+              "type": "string",
+              "enum": [
+                "overwrite",
+                "append"
+              ],
+              "default": "overwrite",
+              "description": "overwrite（默认）：从 start_cell 起写「表头 + 数据」块；append：把数据追加到子表已有数据下方（默认不重复表头）。"
+            },
+            "header": {
+              "type": "boolean",
+              "description": "是否写一行列名表头。省略时按 mode 取默认:overwrite→true、append→false（避免在已有表头下重复）；显式给值可覆盖。"
+            },
+            "allow_overwrite": {
+              "type": "boolean",
+              "default": true,
+              "description": "为 false 时，若写入会落在非空单元格则拒写以保护原数据（返回 partial_success）。默认 true。"
+            },
+            "columns": {
+              "type": "array",
+              "minItems": 1,
+              "description": "列名字符串数组，顺序与 `data` 中每行取值一一对应。同一子表内列名不可重复。",
+              "items": {
+                "type": "string"
+              }
+            },
+            "data": {
+              "type": "array",
+              "description": "数据行；每行是一个数组，长度必须等于 `columns` 数。元素按 `dtypes` 推得的列类型取值（date 列写 ISO yyyy-mm-dd 字符串、number 列写数值、bool 列写布尔、其余写文本），null 表示空单元格。",
+              "items": {
+                "type": "array",
+                "items": {
+                  "type": [
+                    "string",
+                    "number",
+                    "boolean",
+                    "null"
+                  ],
+                  "description": "单元格值：date→ISO yyyy-mm-dd 字符串；number→数值（json.Number 精度保留）；bool→布尔；string→文本；null→空单元格。"
+                }
+              }
+            },
+            "dtypes": {
+              "type": "object",
+              "description": "可选。列名 → pandas dtype 字符串的映射；缺失项默认按 object（string + 文本格式 `@`）处理，所以省略整段时整张表按文本写入（导入 CSV-shaped 数据的最简形态）。dtype 解析规则：`int*` / `uint*` / `Int*` / `UInt*` / `float*` / `Float*` / `complex*` → number（精度保留），`bool` / `boolean` → bool，`datetime64[ns]` / 含时区的 `datetime64[ns, UTC]` 等 → date（默认 `yyyy-mm-dd` 格式），`object` / `string` / `category` / 未识别 → string + 文本格式 `@`（数字样字符串如「00123」不会塌缩成数字）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            },
+            "formats": {
+              "type": "object",
+              "description": "可选。列名 → Excel number_format 字符串的映射，覆盖 dtype 自带的默认格式（金额 `#,##0.00`、百分比 `0.0%`、自定义日期 `yyyy-mm` 等）。percent 列的数值尺度由调用方负责（0.0469 配 `0.00%` 显示 4.69%）。",
+              "additionalProperties": {
+                "type": "string"
+              }
+            }
+          }
+        }
+      },
+      "styles": {
+        "items": {
+          "properties": {
+            "cell_merges": {
+              "description": "单元格合并操作数组；range 使用 A1 单元格范围，merge_type 默认 all。",
+              "items": {
+                "properties": {
+                  "merge_type": {
+                    "enum": [
+                      "all",
+                      "rows",
+                      "columns"
+                    ],
+                    "type": "string"
+                  },
+                  "range": {
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "cell_styles": {
+              "description": "单元格样式操作数组；每项用 A1 单元格 range 指定范围，字段名与 +cells-set-style 对齐。",
+              "items": {
+                "properties": {
+                  "background_color": {
+                    "type": "string"
+                  },
+                  "border_styles": {
+                    "type": "object",
+                    "description": "边框配置，结构同 +cells-set-style --border-styles。",
+                    "properties": {
+                      "bottom": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "left": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "right": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      },
+                      "top": {
+                        "properties": {
+                          "color": {
+                            "description": "边框颜色（十六进制，例如 \"#000000\"）",
+                            "type": "string"
+                          },
+                          "style": {
+                            "description": "边框线型；传 \"none\" 表示清除该方向边框（无边框线）",
+                            "enum": [
+                              "solid",
+                              "dashed",
+                              "dotted",
+                              "double",
+                              "none"
+                            ],
+                            "type": "string"
+                          },
+                          "weight": {
+                            "description": "边框粗细/线宽",
+                            "enum": [
+                              "thin",
+                              "medium",
+                              "thick"
+                            ],
+                            "type": "string"
+                          }
+                        },
+                        "type": "object"
+                      }
+                    }
+                  },
+                  "font_color": {
+                    "type": "string"
+                  },
+                  "font_line": {
+                    "enum": [
+                      "none",
+                      "underline",
+                      "line-through"
+                    ],
+                    "type": "string"
+                  },
+                  "font_size": {
+                    "type": "number"
+                  },
+                  "font_style": {
+                    "enum": [
+                      "normal",
+                      "italic"
+                    ],
+                    "type": "string"
+                  },
+                  "font_weight": {
+                    "enum": [
+                      "normal",
+                      "bold"
+                    ],
+                    "type": "string"
+                  },
+                  "horizontal_alignment": {
+                    "enum": [
+                      "left",
+                      "center",
+                      "right"
+                    ],
+                    "type": "string"
+                  },
+                  "number_format": {
+                    "type": "string"
+                  },
+                  "range": {
+                    "description": "A1 单元格范围，必须落在该子表本次写入区域内；例如 A1:B1、B2。",
+                    "type": "string"
+                  },
+                  "vertical_alignment": {
+                    "enum": [
+                      "top",
+                      "middle",
+                      "bottom"
+                    ],
+                    "type": "string"
+                  },
+                  "word_wrap": {
+                    "enum": [
+                      "overflow",
+                      "auto-wrap",
+                      "word-clip"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "col_sizes": {
+              "description": "列宽操作数组；range 使用列范围如 A:C，type 为 pixel/standard，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            },
+            "name": {
+              "description": "子表名。--sheets 模式下必须与同位置 --sheets.sheets[].name 一致；--values 模式下建议写 Sheet1（其 name 会被忽略）。",
+              "type": "string"
+            },
+            "row_sizes": {
+              "description": "行高操作数组；range 使用行范围如 1:3，type 为 pixel/standard/auto，pixel 需要 size。",
+              "items": {
+                "properties": {
+                  "range": {
+                    "type": "string"
+                  },
+                  "size": {
+                    "type": "number"
+                  },
+                  "type": {
+                    "enum": [
+                      "pixel",
+                      "standard",
+                      "auto"
+                    ],
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "range",
+                  "type"
+                ],
+                "type": "object"
+              },
+              "type": "array"
+            }
+          },
+          "required": [
+            "name"
+          ],
+          "type": "object"
+        },
+        "type": "array"
+      }
     }
   }
 }

shortcuts/sheets/execute_paths_test.go

@@ -365,14 +365,17 @@ func TestExecute_WorkbookCreate(t *testing.T) {
 			},
 		},
 	}
-	// Initial fill first reads the workbook structure to resolve the default
-	// sheet's id (the create response doesn't echo it), then writes.
+	// The write reads the workbook structure to resolve the default sheet's id
+	// (the create response doesn't echo it). lookupFirstSheetID and
+	// writeTypedSheets' listSheetIDsByName both read it — one reusable stub serves
+	// both. The synthesized sheet is named "Sheet1", matching the default sheet,
+	// so it's adopted in place (no rename).
 	structure := toolOutputStub("shtcnBRAND", "read", `{"sheets":[{"sheet_id":"shtFirst","sheet_name":"Sheet1","index":0}]}`)
+	structure.Reusable = true
 	fill := toolOutputStub("shtcnBRAND", "write", `{"updated_cells":4}`)
 	out, err := runShortcutWithStubs(t, WorkbookCreate, []string{
 		"--title", "Sales",
-		"--headers", `["Name","Score"]`,
-		"--values", `[["alice",95]]`,
+		"--values", `[["Name","Score"],["alice",95]]`,
 	}, create, structure, fill)
 	if err != nil {
 		t.Fatalf("execute failed: %v\nout=%s", err, out)
@@ -382,8 +385,8 @@ func TestExecute_WorkbookCreate(t *testing.T) {
 	if ss["spreadsheet_token"] != "shtcnBRAND" {
 		t.Errorf("spreadsheet_token = %v", ss["spreadsheet_token"])
 	}
-	if data["initial_fill"] == nil {
-		t.Errorf("initial_fill missing in envelope")
+	if sheets, _ := data["sheets"].([]interface{}); len(sheets) != 1 {
+		t.Errorf("sheets summary missing in envelope; got %#v", data["sheets"])
 	}
 	// The fill must target the resolved first sheet, not an empty selector.
 	fillInput := decodeToolInput(t, decodeRawEnvelopeBody(t, fill.CapturedBody), "set_cell_range")
@@ -393,14 +396,13 @@ func TestExecute_WorkbookCreate(t *testing.T) {
 }
 
 // TestExecute_WorkbookCreate_EmptyArraysSkipFill locks the fix for the nil-map
-// panic / illegal-range bug: --values '[]' or --headers '[]' must short-circuit
-// the initial fill (no structure/fill calls fire) and finish with the
-// spreadsheet created but no initial_fill — never panic on a nil fill map.
+// panic / illegal-range bug: --values '[]' must short-circuit the initial fill
+// (no structure/fill calls fire) and finish with the spreadsheet created but no
+// sheets summary — never panic on a nil payload.
 func TestExecute_WorkbookCreate_EmptyArraysSkipFill(t *testing.T) {
 	t.Parallel()
 	for _, tc := range []struct{ name, flag, val string }{
 		{"empty values", "--values", "[]"},
-		{"empty headers", "--headers", "[]"},
 	} {
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
@@ -421,8 +423,8 @@ func TestExecute_WorkbookCreate_EmptyArraysSkipFill(t *testing.T) {
 				t.Fatalf("execute failed: %v\nout=%s", err, out)
 			}
 			data := decodeEnvelopeData(t, out)
-			if data["initial_fill"] != nil {
-				t.Errorf("initial_fill should be absent for %s %s; got %#v", tc.flag, tc.val, data["initial_fill"])
+			if data["sheets"] != nil {
+				t.Errorf("sheets should be absent for %s %s; got %#v", tc.flag, tc.val, data["sheets"])
 			}
 			if ss, _ := data["spreadsheet"].(map[string]interface{}); ss["spreadsheet_token"] != "shtNEW" {
 				t.Errorf("spreadsheet_token = %v, want shtNEW", ss["spreadsheet_token"])

shortcuts/sheets/flag_defs_gen.go

@@ -308,7 +308,6 @@ var flagDefs = map[string]commandDef{
 			{Name: "max-chars", Kind: "own", Type: "int", Required: "optional", Desc: "Safety cap; default 200000", Default: "200000", Hidden: true},
 			{Name: "include-row-prefix", Kind: "own", Type: "bool", Required: "optional", Desc: "Whether to prefix each row with `[row=N]`; default `true`", Default: "true"},
 			{Name: "skip-hidden", Kind: "own", Type: "bool", Required: "optional", Desc: "Skip hidden rows and columns; default `false`"},
-			{Name: "rows-json", Kind: "own", Type: "bool", Required: "optional", Desc: "Return structured rows ({row_number, values:{col→cell}}) instead of CSV text; default false", Default: "false"},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional", Desc: "Print the request path and parameters without executing"},
 		},
 	},
@@ -320,7 +319,7 @@ var flagDefs = map[string]commandDef{
 			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
 			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
 			{Name: "start-cell", Kind: "own", Type: "string", Required: "required", Desc: "Top-left A1 anchor (e.g. `A1`, `B5`; no sheet prefix — use `--sheet-id` / `--sheet-name` to select the sheet); must be a single cell, range notation not accepted; the bottom-right is inferred from CSV row/column counts", Default: "A1"},
-			{Name: "csv", Kind: "own", Type: "string", Required: "required", Desc: "RFC 4180 CSV text; plain values only (no formulas / styles / comments)", Input: []string{"file", "stdin"}},
+			{Name: "csv", Kind: "own", Type: "string", Required: "required", Desc: "RFC 4180 CSV text; values or formulas (a leading = is evaluated as a formula); no styles / comments / images (use +cells-set for those).", Input: []string{"file", "stdin"}},
 			{Name: "allow-overwrite", Kind: "own", Type: "bool", Required: "optional", Desc: "Allow overwriting (default true); set false to error if any target cell is non-empty", Default: "true"},
 			{Name: "range", Kind: "own", Type: "string", Required: "optional", Desc: "alias for --start-cell (parity with +csv-get / +cells-set, which locate with --range); a range like A1:H17 collapses to its top-left cell", Hidden: true},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
@@ -633,6 +632,35 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+history-list": {
+		Risk: "read",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "cursor", Kind: "own", Type: "string", Required: "optional", Desc: "Pagination cursor; pass the previous page's next_cursor, omit for first page"},
+			{Name: "count", Kind: "own", Type: "int", Required: "optional", Desc: "History versions per page, default 20"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
+	"+history-revert": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "revision-id", Kind: "own", Type: "string", Required: "required", Desc: "Restore the whole spreadsheet to this version: a revision_id (minor id) from +history-list"},
+			{Name: "edit-time", Kind: "own", Type: "string", Required: "optional", Desc: "The matching edit_time from the same +history-list entry; pass it to locate the version faster"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
+	"+history-revert-status": {
+		Risk: "read",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "task-id", Kind: "own", Type: "string", Required: "required", Desc: "Revert task id returned by +history-revert"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+pivot-create": {
 		Risk: "write",
 		Flags: []flagDef{
@@ -766,7 +794,7 @@ var flagDefs = map[string]commandDef{
 			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
 			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
 			{Name: "title", Kind: "own", Type: "string", Required: "required", Desc: "New sheet title"},
-			{Name: "index", Kind: "own", Type: "int", Required: "optional", Desc: "Insert position; appended to the end when omitted", Default: "-1"},
+			{Name: "index", Kind: "own", Type: "int", Required: "optional", Desc: "Insert position (0-based); appended to the end when omitted", Default: "-1"},
 			{Name: "row-count", Kind: "own", Type: "int", Required: "optional", Desc: "Initial row count (default 200, max 50000)", Default: "200"},
 			{Name: "col-count", Kind: "own", Type: "int", Required: "optional", Desc: "Initial column count (default 20, max 200)", Default: "20"},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
@@ -793,6 +821,16 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+sheet-hide-gridline": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
+			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+sheet-info": {
 		Risk: "read",
 		Flags: []flagDef{
@@ -839,6 +877,16 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+sheet-show-gridline": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "sheet-id", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet reference_id (XOR with `--sheet-name`)"},
+			{Name: "sheet-name", Kind: "public", Type: "string", Required: "xor", Desc: "Sheet name (XOR with `--sheet-id`)"},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+sheet-unhide": {
 		Risk: "write",
 		Flags: []flagDef{
@@ -895,13 +943,39 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+table-get": {
+		Risk: "read",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token (XOR with `--url`)"},
+			{Name: "sheet-id", Kind: "own", Type: "string", Required: "optional", Desc: "Read only this sheet (by id); omit to read all sheets"},
+			{Name: "sheet-name", Kind: "own", Type: "string", Required: "optional", Desc: "Read only this sheet (by name); omit to read all sheets"},
+			{Name: "range", Kind: "own", Type: "string", Required: "optional", Desc: "A1 range to read; omit to read each sheet current region"},
+			{Name: "no-header", Kind: "own", Type: "bool", Required: "optional", Desc: "Treat the first row as data instead of a header (columns get positional names col1, col2, ...)"},
+			{Name: "dataframe-out", Kind: "own", Type: "string", Required: "optional", Desc: "Write the typed table as one Arrow IPC file (Feather v2) instead of the default JSON. Pass `@<path>` for a file or `-` for binary stdout (same convention as other binary I/O flags). Mirror of the input-side `--dataframe` on `+table-put` / `+workbook-create` — pandas users round-trip via `df = pd.read_feather(\"x.arrow\")` or `pd.read_feather(io.BytesIO(stdout))`. Single-sheet only: requires `--sheet-id` or `--sheet-name`; whole-workbook reads keep the default JSON path. Column types come from the typed read-back (string/number/date/bool); per-column `number_format` is preserved as Arrow field metadata so the Arrow file can round-trip straight back through `+table-put --dataframe`."},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
+	"+table-put": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "url", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet URL to write into (XOR with `--spreadsheet-token`)"},
+			{Name: "spreadsheet-token", Kind: "public", Type: "string", Required: "xor", Desc: "Spreadsheet token to write into (XOR with `--url`)"},
+			{Name: "sheets", Kind: "own", Type: "string", Required: "xor", Desc: "Typed table payload (pandas-DataFrame-shaped) as JSON, XOR with `--dataframe`: a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it with `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. `dtypes` values are pandas dtype strings (`int64`, `float64`, `Int64`, `bool`, `boolean`, `datetime64[ns]`, `object`, ...); the writer maps them to internal string/number/date/bool — omit `dtypes` and a column writes as text (good for raw CSV-shaped data). `formats[col]` is an Excel number_format string (e.g. `#,##0.00`, `0.0%`, `yyyy-mm`); when absent, date columns default to `yyyy-mm-dd` and string columns to text format (`@`).", Input: []string{"file", "stdin"}},
+			{Name: "dataframe", Kind: "own", Type: "string", Required: "xor", Desc: "Single-sheet typed table from one Arrow IPC file (a.k.a. Feather v2 — what `pandas.DataFrame.to_feather()` writes), XOR with `--sheets`. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema (int*/uint*/float* → number, date32/date64/timestamp → date, utf8/large_utf8 → string, bool → bool); per-column `number_format` may be set via Arrow field metadata (`pa.field(\"price\", pa.float64(), metadata={b\"number_format\": b\"$#,##0.00\"})`). Writes the sheet at default placement: name `Sheet1` (created when absent), overwrite from A1 with header. For a different sheet name, anchor, mode, or to write multiple sheets, use `--sheets` instead."},
+			{Name: "styles", Kind: "own", Type: "string", Required: "optional", Desc: "Visual operations applied after the typed write, as JSON: top-level `{styles:[...]}`. Each item corresponds to one written sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. The styles array length/order/name must match the written sheets: with --sheets, match --sheets.sheets; with --dataframe (single sheet named Sheet1), pass exactly one styles item with name `Sheet1`.", Input: []string{"file", "stdin"}},
+			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
+		},
+	},
 	"+workbook-create": {
 		Risk: "write",
 		Flags: []flagDef{
 			{Name: "title", Kind: "own", Type: "string", Required: "required", Desc: "Spreadsheet title"},
 			{Name: "folder-token", Kind: "own", Type: "string", Required: "optional", Desc: "Target folder token; placed at the drive root when omitted"},
-			{Name: "headers", Kind: "own", Type: "string", Required: "optional", Desc: "Header row as a JSON array: `[\"Col A\",\"Col B\"]`", Input: []string{"file", "stdin"}},
-			{Name: "values", Kind: "own", Type: "string", Required: "optional", Desc: "Initial data as a 2D JSON array: `[[\"alice\",95]]`", Input: []string{"file", "stdin"}},
+			{Name: "values", Kind: "own", Type: "string", Required: "optional", Desc: "Untyped initial data as one 2D JSON array (`[[\"alice\",95]]`); values are written as-is with their type auto-detected, through the same batched set_cell_range path as --sheets — pair with --styles for number formats, colors, merges, and row/col sizes", Input: []string{"file", "stdin"}},
+			{Name: "sheets", Kind: "own", Type: "string", Required: "optional", Desc: "Typed table payload as JSON (same shape as `+table-put`): a top-level `sheets` array, each item `{name, start_cell?, mode?, header?, allow_overwrite?, columns:[\"colA\",\"colB\",...], data:[[...]], dtypes?:{colA:pandasDtype, ...}, formats?:{colA:numberFormat, ...}}`. Agents typically build it from a DataFrame via `{**json.loads(df.to_json(orient=\"split\")), \"dtypes\": df.dtypes.astype(str).to_dict()}`. Mutually exclusive with --values and --dataframe. Creates the workbook, then writes typed type-faithful data (dates land as real dates, numbers keep precision).", Input: []string{"file", "stdin"}},
+			{Name: "styles", Kind: "own", Type: "string", Required: "optional", Desc: "Initial visual operations as JSON: top-level `{styles:[...]}`. Each item corresponds to one target sheet and must include `name`, plus at least one of `cell_styles` / `row_sizes` / `col_sizes` / `cell_merges`. `cell_styles` entries use +cells-set-style fields with a cell range; row/col sizes use dimension ranges plus type/size; merges use cell ranges plus optional merge_type. With --sheets, styles array length/order/name must match --sheets.sheets. With --values, pass exactly one styles item for the initial sheet (its name is ignored).", Input: []string{"file", "stdin"}},
+			{Name: "dataframe", Kind: "own", Type: "string", Required: "optional", Desc: "Single-sheet typed table from one Arrow IPC file (Feather v2 — what `pandas.DataFrame.to_feather()` writes), mutually exclusive with --values and --sheets. Pass `@<path>` for a file or `-` for binary stdin (same convention as other input flags). Arrow bytes are read raw — no TrimSpace / BOM strip — so the IPC magic survives intact (unlike text input flags). Column types come from the Arrow schema; per-column `number_format` may be set via Arrow field metadata. Creates the workbook and fills its default sheet (`Sheet1` — adopted in place, no empty Sheet1 left behind). For multi-sheet or non-default placement, use `--sheets` instead."},
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
@@ -916,6 +990,14 @@ var flagDefs = map[string]commandDef{
 			{Name: "dry-run", Kind: "system", Type: "bool", Required: "optional"},
 		},
 	},
+	"+workbook-import": {
+		Risk: "write",
+		Flags: []flagDef{
+			{Name: "file", Kind: "own", Type: "string", Required: "required", Desc: "Local file path (.xlsx / .xls / .csv)"},
+			{Name: "folder-token", Kind: "own", Type: "string", Required: "optional", Desc: "Target folder token; imported to the cloud drive root when omitted"},
+			{Name: "name", Kind: "own", Type: "string", Required: "optional", Desc: "Imported spreadsheet name; defaults to the local file name without its extension"},
+		},
+	},
 	"+workbook-info": {
 		Risk: "read",
 		Flags: []flagDef{

shortcuts/sheets/flag_schema_validate.go

@@ -63,6 +63,7 @@ func validateParsedJSONFlag(fv flagView, name string, value interface{}) error {
 var parseJSONFlagSkip = map[string]struct{}{
 	"properties": {},
 	"operations": {},
+	"styles":     {},
 }
 
 // validateValueAgainstSchema is the (command, flag) → schema → check

shortcuts/sheets/flag_schemas_gen.go

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

shortcuts/sheets/lark_sheet_dataframe.go

@@ -0,0 +1,553 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"io"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/apache/arrow/go/v17/arrow"
+	"github.com/apache/arrow/go/v17/arrow/array"
+	"github.com/apache/arrow/go/v17/arrow/ipc"
+	"github.com/apache/arrow/go/v17/arrow/memory"
+
+	"github.com/larksuite/cli/extension/fileio"
+	"github.com/larksuite/cli/internal/cmdutil"
+	"github.com/larksuite/cli/shortcuts/common"
+)
+
+// ─── --dataframe (Arrow IPC / Feather v2 binary input) ────────────────
+//
+// --dataframe is the binary-typed twin of --sheets. The wire payload is one
+// Arrow IPC file (a.k.a. Feather v2 — what `pandas.DataFrame.to_feather()`
+// writes), single schema, optionally multi-batch. Type / format are read off
+// the Arrow schema (no separate dtypes/formats maps), and per-column number
+// format can be set via the field's `number_format` metadata key:
+//
+//	pa.field("price", pa.float64(), metadata={b"number_format": b"$#,##0.00"})
+//
+// One DataFrame writes into one sub-sheet at fixed defaults: name `Sheet1`
+// (adopted in place by +workbook-create; created when absent by +table-put),
+// overwrite from A1 with header on, allow_overwrite=true. The shortcut
+// surface is deliberately the one flag — anything that needs a different
+// sheet name / anchor / mode / multi-sheet falls back to --sheets, whose
+// JSON payload already carries every knob.
+//
+// Binary IO note: --dataframe bypasses the text-oriented Input resolver
+// (`runtime.Str("dataframe")` carries a *path*, not file contents). Reading
+// the Arrow bytes through that resolver would TrimSpace the trailing IPC
+// magic / corrupt non-UTF8 bytes. Path → FileIO.Open → io.ReadAll keeps the
+// stream byte-exact. "-" reads from stdin directly.
+
+// dataframeDefaultSheetName is the sub-sheet name --dataframe writes into.
+// Matches valuesSheetName so +workbook-create adopts the brand-new
+// workbook's default sheet in place (no stray empty Sheet1 left behind);
+// +table-put creates Sheet1 if it doesn't already exist.
+const dataframeDefaultSheetName = valuesSheetName
+
+// parseDataframePayload reads the --dataframe path (Arrow IPC file) and
+// composes a single-sheet tablePayload at the fixed default placement.
+// Network-free: safe from Validate and DryRun. The resulting tableSheetSpec
+// rides the same buildSheetMatrix / buildTypedCell path as a --sheets entry,
+// so downstream is unaware of where the rows came from.
+func parseDataframePayload(rctx *common.RuntimeContext) (*tablePayload, error) {
+	raw := strings.TrimSpace(rctx.Str("dataframe"))
+	if raw == "" {
+		return nil, common.FlagErrorf("--dataframe is required")
+	}
+	data, err := readDataframeBytes(rctx, raw)
+	if err != nil {
+		return nil, err
+	}
+	spec, err := decodeArrowToSheet(data, dataframeDefaultSheetName)
+	if err != nil {
+		return nil, common.FlagErrorf("--dataframe: %v", err)
+	}
+	payload := &tablePayload{Sheets: []tableSheetSpec{spec}}
+	if err := payload.validate(); err != nil {
+		return nil, err
+	}
+	return payload, nil
+}
+
+// dataframeStdinCache holds the bytes read from stdin on the first call so a
+// later call (Validate → Execute / DryRun) gets the same bytes instead of an
+// empty stream — stdin is single-shot, but parseDataframePayload runs
+// multiple times per command invocation. Process-wide is fine: lark-cli is
+// one-shot (one command per process). Tests reset by setting it back to nil.
+var dataframeStdinCache []byte
+
+// readDataframeBytes resolves --dataframe to raw binary. A literal `@` prefix
+// is tolerated for symmetry with --sheets (`@/tmp/x.arrow` and `/tmp/x.arrow`
+// both work). `-` reads stdin verbatim — cached on first call so Validate /
+// Execute / DryRun all see the same bytes. Bytes are returned untouched: no
+// TrimSpace, no BOM strip — both would corrupt an Arrow IPC stream.
+func readDataframeBytes(rctx *common.RuntimeContext, raw string) ([]byte, error) {
+	if raw == "-" {
+		if dataframeStdinCache != nil {
+			return dataframeStdinCache, nil
+		}
+		io := rctx.IO()
+		if io == nil || io.In == nil {
+			return nil, common.FlagErrorf("--dataframe: stdin is not available")
+		}
+		data, err := readAllBytes(io.In)
+		if err != nil {
+			return nil, common.FlagErrorf("--dataframe: read stdin: %v", err)
+		}
+		if len(data) == 0 {
+			return nil, common.FlagErrorf("--dataframe: stdin is empty")
+		}
+		dataframeStdinCache = data
+		return data, nil
+	}
+	path := strings.TrimPrefix(raw, "@")
+	data, err := cmdutil.ReadInputFile(rctx.FileIO(), path)
+	if err != nil {
+		return nil, common.FlagErrorf("--dataframe: %v", err)
+	}
+	if len(data) == 0 {
+		return nil, common.FlagErrorf("--dataframe: file %q is empty", path)
+	}
+	return data, nil
+}
+
+// readAllBytes is a thin wrapper so tests can fake the io.Reader without
+// importing io. Mirrors io.ReadAll exactly.
+func readAllBytes(r io.Reader) ([]byte, error) { return io.ReadAll(r) }
+
+// decodeArrowToSheet reads `data` as an Arrow IPC file (single schema,
+// possibly multi-batch) and produces a tableSheetSpec with name + columns +
+// rows filled in. Sheet placement (start_cell / mode / header / overwrite) is
+// not touched here — parseDataframePayload layers those on from CLI flags.
+func decodeArrowToSheet(data []byte, sheetName string) (tableSheetSpec, error) {
+	reader, err := ipc.NewFileReader(bytes.NewReader(data))
+	if err != nil {
+		return tableSheetSpec{}, fmt.Errorf("invalid Arrow IPC file (expected pandas df.to_feather output): %v", err)
+	}
+	defer reader.Close()
+
+	schema := reader.Schema()
+	if schema == nil || schema.NumFields() == 0 {
+		return tableSheetSpec{}, fmt.Errorf("Arrow schema has no fields")
+	}
+
+	ncols := schema.NumFields()
+	cols := make([]tableColumnSpec, ncols)
+	seen := make(map[string]bool, ncols)
+	for i := 0; i < ncols; i++ {
+		f := schema.Field(i)
+		name := f.Name
+		if strings.TrimSpace(name) == "" {
+			return tableSheetSpec{}, fmt.Errorf("column %d has empty name", i)
+		}
+		if seen[name] {
+			return tableSheetSpec{}, fmt.Errorf("duplicate column name %q", name)
+		}
+		seen[name] = true
+		typ, format, err := arrowFieldToTypeFormat(f)
+		if err != nil {
+			return tableSheetSpec{}, fmt.Errorf("column %q: %v", name, err)
+		}
+		cols[i] = tableColumnSpec{Name: name, Type: typ, Format: format}
+	}
+
+	var rows [][]interface{}
+	for b := 0; b < reader.NumRecords(); b++ {
+		rec, err := reader.RecordAt(b)
+		if err != nil {
+			return tableSheetSpec{}, fmt.Errorf("read record batch %d: %v", b, err)
+		}
+		batchRows, err := arrowRecordToRows(rec, cols)
+		rec.Release()
+		if err != nil {
+			return tableSheetSpec{}, err
+		}
+		rows = append(rows, batchRows...)
+	}
+
+	return tableSheetSpec{Name: sheetName, Columns: cols, Rows: rows}, nil
+}
+
+// arrowFieldToTypeFormat maps an Arrow field to the internal (type, format)
+// pair. The field's `number_format` metadata key — when present — sets the
+// Excel number_format string verbatim; otherwise sensible defaults are
+// applied per type (`@` text for strings, `yyyy-mm-dd` for dates).
+func arrowFieldToTypeFormat(f arrow.Field) (typ, format string, err error) {
+	if v, ok := f.Metadata.GetValue("number_format"); ok {
+		format = strings.TrimSpace(v)
+	}
+	switch f.Type.(type) {
+	case *arrow.StringType, *arrow.LargeStringType:
+		if format == "" {
+			format = "@"
+		}
+		return "string", format, nil
+	case *arrow.BooleanType:
+		return "bool", format, nil
+	case *arrow.Date32Type, *arrow.Date64Type, *arrow.TimestampType:
+		if format == "" {
+			format = "yyyy-mm-dd"
+		}
+		return "date", format, nil
+	}
+	if isArrowNumericType(f.Type) {
+		return "number", format, nil
+	}
+	return "", "", fmt.Errorf("unsupported Arrow type %s (want string/number/date/bool)", f.Type.Name())
+}
+
+func isArrowNumericType(t arrow.DataType) bool {
+	switch t.ID() {
+	case arrow.INT8, arrow.INT16, arrow.INT32, arrow.INT64,
+		arrow.UINT8, arrow.UINT16, arrow.UINT32, arrow.UINT64,
+		arrow.FLOAT16, arrow.FLOAT32, arrow.FLOAT64:
+		return true
+	}
+	return false
+}
+
+// arrowRecordToRows transposes one column-batch into row-major
+// [][]interface{} matched to `cols`. Cells are stamped with the same value
+// shapes buildTypedCell expects from the JSON path: nil for nulls,
+// json.Number for numerics (precision-preserving), `yyyy-mm-dd` strings for
+// dates/timestamps, bool for booleans, string for strings.
+func arrowRecordToRows(rec arrow.Record, cols []tableColumnSpec) ([][]interface{}, error) {
+	if int(rec.NumCols()) != len(cols) {
+		return nil, fmt.Errorf("record has %d cols, schema declared %d", rec.NumCols(), len(cols))
+	}
+	nrows := int(rec.NumRows())
+	rows := make([][]interface{}, nrows)
+	for r := range rows {
+		rows[r] = make([]interface{}, len(cols))
+	}
+	for c := range cols {
+		arr := rec.Column(c)
+		for r := 0; r < nrows; r++ {
+			if arr.IsNull(r) {
+				continue
+			}
+			v, err := arrowCellValue(arr, r)
+			if err != nil {
+				return nil, fmt.Errorf("row %d column %q: %v", r, cols[c].Name, err)
+			}
+			rows[r][c] = v
+		}
+	}
+	return rows, nil
+}
+
+func arrowCellValue(arr arrow.Array, i int) (interface{}, error) {
+	switch a := arr.(type) {
+	case *array.String:
+		return a.Value(i), nil
+	case *array.LargeString:
+		return a.Value(i), nil
+	case *array.Boolean:
+		return a.Value(i), nil
+	case *array.Int8:
+		return json.Number(strconv.FormatInt(int64(a.Value(i)), 10)), nil
+	case *array.Int16:
+		return json.Number(strconv.FormatInt(int64(a.Value(i)), 10)), nil
+	case *array.Int32:
+		return json.Number(strconv.FormatInt(int64(a.Value(i)), 10)), nil
+	case *array.Int64:
+		return json.Number(strconv.FormatInt(a.Value(i), 10)), nil
+	case *array.Uint8:
+		return json.Number(strconv.FormatUint(uint64(a.Value(i)), 10)), nil
+	case *array.Uint16:
+		return json.Number(strconv.FormatUint(uint64(a.Value(i)), 10)), nil
+	case *array.Uint32:
+		return json.Number(strconv.FormatUint(uint64(a.Value(i)), 10)), nil
+	case *array.Uint64:
+		return json.Number(strconv.FormatUint(a.Value(i), 10)), nil
+	case *array.Float16:
+		return json.Number(strconv.FormatFloat(float64(a.Value(i).Float32()), 'f', -1, 32)), nil
+	case *array.Float32:
+		return json.Number(strconv.FormatFloat(float64(a.Value(i)), 'f', -1, 32)), nil
+	case *array.Float64:
+		return json.Number(strconv.FormatFloat(a.Value(i), 'f', -1, 64)), nil
+	case *array.Date32:
+		// Date32: days since 1970-01-01 (epoch). Multiply to seconds, format
+		// in UTC so timezone offset can't flip the calendar date.
+		t := time.Unix(int64(a.Value(i))*86400, 0).UTC()
+		return t.Format("2006-01-02"), nil
+	case *array.Date64:
+		t := time.UnixMilli(int64(a.Value(i))).UTC()
+		return t.Format("2006-01-02"), nil
+	case *array.Timestamp:
+		ts := int64(a.Value(i))
+		unit := a.DataType().(*arrow.TimestampType).Unit
+		var t time.Time
+		switch unit {
+		case arrow.Second:
+			t = time.Unix(ts, 0).UTC()
+		case arrow.Millisecond:
+			t = time.UnixMilli(ts).UTC()
+		case arrow.Microsecond:
+			t = time.UnixMicro(ts).UTC()
+		case arrow.Nanosecond:
+			t = time.Unix(0, ts).UTC()
+		default:
+			return nil, fmt.Errorf("unsupported timestamp unit %v", unit)
+		}
+		return t.Format("2006-01-02"), nil
+	}
+	return nil, fmt.Errorf("unsupported Arrow array %T", arr)
+}
+
+// ─── --dataframe-out (Arrow IPC binary output, mirror of --dataframe) ──
+//
+// +table-get's binary read-back: encode one sheet's typed read-back as an
+// Arrow IPC file (Feather v2), so pandas can `pd.read_feather(path)` /
+// `pd.read_feather(BytesIO(stdout))` symmetrically with the put side.
+// Single-sheet only — Arrow IPC carries one schema per file. The JSON path
+// is unchanged; --dataframe-out swaps the encoder for callers that already
+// have pandas / pyarrow in their pipeline.
+
+// encodeSheetMapToArrowIPC turns one readSheetAsSpec output into an Arrow IPC
+// file blob. Internal column types are recovered from `dtypes` (the wire
+// proxy for the typed protocol), and per-column `number_format` rides through
+// as Arrow field metadata so the file feeds straight back into
+// `+table-put --dataframe`.
+func encodeSheetMapToArrowIPC(sheet map[string]interface{}) ([]byte, error) {
+	columns, _ := sheet["columns"].([]interface{})
+	if len(columns) == 0 {
+		return nil, fmt.Errorf("sheet has no columns")
+	}
+	dtypes, _ := sheet["dtypes"].(map[string]interface{})
+	formats, _ := sheet["formats"].(map[string]interface{})
+	// `data` arrives as either []interface{} (when the sheet came through a
+	// JSON round-trip / unit-test fixture) or [][]interface{} (the shape
+	// readSheetAsSpec directly emits in production). Accept both — anything
+	// else falls through to a zero-row table.
+	var rawData [][]interface{}
+	switch d := sheet["data"].(type) {
+	case [][]interface{}:
+		rawData = d
+	case []interface{}:
+		rawData = make([][]interface{}, len(d))
+		for i, r := range d {
+			rawData[i], _ = r.([]interface{})
+		}
+	}
+
+	ncols := len(columns)
+	colNames := make([]string, ncols)
+	colTypes := make([]string, ncols)
+	fields := make([]arrow.Field, ncols)
+	for i, c := range columns {
+		name, _ := c.(string)
+		if name == "" {
+			return nil, fmt.Errorf("column %d has empty name", i)
+		}
+		colNames[i] = name
+		dt, _ := dtypes[name].(string)
+		colTypes[i] = dtypeToInternalType(dt)
+		var meta arrow.Metadata
+		if formats != nil {
+			if nf, ok := formats[name].(string); ok && strings.TrimSpace(nf) != "" {
+				meta = arrow.NewMetadata([]string{"number_format"}, []string{nf})
+			}
+		}
+		fields[i] = arrow.Field{
+			Name:     name,
+			Type:     internalTypeToArrowType(colTypes[i]),
+			Nullable: true,
+			Metadata: meta,
+		}
+	}
+	schema := arrow.NewSchema(fields, nil)
+
+	mem := memory.NewGoAllocator()
+	rb := array.NewRecordBuilder(mem, schema)
+	defer rb.Release()
+	for r, row := range rawData {
+		if len(row) != ncols {
+			return nil, fmt.Errorf("row %d has %d cells, want %d", r, len(row), ncols)
+		}
+		for c := 0; c < ncols; c++ {
+			if err := appendArrowCell(rb.Field(c), colTypes[c], row[c]); err != nil {
+				return nil, fmt.Errorf("row %d column %q: %v", r, colNames[c], err)
+			}
+		}
+	}
+	rec := rb.NewRecord()
+	defer rec.Release()
+
+	var buf bytesWriterSeeker
+	w, err := ipc.NewFileWriter(&buf, ipc.WithSchema(schema), ipc.WithAllocator(mem))
+	if err != nil {
+		return nil, fmt.Errorf("ipc.NewFileWriter: %v", err)
+	}
+	if err := w.Write(rec); err != nil {
+		return nil, fmt.Errorf("write record: %v", err)
+	}
+	if err := w.Close(); err != nil {
+		return nil, fmt.Errorf("close writer: %v", err)
+	}
+	return buf.buf, nil
+}
+
+// dtypeToInternalType inverts typeToDtype so the Arrow encoder can pick an
+// internal column type from the wire-level dtype string. Unknown / object
+// falls back to string (lossless: every cell is already typed as such).
+func dtypeToInternalType(dtype string) string {
+	switch strings.ToLower(strings.TrimSpace(dtype)) {
+	case "float64", "float32", "int64", "int32", "int16", "int8",
+		"uint64", "uint32", "uint16", "uint8":
+		return "number"
+	case "bool", "boolean":
+		return "bool"
+	}
+	if strings.HasPrefix(strings.ToLower(dtype), "datetime") {
+		return "date"
+	}
+	return "string"
+}
+
+// internalTypeToArrowType is the put-side dtypeToTypeFormat dual: maps the
+// internal column type to the Arrow data type the encoder builds a column
+// with. Numbers go to float64 because +table-get can't tell int from float
+// from a number_format alone — float64 covers both losslessly for the cell
+// ranges Lark Sheets accepts.
+func internalTypeToArrowType(typ string) arrow.DataType {
+	switch typ {
+	case "number":
+		return arrow.PrimitiveTypes.Float64
+	case "date":
+		return arrow.FixedWidthTypes.Date32
+	case "bool":
+		return arrow.FixedWidthTypes.Boolean
+	}
+	return arrow.BinaryTypes.String
+}
+
+// appendArrowCell stamps one cell into its column builder. Cell shape matches
+// what cellToTyped emits on the JSON path: json.Number for numbers, ISO
+// `yyyy-mm-dd` string for dates, plain string for strings, bool for bools,
+// nil for empty. Anything off-shape errors so the caller doesn't silently
+// emit nulls for malformed data.
+func appendArrowCell(b array.Builder, typ string, v interface{}) error {
+	if v == nil {
+		b.AppendNull()
+		return nil
+	}
+	switch typ {
+	case "string":
+		s, ok := v.(string)
+		if !ok {
+			return fmt.Errorf("string expects string value, got %T", v)
+		}
+		b.(*array.StringBuilder).Append(s)
+	case "number":
+		f, err := arrowNumber(v)
+		if err != nil {
+			return err
+		}
+		b.(*array.Float64Builder).Append(f)
+	case "date":
+		s, ok := v.(string)
+		if !ok {
+			return fmt.Errorf("date expects ISO yyyy-mm-dd string, got %T", v)
+		}
+		t, err := time.Parse("2006-01-02", strings.TrimSpace(s))
+		if err != nil {
+			return fmt.Errorf("date parse %q: %v", s, err)
+		}
+		b.(*array.Date32Builder).Append(arrow.Date32FromTime(t))
+	case "bool":
+		bb, ok := v.(bool)
+		if !ok {
+			return fmt.Errorf("bool expects bool, got %T", v)
+		}
+		b.(*array.BooleanBuilder).Append(bb)
+	default:
+		return fmt.Errorf("unsupported internal type %q", typ)
+	}
+	return nil
+}
+
+// arrowNumber converts the number cell shape readSheetAsSpec emits
+// (json.Number) plus the float fallback to float64 for the Arrow builder.
+func arrowNumber(v interface{}) (float64, error) {
+	switch n := v.(type) {
+	case json.Number:
+		f, err := n.Float64()
+		if err != nil {
+			return 0, fmt.Errorf("number parse %q: %v", n.String(), err)
+		}
+		return f, nil
+	case float64:
+		return n, nil
+	}
+	return 0, fmt.Errorf("number expects numeric value, got %T", v)
+}
+
+// bytesWriterSeeker is a 10-line in-memory io.WriteSeeker for
+// ipc.NewFileWriter, which seeks back to patch a footer offset. Using a
+// buffer (instead of a temp file or os.Stdout, which isn't seekable) keeps
+// --dataframe-out's stdout path zero-IO and stays straightforward.
+type bytesWriterSeeker struct {
+	buf []byte
+	pos int64
+}
+
+func (w *bytesWriterSeeker) Write(p []byte) (int, error) {
+	end := w.pos + int64(len(p))
+	if end > int64(len(w.buf)) {
+		w.buf = append(w.buf, make([]byte, end-int64(len(w.buf)))...)
+	}
+	n := copy(w.buf[w.pos:], p)
+	w.pos = end
+	return n, nil
+}
+
+func (w *bytesWriterSeeker) Seek(offset int64, whence int) (int64, error) {
+	switch whence {
+	case io.SeekStart:
+		w.pos = offset
+	case io.SeekCurrent:
+		w.pos += offset
+	case io.SeekEnd:
+		w.pos = int64(len(w.buf)) + offset
+	default:
+		return 0, fmt.Errorf("unknown whence %d", whence)
+	}
+	return w.pos, nil
+}
+
+// writeDataframeOut dispatches the encoded Arrow bytes to wherever --dataframe-out
+// points: `-` → process stdout, `@<path>` or plain path → local file. Symmetric
+// with readDataframeBytes on the input side: same `@` tolerance, same TrimPrefix
+// semantics, and an absolute path will still get rejected by FileIO's SafePath.
+func writeDataframeOut(rctx *common.RuntimeContext, raw string, data []byte) error {
+	if raw == "-" {
+		out := rctx.IO()
+		if out == nil || out.Out == nil {
+			return common.FlagErrorf("--dataframe-out: stdout is not available")
+		}
+		if _, err := out.Out.Write(data); err != nil {
+			return fmt.Errorf("--dataframe-out: write stdout: %v", err)
+		}
+		return nil
+	}
+	path := strings.TrimPrefix(raw, "@")
+	fio := rctx.FileIO()
+	if fio == nil {
+		return common.FlagErrorf("--dataframe-out: file output is not available in this context")
+	}
+	// FileIO.Save validates the path via SafeOutputPath (the same sandbox
+	// readDataframeBytes hits on the input side) and writes atomically, so we
+	// don't need an extra ValidatePath call here.
+	if _, err := fio.Save(path, fileio.SaveOptions{ContentLength: int64(len(data))}, bytes.NewReader(data)); err != nil {
+		return fmt.Errorf("--dataframe-out: write %q: %v", path, err)
+	}
+	return nil
+}

shortcuts/sheets/lark_sheet_dataframe_test.go

@@ -0,0 +1,378 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/apache/arrow/go/v17/arrow"
+	"github.com/apache/arrow/go/v17/arrow/array"
+	"github.com/apache/arrow/go/v17/arrow/ipc"
+	"github.com/apache/arrow/go/v17/arrow/memory"
+)
+
+// buildArrowIPC writes one record into a Feather v2 (Arrow IPC file) blob.
+// Used by the round-trip tests below to stand in for what
+// `pandas.DataFrame.to_feather(path)` would produce; saves the tests from
+// depending on a pandas-shaped fixture file.
+//
+// ipc.NewFileWriter wants an io.WriteSeeker (it back-patches a footer
+// offset), so we write to a temp file and read the bytes back — simpler than
+// re-implementing a seekable in-memory buffer.
+func buildArrowIPC(t *testing.T, schema *arrow.Schema, build func(b *array.RecordBuilder)) []byte {
+	t.Helper()
+	mem := memory.NewGoAllocator()
+	rb := array.NewRecordBuilder(mem, schema)
+	defer rb.Release()
+	build(rb)
+	rec := rb.NewRecord()
+	defer rec.Release()
+
+	path := filepath.Join(t.TempDir(), "df.arrow")
+	f, err := os.Create(path)
+	if err != nil {
+		t.Fatalf("create temp arrow file: %v", err)
+	}
+	w, err := ipc.NewFileWriter(f, ipc.WithSchema(schema), ipc.WithAllocator(mem))
+	if err != nil {
+		f.Close()
+		t.Fatalf("ipc.NewFileWriter: %v", err)
+	}
+	if err := w.Write(rec); err != nil {
+		t.Fatalf("write record: %v", err)
+	}
+	if err := w.Close(); err != nil {
+		t.Fatalf("close writer: %v", err)
+	}
+	if err := f.Close(); err != nil {
+		t.Fatalf("close file: %v", err)
+	}
+	data, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("read temp arrow file: %v", err)
+	}
+	return data
+}
+
+// TestDataframe_RoundTripCoreTypes pins down the Arrow-schema → internal
+// (type, format) mapping and the per-cell value shape that buildTypedCell
+// expects: number cells are json.Number (precision-preserving), date cells
+// are `yyyy-mm-dd` strings, bool/string come through verbatim. Numbers, dates,
+// strings, bools, and nulls all in one record so a future Arrow-Go bump can't
+// quietly regress any one family.
+func TestDataframe_RoundTripCoreTypes(t *testing.T) {
+	t.Parallel()
+
+	schema := arrow.NewSchema([]arrow.Field{
+		{Name: "name", Type: arrow.BinaryTypes.String},
+		{Name: "qty", Type: arrow.PrimitiveTypes.Int64},
+		{Name: "price", Type: arrow.PrimitiveTypes.Float64, Metadata: arrow.NewMetadata(
+			[]string{"number_format"}, []string{"$#,##0.00"},
+		)},
+		{Name: "active", Type: arrow.FixedWidthTypes.Boolean},
+		{Name: "shipped_on", Type: arrow.FixedWidthTypes.Date32},
+	}, nil)
+
+	jan15 := arrow.Date32FromTime(time.Date(2024, 1, 15, 0, 0, 0, 0, time.UTC))
+	feb02 := arrow.Date32FromTime(time.Date(2024, 2, 2, 0, 0, 0, 0, time.UTC))
+
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		b.Field(0).(*array.StringBuilder).AppendValues([]string{"alice", ""}, []bool{true, false})
+		b.Field(1).(*array.Int64Builder).AppendValues([]int64{42, 0}, []bool{true, false})
+		b.Field(2).(*array.Float64Builder).AppendValues([]float64{19.95, 0}, []bool{true, false})
+		b.Field(3).(*array.BooleanBuilder).AppendValues([]bool{true, false}, []bool{true, true})
+		b.Field(4).(*array.Date32Builder).AppendValues([]arrow.Date32{jan15, feb02}, []bool{true, true})
+	})
+
+	spec, err := decodeArrowToSheet(buf, "S1")
+	if err != nil {
+		t.Fatalf("decodeArrowToSheet: %v", err)
+	}
+	if spec.Name != "S1" {
+		t.Errorf("sheet name = %q, want S1", spec.Name)
+	}
+	if len(spec.Columns) != 5 {
+		t.Fatalf("got %d columns, want 5", len(spec.Columns))
+	}
+	want := []struct{ typ, format string }{
+		{"string", "@"},
+		{"number", ""},
+		{"number", "$#,##0.00"},
+		{"bool", ""},
+		{"date", "yyyy-mm-dd"},
+	}
+	for i, w := range want {
+		if spec.Columns[i].Type != w.typ {
+			t.Errorf("columns[%d].Type = %q, want %q", i, spec.Columns[i].Type, w.typ)
+		}
+		if spec.Columns[i].Format != w.format {
+			t.Errorf("columns[%d].Format = %q, want %q", i, spec.Columns[i].Format, w.format)
+		}
+	}
+
+	if len(spec.Rows) != 2 {
+		t.Fatalf("got %d rows, want 2", len(spec.Rows))
+	}
+	// Row 0: every field present, types match what buildTypedCell will accept.
+	row0 := spec.Rows[0]
+	if row0[0] != "alice" {
+		t.Errorf("row0[name] = %#v, want \"alice\"", row0[0])
+	}
+	if n, ok := row0[1].(json.Number); !ok || n.String() != "42" {
+		t.Errorf("row0[qty] = %#v, want json.Number(\"42\")", row0[1])
+	}
+	if n, ok := row0[2].(json.Number); !ok || n.String() != "19.95" {
+		t.Errorf("row0[price] = %#v, want json.Number(\"19.95\")", row0[2])
+	}
+	if row0[3] != true {
+		t.Errorf("row0[active] = %#v, want true", row0[3])
+	}
+	if row0[4] != "2024-01-15" {
+		t.Errorf("row0[shipped_on] = %#v, want \"2024-01-15\"", row0[4])
+	}
+
+	// Row 1: nulls on name/qty/price (despite the buffer values) must become nil
+	// so buildTypedCell paints an empty cell that still carries number_format.
+	row1 := spec.Rows[1]
+	for _, c := range []int{0, 1, 2} {
+		if row1[c] != nil {
+			t.Errorf("row1[%d] = %#v, want nil (null in arrow)", c, row1[c])
+		}
+	}
+	if row1[3] != false {
+		t.Errorf("row1[active] = %#v, want false", row1[3])
+	}
+	if row1[4] != "2024-02-02" {
+		t.Errorf("row1[shipped_on] = %#v, want \"2024-02-02\"", row1[4])
+	}
+}
+
+// TestDataframe_Timestamp pins the timestamp → date conversion for the
+// timestamp[us] case (pandas default for `pd.Timestamp` columns once written
+// via `to_feather`). Only the calendar date matters for our `yyyy-mm-dd`
+// landing — guard against TZ drift from the wrong unit pick.
+func TestDataframe_Timestamp(t *testing.T) {
+	t.Parallel()
+	schema := arrow.NewSchema([]arrow.Field{
+		{Name: "ts", Type: &arrow.TimestampType{Unit: arrow.Microsecond}},
+	}, nil)
+	ts := arrow.Timestamp(time.Date(2024, 6, 12, 14, 30, 0, 0, time.UTC).UnixMicro())
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		b.Field(0).(*array.TimestampBuilder).AppendValues([]arrow.Timestamp{ts}, []bool{true})
+	})
+	spec, err := decodeArrowToSheet(buf, "S")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if spec.Columns[0].Type != "date" {
+		t.Errorf("type = %q, want date", spec.Columns[0].Type)
+	}
+	if got := spec.Rows[0][0]; got != "2024-06-12" {
+		t.Errorf("ts = %#v, want \"2024-06-12\"", got)
+	}
+}
+
+// TestDataframe_EmptySchema rejects an Arrow file whose schema has no fields:
+// a 0-column "DataFrame" would write a header-less, data-less block that
+// validates as "writer ran successfully" but produces nothing — the test ties
+// that off as an explicit error rather than letting it slip through.
+func TestDataframe_EmptySchema(t *testing.T) {
+	t.Parallel()
+	schema := arrow.NewSchema(nil, nil)
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {})
+	_, err := decodeArrowToSheet(buf, "S")
+	if err == nil || !strings.Contains(err.Error(), "no fields") {
+		t.Errorf("err = %v, want 'no fields' error", err)
+	}
+}
+
+// TestDataframe_DuplicateColumn catches duplicate-name columns at decode
+// time. Validate already rejects duplicate column names for the JSON path;
+// the Arrow path mirrors that so the error surfaces with the same shape.
+func TestDataframe_DuplicateColumn(t *testing.T) {
+	t.Parallel()
+	schema := arrow.NewSchema([]arrow.Field{
+		{Name: "x", Type: arrow.BinaryTypes.String},
+		{Name: "x", Type: arrow.PrimitiveTypes.Int64},
+	}, nil)
+	buf := buildArrowIPC(t, schema, func(b *array.RecordBuilder) {
+		b.Field(0).(*array.StringBuilder).Append("")
+		b.Field(1).(*array.Int64Builder).Append(0)
+	})
+	_, err := decodeArrowToSheet(buf, "S")
+	if err == nil || !strings.Contains(err.Error(), "duplicate") {
+		t.Errorf("err = %v, want duplicate-column error", err)
+	}
+}
+
+// TestDataframe_BadBytes rejects a non-Arrow blob with a hint pointing at
+// pandas df.to_feather so users see what producer is expected without having
+// to grep the docs.
+func TestDataframe_BadBytes(t *testing.T) {
+	t.Parallel()
+	_, err := decodeArrowToSheet([]byte("not arrow"), "S")
+	if err == nil || !strings.Contains(err.Error(), "Arrow") {
+		t.Errorf("err = %v, want Arrow-decode error", err)
+	}
+}
+
+// TestDataframe_EncodeRoundTrip checks --dataframe-out's encoder against its
+// own decoder: build a +table-get-shaped sheet map (the same one
+// readSheetAsSpec emits), encode to Arrow IPC, decode back via the put-side
+// decoder, and require the column types / formats / row values to match. If
+// any encoder choice drifts from what the decoder expects, the round-trip
+// breaks here long before a real put → get round-trip in production would.
+func TestDataframe_EncodeRoundTrip(t *testing.T) {
+	t.Parallel()
+	sheet := map[string]interface{}{
+		"name":    "S1",
+		"columns": []interface{}{"name", "qty", "price", "active", "ts"},
+		"dtypes": map[string]interface{}{
+			"name":   "object",
+			"qty":    "float64",
+			"price":  "float64",
+			"active": "bool",
+			"ts":     "datetime64[ns]",
+		},
+		"formats": map[string]interface{}{
+			// `@` is the writer convention for string columns; readSheetAsSpec
+			// strips it via isTextNumberFormat, so an Arrow file built from a
+			// real read won't carry @ either. Keep it absent here to mirror
+			// the production wire shape.
+			"price": "$#,##0.00",
+		},
+		"data": []interface{}{
+			[]interface{}{"alice", json.Number("42"), json.Number("19.95"), true, "2024-01-15"},
+			[]interface{}{"bob", nil, json.Number("8.5"), false, "2024-02-02"},
+		},
+	}
+	blob, err := encodeSheetMapToArrowIPC(sheet)
+	if err != nil {
+		t.Fatalf("encodeSheetMapToArrowIPC: %v", err)
+	}
+	spec, err := decodeArrowToSheet(blob, "S1")
+	if err != nil {
+		t.Fatalf("decodeArrowToSheet: %v", err)
+	}
+	wantTypes := []string{"string", "number", "number", "bool", "date"}
+	wantFormats := []string{"@", "", "$#,##0.00", "", "yyyy-mm-dd"}
+	if len(spec.Columns) != len(wantTypes) {
+		t.Fatalf("got %d columns, want %d", len(spec.Columns), len(wantTypes))
+	}
+	for i, w := range wantTypes {
+		if spec.Columns[i].Type != w {
+			t.Errorf("columns[%d].Type = %q, want %q", i, spec.Columns[i].Type, w)
+		}
+		if spec.Columns[i].Format != wantFormats[i] {
+			t.Errorf("columns[%d].Format = %q, want %q", i, spec.Columns[i].Format, wantFormats[i])
+		}
+	}
+	if len(spec.Rows) != 2 {
+		t.Fatalf("got %d rows, want 2", len(spec.Rows))
+	}
+	if spec.Rows[0][0] != "alice" {
+		t.Errorf("row0[name] = %#v, want alice", spec.Rows[0][0])
+	}
+	if n, ok := spec.Rows[0][1].(json.Number); !ok || n.String() != "42" {
+		t.Errorf("row0[qty] = %#v, want json.Number(\"42\")", spec.Rows[0][1])
+	}
+	if spec.Rows[0][3] != true {
+		t.Errorf("row0[active] = %#v, want true", spec.Rows[0][3])
+	}
+	if spec.Rows[0][4] != "2024-01-15" {
+		t.Errorf("row0[ts] = %#v, want 2024-01-15", spec.Rows[0][4])
+	}
+	// qty is null on row1, must come back as nil (not a zero-valued
+	// json.Number that would later round-trip as 0).
+	if spec.Rows[1][1] != nil {
+		t.Errorf("row1[qty] = %#v, want nil (null arrow cell)", spec.Rows[1][1])
+	}
+}
+
+// TestDataframe_EncodeAcceptsBothRowShapes pins the encoder against the two
+// shapes `sheet["data"]` actually arrives in: `[][]interface{}` from a live
+// readSheetAsSpec call (production), and `[]interface{}` from a JSON
+// unmarshal (round-trip / fixtures). Either must produce non-empty Arrow
+// output — early on the production shape silently fell through the
+// `[]interface{}` type assertion and we shipped a 0-row Arrow blob.
+func TestDataframe_EncodeAcceptsBothRowShapes(t *testing.T) {
+	t.Parallel()
+	base := func(data interface{}) map[string]interface{} {
+		return map[string]interface{}{
+			"name":    "S",
+			"columns": []interface{}{"city"},
+			"dtypes":  map[string]interface{}{"city": "object"},
+			"data":    data,
+		}
+	}
+	for label, data := range map[string]interface{}{
+		"production [][]interface{}": [][]interface{}{{"BJ"}, {"SH"}},
+		"unmarshal []interface{}":    []interface{}{[]interface{}{"BJ"}, []interface{}{"SH"}},
+	} {
+		blob, err := encodeSheetMapToArrowIPC(base(data))
+		if err != nil {
+			t.Errorf("%s: encode: %v", label, err)
+			continue
+		}
+		spec, err := decodeArrowToSheet(blob, "S")
+		if err != nil {
+			t.Errorf("%s: decode: %v", label, err)
+			continue
+		}
+		if len(spec.Rows) != 2 {
+			t.Errorf("%s: got %d rows, want 2", label, len(spec.Rows))
+		}
+	}
+}
+
+// TestDataframe_DtypeToInternalType pins the inverse of typeToDtype so
+// readSheetAsSpec's dtype labels recover the right internal type. Covers the
+// dtype families +table-get emits today plus the safe fallback for unknown
+// labels (string, lossless).
+func TestDataframe_DtypeToInternalType(t *testing.T) {
+	t.Parallel()
+	cases := map[string]string{
+		"float64":         "number",
+		"int64":           "number",
+		"Int64":           "number",
+		"bool":            "bool",
+		"boolean":         "bool",
+		"datetime64[ns]":  "date",
+		"datetime64[ms]":  "date",
+		"object":          "string",
+		"":                "string",
+		"weird-new-dtype": "string",
+	}
+	for in, want := range cases {
+		if got := dtypeToInternalType(in); got != want {
+			t.Errorf("dtypeToInternalType(%q) = %q, want %q", in, got, want)
+		}
+	}
+}
+
+// TestDataframe_BytesWriterSeeker confirms the in-memory WriteSeeker handles
+// the Seek-and-overwrite pattern ipc.NewFileWriter uses to patch the footer
+// offset: write some bytes, seek back to the middle, overwrite, end up with
+// the buffer reflecting the overwritten bytes (not a tail-extended duplicate).
+func TestDataframe_BytesWriterSeeker(t *testing.T) {
+	t.Parallel()
+	var w bytesWriterSeeker
+	if _, err := w.Write([]byte("hello world")); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := w.Seek(6, 0); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := w.Write([]byte("WORLD")); err != nil {
+		t.Fatal(err)
+	}
+	if got := string(w.buf); got != "hello WORLD" {
+		t.Errorf("buf = %q, want \"hello WORLD\"", got)
+	}
+}

shortcuts/sheets/lark_sheet_history.go

@@ -0,0 +1,197 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"context"
+	"strings"
+
+	"github.com/larksuite/cli/shortcuts/common"
+)
+
+// ─── lark_sheet_history ───────────────────────────────────────────────
+//
+// Wraps:
+//   - list_history_versions (read)  — powers +history-list
+//   - revert_to_revision    (write) — powers +history-revert
+//   - get_revert_status     (read)  — powers +history-revert-status
+//
+// The version-history "方案 B": list the spreadsheet's saved revisions, submit
+// a whole-document revert to one of them, then poll the async revert task for
+// completion. The facade gateway owns the work; the CLI only forwards the
+// target revision (and later the task id) and the server does the rest.
+//
+// ⚠️ Full-table overwrite: +history-revert rolls the WHOLE spreadsheet back to
+// the target revision, discarding every change made afterwards — including
+// other collaborators' (and the web UI's) edits. Use it only on agent scratch
+// spreadsheets, or when a whole-document rollback is acceptable.
+
+// HistoryList wraps list_history_versions: page through the spreadsheet's saved
+// revisions so a later +history-revert can target one by revision_id.
+var HistoryList = common.Shortcut{
+	Service:     "sheets",
+	Command:     "+history-list",
+	Description: "List the spreadsheet's saved history versions (paginated); use a returned revision_id with +history-revert.",
+	Risk:        "read",
+	Scopes:      []string{"sheets:spreadsheet:read"},
+	AuthTypes:   []string{"user", "bot"},
+	HasFormat:   true,
+	Flags:       flagsFor("+history-list"),
+	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		_, err := resolveSpreadsheetToken(runtime)
+		return err
+	},
+	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
+		token, _ := resolveSpreadsheetToken(runtime)
+		return invokeToolDryRun(token, ToolKindRead, "list_history_versions", historyListInput(runtime))
+	},
+	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		token, err := resolveSpreadsheetToken(runtime)
+		if err != nil {
+			return err
+		}
+		out, err := callTool(ctx, runtime, token, ToolKindRead, "list_history_versions", historyListInput(runtime))
+		if err != nil {
+			return err
+		}
+		runtime.Out(out, nil)
+		return nil
+	},
+	Tips: []string{
+		"Omit --cursor for the first page; pass the previous response's next_cursor to fetch the next page.",
+		"Pick a revision_id from a listing entry and pass it (plus that entry's edit_time to --edit-time) to +history-revert to roll the whole spreadsheet back.",
+	},
+}
+
+// HistoryRevert wraps revert_to_revision: roll the whole spreadsheet back to a
+// past revision. Returns an async task id to poll via +history-revert-status.
+var HistoryRevert = common.Shortcut{
+	Service:     "sheets",
+	Command:     "+history-revert",
+	Description: "Roll the whole spreadsheet back to a past revision (full-document restore; discards all later edits).",
+	Risk:        "write",
+	Scopes:      []string{"sheets:spreadsheet:write_only"},
+	AuthTypes:   []string{"user", "bot"},
+	HasFormat:   true,
+	Flags:       flagsFor("+history-revert"),
+	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		if _, err := resolveSpreadsheetToken(runtime); err != nil {
+			return err
+		}
+		_, err := historyRevertInput(runtime)
+		return err
+	},
+	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
+		token, _ := resolveSpreadsheetToken(runtime)
+		input, _ := historyRevertInput(runtime)
+		return invokeToolDryRun(token, ToolKindWrite, "revert_to_revision", input)
+	},
+	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		token, err := resolveSpreadsheetToken(runtime)
+		if err != nil {
+			return err
+		}
+		input, err := historyRevertInput(runtime)
+		if err != nil {
+			return err
+		}
+		out, err := callTool(ctx, runtime, token, ToolKindWrite, "revert_to_revision", input)
+		if err != nil {
+			return err
+		}
+		runtime.Out(out, nil)
+		return nil
+	},
+	Tips: []string{
+		"+history-revert is a FULL-DOCUMENT rollback — it discards every edit made after the target version, including other collaborators'.",
+		"--revision-id takes a revision_id (minor id) from +history-list; pass the same entry's edit_time to --edit-time to locate the version faster. Poll the returned task id with +history-revert-status.",
+	},
+}
+
+// HistoryRevertStatus wraps get_revert_status: poll an async revert task
+// started by +history-revert for completion.
+var HistoryRevertStatus = common.Shortcut{
+	Service:     "sheets",
+	Command:     "+history-revert-status",
+	Description: "Poll the status of an async revert task started by +history-revert.",
+	Risk:        "read",
+	Scopes:      []string{"sheets:spreadsheet:read"},
+	AuthTypes:   []string{"user", "bot"},
+	HasFormat:   true,
+	Flags:       flagsFor("+history-revert-status"),
+	Validate: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		if _, err := resolveSpreadsheetToken(runtime); err != nil {
+			return err
+		}
+		_, err := historyRevertStatusInput(runtime)
+		return err
+	},
+	DryRun: func(ctx context.Context, runtime *common.RuntimeContext) *common.DryRunAPI {
+		token, _ := resolveSpreadsheetToken(runtime)
+		input, _ := historyRevertStatusInput(runtime)
+		return invokeToolDryRun(token, ToolKindRead, "get_revert_status", input)
+	},
+	Execute: func(ctx context.Context, runtime *common.RuntimeContext) error {
+		token, err := resolveSpreadsheetToken(runtime)
+		if err != nil {
+			return err
+		}
+		input, err := historyRevertStatusInput(runtime)
+		if err != nil {
+			return err
+		}
+		out, err := callTool(ctx, runtime, token, ToolKindRead, "get_revert_status", input)
+		if err != nil {
+			return err
+		}
+		runtime.Out(out, nil)
+		return nil
+	},
+	Tips: []string{
+		"--task-id is the task id returned by +history-revert; re-run this until the task reports completion.",
+	},
+}
+
+// historyListInput builds the list_history_versions tool body. Network-free;
+// shared by DryRun and Execute. Both flags are optional: cursor is forwarded
+// only when set, count only when a positive value is given.
+func historyListInput(runtime flagView) map[string]interface{} {
+	input := map[string]interface{}{}
+	if cursor := strings.TrimSpace(runtime.Str("cursor")); cursor != "" {
+		input["cursor"] = cursor
+	}
+	if count := runtime.Int("count"); count > 0 {
+		input["count"] = count
+	}
+	return input
+}
+
+// historyRevertInput builds the revert_to_revision tool body. Network-free;
+// shared by Validate, DryRun, and Execute. revision_id 是 +history-list 返回的
+// revision_id（minor id）；edit_time 可选，传同一条 entry 的 edit_time 让服务端更快定位该版本。
+func historyRevertInput(runtime flagView) (map[string]interface{}, error) {
+	rev := strings.TrimSpace(runtime.Str("revision-id"))
+	if rev == "" {
+		return nil, common.FlagErrorf("--revision-id is required (a revision_id from +history-list)")
+	}
+	input := map[string]interface{}{
+		"revision_id": rev,
+	}
+	if et := strings.TrimSpace(runtime.Str("edit-time")); et != "" {
+		input["edit_time"] = et
+	}
+	return input, nil
+}
+
+// historyRevertStatusInput builds the get_revert_status tool body.
+// Network-free; shared by Validate, DryRun, and Execute.
+func historyRevertStatusInput(runtime flagView) (map[string]interface{}, error) {
+	tid := strings.TrimSpace(runtime.Str("task-id"))
+	if tid == "" {
+		return nil, common.FlagErrorf("--task-id is required")
+	}
+	return map[string]interface{}{
+		"task_id": tid,
+	}, nil
+}

shortcuts/sheets/lark_sheet_object_crud.go

@@ -49,6 +49,12 @@ type objectCRUDSpec struct {
 	// right nesting level.
 	enhanceCreateInput func(rt flagView, input map[string]interface{})
 	enhanceUpdateInput func(rt flagView, input map[string]interface{})
+	// validateCreateInput, when set, runs after enhanceCreateInput to
+	// enforce *cross-flag, create-only* constraints JSON Schema can't
+	// express (e.g. pivot rejects --target-position vs --range when
+	// both carry non-default values — they map to the same wire field
+	// and conflicting values are ambiguous). Mirrors validateUpdateInput.
+	validateCreateInput func(rt flagView, input map[string]interface{}) error
 	// validateUpdateInput, when set, runs after enhanceUpdateInput to
 	// enforce *cross-field, update-only* constraints JSON Schema can't
 	// express (e.g. sparkline requires properties.sparklines[i] to
@@ -190,6 +196,11 @@ func objectCreateInput(runtime flagView, token, sheetID, sheetName string, spec
 	if spec.enhanceCreateInput != nil {
 		spec.enhanceCreateInput(runtime, input)
 	}
+	if spec.validateCreateInput != nil {
+		if err := spec.validateCreateInput(runtime, input); err != nil {
+			return nil, err
+		}
+	}
 	if err := validateInputAgainstSchema(runtime, input); err != nil {
 		return nil, err
 	}
@@ -381,9 +392,6 @@ var pivotSpec = objectCRUDSpec{
 	},
 	createWarn: pivotPlacementWarn,
 	enhanceCreateInput: func(rt flagView, input map[string]interface{}) {
-		if v := strings.TrimSpace(rt.Str("target-position")); v != "" && v != "A1" {
-			input["target_position"] = v
-		}
 		props, _ := input["properties"].(map[string]interface{})
 		if props == nil {
 			return
@@ -391,10 +399,26 @@ var pivotSpec = objectCRUDSpec{
 		if v := strings.TrimSpace(rt.Str("source")); v != "" {
 			props["source"] = v
 		}
-		if v := strings.TrimSpace(rt.Str("range")); v != "" {
+		// --target-position 与 --range 都映射到 properties.range；
+		// --target-position 优先，未给（或为默认值 A1）时回落到 --range。
+		// 互斥校验在 validateCreateInput 里做。
+		if v := strings.TrimSpace(rt.Str("target-position")); v != "" && v != "A1" {
+			props["range"] = v
+		} else if v := strings.TrimSpace(rt.Str("range")); v != "" {
 			props["range"] = v
 		}
 	},
+	// --target-position 与 --range 落到同一 wire 字段（properties.range），
+	// 同时给非默认值时无法判断意图——按 --target-sheet-id / --target-sheet-name
+	// 的处理方式，CLI 端直接拒绝（优于静默丢弃其一）。
+	validateCreateInput: func(rt flagView, _ map[string]interface{}) error {
+		pos := strings.TrimSpace(rt.Str("target-position"))
+		rng := strings.TrimSpace(rt.Str("range"))
+		if pos != "" && pos != "A1" && rng != "" {
+			return common.FlagErrorf("--target-position and --range are mutually exclusive (both map to properties.range; pass only one)")
+		}
+		return nil
+	},
 }
 var PivotCreate = newObjectCreateShortcut(pivotSpec)
 var PivotUpdate = newObjectUpdateShortcut(pivotSpec)

shortcuts/sheets/lark_sheet_object_crud_test.go

@@ -137,25 +137,24 @@ func TestObjectCRUDShortcuts_DryRun(t *testing.T) {
 		// covered separately in the +pivot-create empty-selector / mutex
 		// tests below.
 		{
-			name: "+pivot-create with placement / source / range flags",
+			name: "+pivot-create with placement / source / target-position flags",
 			sc:   PivotCreate,
 			args: []string{
 				"--url", testURL, "--target-sheet-id", testSheetID,
 				"--properties", `{"rows":[{"field":"A"}]}`,
 				"--source", "Sheet1!A1:F1000",
-				"--range", "F1",
 				"--target-position", "B5",
 			},
 			toolName: "manage_pivot_table_object",
 			wantInput: map[string]interface{}{
-				"excel_id":        testToken,
-				"sheet_id":        testSheetID,
-				"operation":       "create",
-				"target_position": "B5",
+				"excel_id":  testToken,
+				"sheet_id":  testSheetID,
+				"operation": "create",
 				"properties": map[string]interface{}{
 					"rows":   []interface{}{map[string]interface{}{"field": "A"}},
 					"source": "Sheet1!A1:F1000",
-					"range":  "F1",
+					// --target-position 映射到 properties.range。
+					"range": "B5",
 				},
 			},
 		},
@@ -507,6 +506,55 @@ func TestPivotCreate_SheetSelectorSemantics(t *testing.T) {
 	})
 }
 
+// TestPivotCreate_TargetPositionRangeMutex regresses the "--target-position
+// and --range cannot both be set" guardrail on +pivot-create. They map to
+// the same wire field (properties.range), so two non-default values are
+// ambiguous; the CLI rejects up front (mirrors the --target-sheet-id /
+// --target-sheet-name mutex). --target-position=A1 is the documented default
+// and is treated as "not set" — pairing it with --range still works.
+func TestPivotCreate_TargetPositionRangeMutex(t *testing.T) {
+	t.Parallel()
+
+	t.Run("both non-default values rejected", func(t *testing.T) {
+		t.Parallel()
+		_, stderr, err := runShortcutCapturingErr(t, PivotCreate, []string{
+			"--url", testURL,
+			"--target-sheet-id", testSheetID,
+			"--properties", `{"rows":[{"field":"A"}]}`,
+			"--source", "Sheet1!A1:F1000",
+			"--target-position", "B5",
+			"--range", "F1",
+		})
+		if err == nil {
+			t.Fatalf("expected CLI to reject --target-position with --range; stderr=%s", stderr)
+		}
+		combined := stderr + err.Error()
+		if !strings.Contains(combined, "mutually exclusive") {
+			t.Errorf("expected error to say 'mutually exclusive'; got=%s|%v", stderr, err)
+		}
+		if !strings.Contains(combined, "--target-position") || !strings.Contains(combined, "--range") {
+			t.Errorf("expected error to quote both --target-position and --range; got=%s|%v", stderr, err)
+		}
+	})
+
+	t.Run("default A1 with --range is accepted (range wins)", func(t *testing.T) {
+		t.Parallel()
+		body := parseDryRunBody(t, PivotCreate, []string{
+			"--url", testURL,
+			"--target-sheet-id", testSheetID,
+			"--properties", `{"rows":[{"field":"A"}]}`,
+			"--source", "Sheet1!A1:F1000",
+			"--target-position", "A1",
+			"--range", "F1",
+		})
+		input := decodeToolInput(t, body, "manage_pivot_table_object")
+		props, _ := input["properties"].(map[string]interface{})
+		if got, _ := props["range"].(string); got != "F1" {
+			t.Errorf("properties.range = %q, want %q", got, "F1")
+		}
+	})
+}
+
 // TestPivotCreate_SchemaValidates exercises the schema-driven
 // validator wired into objectCreateInput. The pivot create schema
 // doesn't constrain rows/columns/values to be present (the backend

shortcuts/sheets/lark_sheet_read_data.go

@@ -5,8 +5,6 @@ package sheets
 
 import (
 	"context"
-	"encoding/csv"
-	"regexp"
 	"strconv"
 	"strings"
 
@@ -164,12 +162,7 @@ var CsvGet = common.Shortcut{
 		if err != nil {
 			return err
 		}
-		switch {
-		case runtime.Bool("rows-json"):
-			// --rows-json reshapes the CSV response into structured rows
-			// ({row_number, values:{col→cell}}); see assembleRowsJSON.
-			out = assembleRowsJSON(out, strings.TrimSpace(runtime.Str("range")))
-		case !runtime.Bool("include-row-prefix"):
+		if !runtime.Bool("include-row-prefix") {
 			out = stripRowPrefixFromCsvOutput(out)
 		}
 		runtime.Out(out, nil)
@@ -219,141 +212,6 @@ func stripRowPrefixFromCsvOutput(out interface{}) interface{} {
 	return m
 }
 
-// rowPrefixRe matches the leading "[row=N] " (or "[row=N],") annotation that
-// the tool prepends to the first physical line of each logical CSV record.
-var rowPrefixRe = regexp.MustCompile(`^\[row=(\d+)\][ ,]?`)
-
-// assembleRowsJSON reshapes the tool's annotated_csv string into structured
-// rows so callers never have to regex-parse "[row=N]" or RFC-4180 CSV by hand:
-//
-//	{
-//	  "range": "A1:K3380",
-//	  "current_region": "...",        // passthrough, if the tool returned it
-//	  "rows": [{"row_number":1,"values":{"A":"姓名", ..., "K":"时间差_分钟"}},
-//	           {"row_number":2,"values":{"A":"张三", ..., "K":"8.5"}}, ...]
-//	}
-//
-// Every logical row is emitted, including the first — no row is assumed to be a
-// header, since sheet data is not always tabular. Each cell is keyed by its
-// column letter (from the tool's col_indices when present, else derived from the
-// requested range's start column). On any parsing trouble it returns the
-// original output unchanged.
-func assembleRowsJSON(out interface{}, requestedRange string) interface{} {
-	m, ok := out.(map[string]interface{})
-	if !ok {
-		return out
-	}
-	csvStr, ok := m["annotated_csv"].(string)
-	if !ok {
-		return out
-	}
-
-	// Group physical lines into logical records by [row=N] boundaries; lines
-	// without a prefix are embedded-newline continuations of the current record.
-	type logicalRow struct {
-		num  int
-		text string
-	}
-	var groups []logicalRow
-	for _, line := range strings.Split(csvStr, "\n") {
-		if mm := rowPrefixRe.FindStringSubmatch(line); mm != nil {
-			n, _ := strconv.Atoi(mm[1])
-			groups = append(groups, logicalRow{num: n, text: line[len(mm[0]):]})
-		} else if len(groups) > 0 {
-			groups[len(groups)-1].text += "\n" + line
-		}
-	}
-	if len(groups) == 0 {
-		return out
-	}
-
-	// Parse every logical row; widest row sets the column count. No row is
-	// singled out as a header — that would assume the data is tabular, which it
-	// often is not. The model reads row 1 like any other row and decides for
-	// itself whether it is a header.
-	parsed := make([][]string, len(groups))
-	maxCols := 0
-	for i, g := range groups {
-		parsed[i] = parseCSVRecord(g.text)
-		if len(parsed[i]) > maxCols {
-			maxCols = len(parsed[i])
-		}
-	}
-	if maxCols == 0 {
-		return out
-	}
-
-	// Column letters key each cell. Prefer the tool's col_indices (authoritative,
-	// length == col_count); otherwise derive from the requested range's start col.
-	letters := coerceStringSlice(m["col_indices"])
-	if len(letters) < maxCols {
-		start := csvStartColIndex(requestedRange)
-		letters = make([]string, maxCols)
-		for j := 0; j < maxCols; j++ {
-			letters[j] = csvColLetter(start + j)
-		}
-	}
-
-	rows := make([]map[string]interface{}, 0, len(groups))
-	for i := range groups {
-		fields := parsed[i]
-		values := make(map[string]interface{}, len(letters))
-		for j := range letters {
-			v := ""
-			if j < len(fields) {
-				v = fields[j]
-			}
-			values[letters[j]] = v
-		}
-		rows = append(rows, map[string]interface{}{
-			"row_number": groups[i].num,
-			"values":     values,
-		})
-	}
-
-	result := map[string]interface{}{}
-	for k, v := range m {
-		result[k] = v
-	}
-	result["range"] = requestedRange
-	result["rows"] = rows
-
-	// Surface the backend's "数据没读全" signal structurally instead of leaving it
-	// buried in warning_message prose. The tool flags it when current_region (the
-	// true data extent) reaches past actual_range (what was actually read) — the
-	// single most important anti-under-read hint. Mirror that same comparison
-	// (regionEndRow > actualEndRow) from the already-passthrough A1 ranges so the
-	// model gets the real data range as a first-class field, never having to
-	// parse it out of prose.
-	if cr, _ := m["current_region"].(string); cr != "" {
-		ar, _ := m["actual_range"].(string)
-		regionEnd := a1EndRow(cr)
-		readEnd := a1EndRow(ar)
-		if regionEnd > 0 && readEnd > 0 && regionEnd > readEnd {
-			result["data_not_fully_read"] = map[string]interface{}{
-				"read_through_row":         readEnd,
-				"data_extends_through_row": regionEnd,
-				"unread_rows":              regionEnd - readEnd,
-				"reread_range":             cr,
-			}
-		}
-	}
-
-	// Drop the fields whose information rows-json fully carries elsewhere:
-	//   - annotated_csv / row_indices / col_indices → reconstructed into
-	//     columns + rows (with integer row_number), losslessly.
-	//   - warning_message → its two halves are both handled: the static
-	//     "[row=N] / col_indices[j]" parse nag is moot once those fields exist,
-	//     and the dynamic "数据没读全" half is now the structured
-	//     data_not_fully_read field above. (Confirmed against the backend's
-	//     get-range-as-csv.ts — warning_message has no other content.)
-	delete(result, "annotated_csv")
-	delete(result, "row_indices")
-	delete(result, "col_indices")
-	delete(result, "warning_message")
-	return result
-}
-
 // a1EndRow extracts the ending row number from an A1 range, e.g. "A1:N51" → 51,
 // "Sheet1!B2:D9" → 9, "C5" → 5. Returns 0 when no row number is present.
 func a1EndRow(rng string) int {
@@ -377,89 +235,6 @@ func a1EndRow(rng string) int {
 	return n
 }
 
-// parseCSVRecord parses a single logical CSV record (which may span multiple
-// physical lines via quoted embedded newlines) into its fields. An empty record
-// yields no fields; a malformed record falls back to a naive comma split so a
-// stray quote never drops a whole row.
-func parseCSVRecord(text string) []string {
-	if strings.TrimSpace(text) == "" {
-		return nil
-	}
-	r := csv.NewReader(strings.NewReader(text))
-	r.FieldsPerRecord = -1
-	fields, err := r.Read()
-	if err != nil {
-		return strings.Split(text, ",")
-	}
-	return fields
-}
-
-// coerceStringSlice returns v as []string when it is a homogeneous []interface{}
-// of strings (the shape of the tool's col_indices), else nil.
-func coerceStringSlice(v interface{}) []string {
-	arr, ok := v.([]interface{})
-	if !ok {
-		return nil
-	}
-	out := make([]string, 0, len(arr))
-	for _, e := range arr {
-		s, ok := e.(string)
-		if !ok {
-			return nil
-		}
-		out = append(out, s)
-	}
-	return out
-}
-
-// csvStartColIndex returns the 0-based column index of a range's start column,
-// e.g. "A1:K3380" → 0, "C5:F9" → 2, "Sheet1!D2" → 3. Unparseable input → 0.
-func csvStartColIndex(rng string) int {
-	rng = strings.TrimSpace(rng)
-	if i := strings.LastIndex(rng, "!"); i >= 0 {
-		rng = rng[i+1:]
-	}
-	var letters strings.Builder
-	for _, c := range rng {
-		if (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') {
-			letters.WriteRune(c)
-			continue
-		}
-		break
-	}
-	if letters.Len() == 0 {
-		return 0
-	}
-	return csvColToIndex(letters.String())
-}
-
-// csvColToIndex converts a column letter to its 0-based index ("A"→0, "K"→10,
-// "AA"→26). Non-letter input → -1.
-func csvColToIndex(s string) int {
-	n := 0
-	for _, c := range strings.ToUpper(s) {
-		if c < 'A' || c > 'Z' {
-			break
-		}
-		n = n*26 + int(c-'A'+1)
-	}
-	return n - 1
-}
-
-// csvColLetter converts a 0-based column index back to its letter (0→"A",
-// 25→"Z", 26→"AA"). Negative input → "".
-func csvColLetter(idx int) string {
-	if idx < 0 {
-		return ""
-	}
-	var b []byte
-	for idx >= 0 {
-		b = append([]byte{byte('A' + idx%26)}, b...)
-		idx = idx/26 - 1
-	}
-	return string(b)
-}
-
 // DropdownGet wraps get_cell_ranges scoped to data_validation: read the
 // dropdown configuration on a range. Aligned with its sibling +cells-get
 // — sheet selection is via --sheet-id / --sheet-name (XOR), and --range

shortcuts/sheets/lark_sheet_read_data_test.go

@@ -63,20 +63,6 @@ func TestReadDataShortcuts_DryRun(t *testing.T) {
 				"value_render_option": "formatted_value",
 			},
 		},
-		{
-			// --rows-json is post-processing on +csv-get's response; it must
-			// NOT leak into the get_range_as_csv input.
-			name:     "+csv-get --rows-json builds the same input (flag is post-process)",
-			sc:       CsvGet,
-			args:     []string{"--url", testURL, "--sheet-id", testSheetID, "--range", "A1:C10", "--rows-json"},
-			toolName: "get_range_as_csv",
-			wantInput: map[string]interface{}{
-				"excel_id": testToken,
-				"sheet_id": testSheetID,
-				"range":    "A1:C10",
-				"max_rows": float64(unboundedReadLimit),
-			},
-		},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
@@ -179,113 +165,3 @@ func TestCsvGet_StripRowPrefix(t *testing.T) {
 		t.Errorf("other field corrupted: %v", out["other"])
 	}
 }
-
-// TestAssembleRowsJSON covers the --rows-json reshaping: every logical row
-// emitted (no header singled out), integer row_number, column-letter keyed
-// values, embedded newlines inside quoted fields, and current_region passthrough.
-func TestAssembleRowsJSON(t *testing.T) {
-	t.Parallel()
-	in := map[string]interface{}{
-		"annotated_csv":   "[row=1] 姓名,备注,时间差_分钟\n[row=2] 张三,\"line1\nline2\",8.5\n[row=3] 李四,ok,3",
-		"current_region":  "A1:C3",
-		"col_indices":     []interface{}{"A", "B", "C"},
-		"row_indices":     []interface{}{1, 2, 3},
-		"warning_message": "①定位行号…②定位列字母…",
-	}
-	out, ok := assembleRowsJSON(in, "A1:C3").(map[string]interface{})
-	if !ok {
-		t.Fatalf("assembleRowsJSON did not return a map")
-	}
-
-	// Fields whose info rows-json carries elsewhere are dropped (annotated_csv /
-	// indices → rows; warning_message → moot static nag + structured
-	// data_not_fully_read). Unrelated metadata like current_region is preserved.
-	if _, exists := out["annotated_csv"]; exists {
-		t.Errorf("annotated_csv should be dropped")
-	}
-	if _, exists := out["col_indices"]; exists {
-		t.Errorf("col_indices should be dropped")
-	}
-	if _, exists := out["warning_message"]; exists {
-		t.Errorf("warning_message should be dropped in rows-json mode")
-	}
-	if _, exists := out["columns"]; exists {
-		t.Errorf("columns field should not exist (no header assumption)")
-	}
-	if out["current_region"] != "A1:C3" {
-		t.Errorf("current_region passthrough lost: %v", out["current_region"])
-	}
-
-	rows, _ := out["rows"].([]map[string]interface{})
-	if len(rows) != 3 {
-		t.Fatalf("want all 3 rows (incl. row 1), got %d: %+v", len(rows), rows)
-	}
-	// Row 1 is emitted as a normal row, not consumed as a header.
-	if rows[0]["row_number"].(int) != 1 {
-		t.Errorf("first row_number = %v, want 1", rows[0]["row_number"])
-	}
-	if v := rows[0]["values"].(map[string]interface{}); v["A"] != "姓名" || v["C"] != "时间差_分钟" {
-		t.Errorf("row 1 values wrong: %+v", v)
-	}
-	// Row 2 keeps its embedded newline inside a single cell.
-	v1 := rows[1]["values"].(map[string]interface{})
-	if rows[1]["row_number"].(int) != 2 || v1["A"] != "张三" || v1["B"] != "line1\nline2" || v1["C"] != "8.5" {
-		t.Errorf("row 2 wrong (embedded newline?): %+v", rows[1])
-	}
-}
-
-// TestAssembleRowsJSON_DerivedLetters verifies cell letters are derived from the
-// range start when the tool omits col_indices (e.g. a C-anchored read).
-func TestAssembleRowsJSON_DerivedLetters(t *testing.T) {
-	t.Parallel()
-	in := map[string]interface{}{
-		"annotated_csv": "[row=5] h1,h2\n[row=6] a,b",
-	}
-	out := assembleRowsJSON(in, "C5:D6").(map[string]interface{})
-	rows := out["rows"].([]map[string]interface{})
-	if len(rows) != 2 {
-		t.Fatalf("want 2 rows, got %d", len(rows))
-	}
-	if rows[0]["row_number"].(int) != 5 {
-		t.Errorf("first row_number = %v, want 5", rows[0]["row_number"])
-	}
-	if v := rows[0]["values"].(map[string]interface{}); v["C"] != "h1" || v["D"] != "h2" {
-		t.Errorf("derived-letter values wrong: %+v", v)
-	}
-	if v := rows[1]["values"].(map[string]interface{}); v["C"] != "a" || v["D"] != "b" {
-		t.Errorf("row 6 values wrong: %+v", v)
-	}
-}
-
-// TestAssembleRowsJSON_DataNotFullyRead verifies the structured under-read hint:
-// when current_region extends past actual_range, rows-json surfaces the true data
-// range as a first-class field (mirroring the backend's prose warning).
-func TestAssembleRowsJSON_DataNotFullyRead(t *testing.T) {
-	t.Parallel()
-	// Read only A1:D2, but the data region reaches D4 → 2 rows unread.
-	in := map[string]interface{}{
-		"annotated_csv":  "[row=1] 序号,姓名\n[row=2] 101,张三",
-		"actual_range":   "A1:D2",
-		"current_region": "A1:D4",
-	}
-	out := assembleRowsJSON(in, "A1:D2").(map[string]interface{})
-	hint, ok := out["data_not_fully_read"].(map[string]interface{})
-	if !ok {
-		t.Fatalf("data_not_fully_read missing; out=%+v", out)
-	}
-	if hint["read_through_row"] != 2 || hint["data_extends_through_row"] != 4 ||
-		hint["unread_rows"] != 2 || hint["reread_range"] != "A1:D4" {
-		t.Errorf("data_not_fully_read wrong: %+v", hint)
-	}
-
-	// Fully-read case: no hint emitted.
-	in2 := map[string]interface{}{
-		"annotated_csv":  "[row=1] 序号,姓名\n[row=2] 101,张三",
-		"actual_range":   "A1:D2",
-		"current_region": "A1:D2",
-	}
-	out2 := assembleRowsJSON(in2, "A1:D2").(map[string]interface{})
-	if _, exists := out2["data_not_fully_read"]; exists {
-		t.Errorf("data_not_fully_read should be absent when fully read")
-	}
-}

shortcuts/sheets/lark_sheet_workbook_export_test.go

@@ -0,0 +1,72 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"encoding/json"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/internal/httpmock"
+)
+
+// TestWorkbookExport_ExecuteExportOnly covers the no-download path: without
+// --output-path, +workbook-export delegates to the shared drive export core
+// with OutputDir="" so it creates + polls the export task and returns the ready
+// file token without writing a local file (downloaded=false).
+func TestWorkbookExport_ExecuteExportOnly(t *testing.T) {
+	stubs := []*httpmock.Stub{
+		{
+			Method: "POST",
+			URL:    "/open-apis/drive/v1/export_tasks",
+			Body: map[string]interface{}{
+				"code": 0, "msg": "ok",
+				"data": map[string]interface{}{"ticket": "tk_export"},
+			},
+		},
+		{
+			Method: "GET",
+			URL:    "/open-apis/drive/v1/export_tasks/tk_export",
+			Body: map[string]interface{}{
+				"code": 0, "msg": "ok",
+				"data": map[string]interface{}{"result": map[string]interface{}{
+					"job_status": float64(0),
+					"file_token": "ftk_xlsx",
+					"file_name":  "report.xlsx",
+					"file_size":  float64(2048),
+				}},
+			},
+		},
+	}
+
+	out, err := runShortcutWithStubs(t, WorkbookExport, []string{
+		"--url", testURL, "--file-extension", "xlsx", "--as", "user",
+	}, stubs...)
+	if err != nil {
+		t.Fatalf("export-only execute failed: %v\n%s", err, out)
+	}
+
+	idx := strings.Index(out, "{")
+	if idx < 0 {
+		t.Fatalf("no JSON envelope:\n%s", out)
+	}
+	var env struct {
+		Data map[string]interface{} `json:"data"`
+	}
+	if err := json.Unmarshal([]byte(out[idx:]), &env); err != nil {
+		t.Fatalf("decode envelope: %v\nraw=%s", err, out)
+	}
+	if env.Data["ready"] != true {
+		t.Errorf("ready = %v, want true", env.Data["ready"])
+	}
+	if env.Data["downloaded"] != false {
+		t.Errorf("downloaded = %v, want false (no --output-path)", env.Data["downloaded"])
+	}
+	if env.Data["file_token"] != "ftk_xlsx" {
+		t.Errorf("file_token = %v, want ftk_xlsx", env.Data["file_token"])
+	}
+	if env.Data["doc_type"] != "sheet" {
+		t.Errorf("doc_type = %v, want sheet", env.Data["doc_type"])
+	}
+}

shortcuts/sheets/lark_sheet_workbook_import_test.go

@@ -0,0 +1,135 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package sheets
+
+import (
+	"encoding/json"
+	"os"
+	"strings"
+	"testing"
+
+	"github.com/larksuite/cli/internal/httpmock"
+	_ "github.com/larksuite/cli/internal/vfs/localfileio"
+)
+
+// chdirTemp switches into a fresh temp dir for the duration of the test and
+// restores the original cwd afterwards. +workbook-import is the first sheets
+// shortcut that stat()s a real local file, so these tests need a working dir.
+func chdirTemp(t *testing.T) {
+	t.Helper()
+	orig, err := os.Getwd()
+	if err != nil {
+		t.Fatalf("getwd: %v", err)
+	}
+	if err := os.Chdir(t.TempDir()); err != nil {
+		t.Fatalf("chdir: %v", err)
+	}
+	t.Cleanup(func() { _ = os.Chdir(orig) })
+}
+
+// TestWorkbookImport_DryRunPinsSheetType verifies the shortcut delegates to the
+// shared drive import core and hard-codes the import target type to "sheet".
+func TestWorkbookImport_DryRunPinsSheetType(t *testing.T) {
+	chdirTemp(t)
+	if err := os.WriteFile("data.xlsx", []byte("fake-xlsx"), 0o644); err != nil {
+		t.Fatalf("write file: %v", err)
+	}
+
+	calls := parseDryRunAPI(t, WorkbookImport, []string{"--file", "./data.xlsx"})
+
+	var createBody map[string]interface{}
+	for _, c := range calls {
+		cm, _ := c.(map[string]interface{})
+		if u, _ := cm["url"].(string); u == "/open-apis/drive/v1/import_tasks" {
+			createBody, _ = cm["body"].(map[string]interface{})
+		}
+	}
+	if createBody == nil {
+		t.Fatalf("no import_tasks create call in dry-run: %#v", calls)
+	}
+	if createBody["type"] != "sheet" {
+		t.Errorf("import type = %v, want sheet (must be pinned regardless of file)", createBody["type"])
+	}
+	if createBody["file_extension"] != "xlsx" {
+		t.Errorf("file_extension = %v, want xlsx", createBody["file_extension"])
+	}
+}
+
+// TestWorkbookImport_RejectsNonSheetFile ensures a file that cannot become a
+// spreadsheet (e.g. .docx) is rejected up front by the pinned-sheet validation.
+func TestWorkbookImport_RejectsNonSheetFile(t *testing.T) {
+	chdirTemp(t)
+	if err := os.WriteFile("notes.docx", []byte("fake-docx"), 0o644); err != nil {
+		t.Fatalf("write file: %v", err)
+	}
+
+	// Validate runs before DryRun, so the pinned-sheet check rejects .docx up
+	// front and the error surfaces through the normal envelope/err path.
+	stdout, stderr, err := runShortcutCapturingErr(t, WorkbookImport, []string{"--file", "./notes.docx", "--dry-run"})
+	if err == nil || !strings.Contains(stdout+stderr+err.Error(), "can only be imported") {
+		t.Errorf("expected .docx → sheet type-mismatch rejection; got stdout=%s stderr=%s err=%v", stdout, stderr, err)
+	}
+}
+
+// TestWorkbookImport_ExecuteCreatesSheet runs the full upload → create → poll
+// flow against stubs and asserts the resulting URL is a /sheets/ link.
+func TestWorkbookImport_ExecuteCreatesSheet(t *testing.T) {
+	chdirTemp(t)
+	if err := os.WriteFile("data.csv", []byte("a,b\n1,2\n"), 0o644); err != nil {
+		t.Fatalf("write file: %v", err)
+	}
+
+	stubs := []*httpmock.Stub{
+		{
+			Method: "POST",
+			URL:    "/open-apis/drive/v1/medias/upload_all",
+			Body: map[string]interface{}{
+				"code": 0, "msg": "ok",
+				"data": map[string]interface{}{"file_token": "file_import_media"},
+			},
+		},
+		{
+			Method: "POST",
+			URL:    "/open-apis/drive/v1/import_tasks",
+			Body: map[string]interface{}{
+				"code": 0, "msg": "ok",
+				"data": map[string]interface{}{"ticket": "tk_sheet"},
+			},
+		},
+		{
+			Method: "GET",
+			URL:    "/open-apis/drive/v1/import_tasks/tk_sheet",
+			Body: map[string]interface{}{
+				"code": 0, "msg": "ok",
+				"data": map[string]interface{}{"result": map[string]interface{}{
+					"token":      "shtcn_imported",
+					"type":       "sheet",
+					"job_status": float64(0),
+				}},
+			},
+		},
+	}
+
+	out, err := runShortcutWithStubs(t, WorkbookImport, []string{"--file", "./data.csv", "--as", "user"}, stubs...)
+	if err != nil {
+		t.Fatalf("import execute failed: %v\n%s", err, out)
+	}
+
+	idx := strings.Index(out, "{")
+	if idx < 0 {
+		t.Fatalf("execute output has no JSON envelope:\n%s", out)
+	}
+	var env struct {
+		Data map[string]interface{} `json:"data"`
+	}
+	if err := json.Unmarshal([]byte(out[idx:]), &env); err != nil {
+		t.Fatalf("decode envelope: %v\nraw=%s", err, out)
+	}
+	if url, _ := env.Data["url"].(string); !strings.Contains(url, "/sheets/") {
+		t.Errorf("imported url = %q, want a /sheets/ link", url)
+	}
+	if tok, _ := env.Data["token"].(string); tok != "shtcn_imported" {
+		t.Errorf("token = %q, want shtcn_imported", tok)
+	}
+}

shortcuts/sheets/lark_sheet_workbook_test.go

@@ -4,14 +4,10 @@
 package sheets
 
 import (
-	"errors"
-	"net/http"
+	"encoding/json"
 	"strings"
 	"testing"
 
-	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
-
-	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/shortcuts/common"
 )
 
@@ -145,6 +141,28 @@ func TestWorkbookShortcuts_DryRun(t *testing.T) {
 				"tab_color": "",
 			},
 		},
+		{
+			name:     "+sheet-show-gridline",
+			sc:       SheetShowGridline,
+			args:     []string{"--url", testURL, "--sheet-id", testSheetID},
+			toolName: "modify_workbook_structure",
+			wantInput: map[string]interface{}{
+				"excel_id":  testToken,
+				"operation": "show_gridline",
+				"sheet_id":  testSheetID,
+			},
+		},
+		{
+			name:     "+sheet-hide-gridline",
+			sc:       SheetHideGridline,
+			args:     []string{"--url", testURL, "--sheet-id", testSheetID},
+			toolName: "modify_workbook_structure",
+			wantInput: map[string]interface{}{
+				"excel_id":  testToken,
+				"operation": "hide_gridline",
+				"sheet_id":  testSheetID,
+			},
+		},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
@@ -288,7 +306,7 @@ func TestWorkbookCreate_DryRun(t *testing.T) {
 		t.Parallel()
 		calls := parseDryRunAPI(t, WorkbookCreate, []string{"--title", "MySheet"})
 		if len(calls) != 1 {
-			t.Fatalf("api calls = %d, want 1 (no headers/data)", len(calls))
+			t.Fatalf("api calls = %d, want 1 (no values)", len(calls))
 		}
 		c := calls[0].(map[string]interface{})
 		if c["url"] != "/open-apis/sheets/v3/spreadsheets" {
@@ -300,12 +318,11 @@ func TestWorkbookCreate_DryRun(t *testing.T) {
 		}
 	})
 
-	t.Run("with headers and data → 2-step plan", func(t *testing.T) {
+	t.Run("with values → 2-step plan", func(t *testing.T) {
 		t.Parallel()
 		calls := parseDryRunAPI(t, WorkbookCreate, []string{
 			"--title", "Sales",
-			"--headers", `["Name","Score"]`,
-			"--values", `[["alice",95],["bob",88]]`,
+			"--values", `[["Name","Score"],["alice",95],["bob",88]]`,
 		})
 		if len(calls) != 2 {
 			t.Fatalf("api calls = %d, want 2 (create + fill)", len(calls))
@@ -317,7 +334,88 @@ func TestWorkbookCreate_DryRun(t *testing.T) {
 		body, _ := fill["body"].(map[string]interface{})
 		input := decodeToolInput(t, body, "set_cell_range")
 		if input["range"] != "A1:B3" {
-			t.Errorf("fill range = %v, want A1:B3 (1 header + 2 data rows × 2 cols)", input["range"])
+			t.Errorf("fill range = %v, want A1:B3 (3 rows × 2 cols)", input["range"])
+		}
+	})
+
+	t.Run("with styles merges into set_cell_range cells", func(t *testing.T) {
+		t.Parallel()
+		calls := parseDryRunAPI(t, WorkbookCreate, []string{
+			"--title", "Sales",
+			"--values", `[["Name","Score"],["alice",95]]`,
+			"--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1","font_weight":"bold","background_color":"#f5f5f5"},{"range":"B1","number_format":"0","border_styles":{"bottom":{"style":"solid","weight":"thin","color":"#000000"}}},{"range":"B2","font_color":"#0f7b0f"}]}]}`,
+		})
+		if len(calls) != 2 {
+			t.Fatalf("api calls = %d, want 2 (create + fill)", len(calls))
+		}
+		body, _ := calls[1].(map[string]interface{})["body"].(map[string]interface{})
+		input := decodeToolInput(t, body, "set_cell_range")
+		cells, _ := input["cells"].([]interface{})
+		if len(cells) != 2 {
+			t.Fatalf("cells rows = %#v, want 2", input["cells"])
+		}
+		headerRow, _ := cells[0].([]interface{})
+		firstHeader, _ := headerRow[0].(map[string]interface{})
+		firstStyle, _ := firstHeader["cell_styles"].(map[string]interface{})
+		if firstStyle["font_weight"] != "bold" || firstStyle["background_color"] != "#f5f5f5" {
+			t.Errorf("first header style = %#v, want bold + background", firstStyle)
+		}
+		secondHeader, _ := headerRow[1].(map[string]interface{})
+		if secondHeader["border_styles"] == nil {
+			t.Errorf("second header missing border_styles: %#v", secondHeader)
+		}
+		secondStyle, _ := secondHeader["cell_styles"].(map[string]interface{})
+		if secondStyle["number_format"] != "0" {
+			t.Errorf("second header number_format = %#v, want 0", secondStyle)
+		}
+		dataRow, _ := cells[1].([]interface{})
+		firstData, _ := dataRow[0].(map[string]interface{})
+		if _, ok := firstData["cell_styles"]; ok {
+			t.Errorf("null style should leave first data cell unstyled: %#v", firstData)
+		}
+		secondData, _ := dataRow[1].(map[string]interface{})
+		secondDataStyle, _ := secondData["cell_styles"].(map[string]interface{})
+		if secondDataStyle["font_color"] != "#0f7b0f" {
+			t.Errorf("second data style = %#v, want font color", secondDataStyle)
+		}
+	})
+
+	t.Run("cell style range can cover the whole initial range", func(t *testing.T) {
+		t.Parallel()
+		calls := parseDryRunAPI(t, WorkbookCreate, []string{
+			"--title", "Sales",
+			"--values", `[["Name","Score"],["alice",95]]`,
+			"--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1:B2","horizontal_alignment":"center"}]}]}`,
+		})
+		body, _ := calls[1].(map[string]interface{})["body"].(map[string]interface{})
+		input := decodeToolInput(t, body, "set_cell_range")
+		raw, _ := json.Marshal(input["cells"])
+		if got := strings.Count(string(raw), "horizontal_alignment"); got != 4 {
+			t.Errorf("horizontal_alignment occurrences = %d, want 4 in 2x2 range; cells=%s", got, raw)
+		}
+	})
+	t.Run("overlapping cell_styles deep-merge fields, no cross-cell pollution", func(t *testing.T) {
+		t.Parallel()
+		calls := parseDryRunAPI(t, WorkbookCreate, []string{
+			"--title", "X",
+			"--values", `[["a","b"]]`,
+			"--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1:B1","font_weight":"bold"},{"range":"B1","font_color":"#ff0000"}]}]}`,
+		})
+		body, _ := calls[1].(map[string]interface{})["body"].(map[string]interface{})
+		input := decodeToolInput(t, body, "set_cell_range")
+		cells, _ := input["cells"].([]interface{})
+		row0, _ := cells[0].([]interface{})
+		// B1 hit by both ops → must keep BOTH font_weight (op1) and font_color (op2).
+		b1, _ := row0[1].(map[string]interface{})
+		b1s, _ := b1["cell_styles"].(map[string]interface{})
+		if b1s["font_weight"] != "bold" || b1s["font_color"] != "#ff0000" {
+			t.Errorf("B1 should deep-merge both ops, got %#v", b1s)
+		}
+		// A1 hit only by op1 → must NOT be polluted by op2's font_color (shared submap).
+		a1, _ := row0[0].(map[string]interface{})
+		a1s, _ := a1["cell_styles"].(map[string]interface{})
+		if a1s["font_color"] != nil {
+			t.Errorf("A1 must not be polluted by op2, got %#v", a1s)
 		}
 	})
 }
@@ -330,8 +428,18 @@ func TestWorkbookCreate_DataValidation(t *testing.T) {
 		args []string
 		want string
 	}{
-		{"headers not array", []string{"--title", "X", "--headers", `"abc"`}, "must be a JSON array"},
 		{"values not 2D", []string{"--title", "X", "--values", `["a","b"]`}, "must be an array"},
+		{"styles not object", []string{"--title", "X", "--styles", `"bold"`}, `shaped as {"styles":[...]}`},
+		{"styles missing array", []string{"--title", "X", "--styles", `{"value":"x"}`}, "--styles.styles is required"},
+		{"styles item missing groups", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","value":"x"}]}`}, "must include at least one of cell_styles/row_sizes/col_sizes/cell_merges"},
+		{"cell styles must be array", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","cell_styles":{"range":"A1","font_weight":"bold"}}]}`}, "cell_styles must be an array"},
+		{"cell style needs range", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"font_weight":"bold"}]}]}`}, "range is required"},
+		{"nested cell_styles rejected", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1","cell_styles":{"font_weight":"bold"}}]}]}`}, "put style fields directly"},
+		{"row size needs row range", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","row_sizes":[{"range":"A1","type":"pixel","size":20}]}]}`}, "must use row numbers"},
+		{"col size needs pixel size", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","col_sizes":[{"range":"A:A","type":"pixel"}]}]}`}, "requires size"},
+		{"border bad style enum", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1","border_styles":{"bottom":{"style":"NONSENSE"}}}]}]}`}, `style "NONSENSE" is invalid`},
+		{"border invalid side", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1","border_styles":{"diagonal":{"style":"solid"}}}]}]}`}, "not a valid side"},
+		{"border bad weight", []string{"--title", "X", "--values", `[["a"]]`, "--styles", `{"styles":[{"name":"Sheet1","cell_styles":[{"range":"A1","border_styles":{"top":{"weight":"xxl"}}}]}]}`}, `weight "xxl" is invalid`},
 	}
 	for _, tt := range cases {
 		t.Run(tt.name, func(t *testing.T) {
@@ -344,21 +452,21 @@ func TestWorkbookCreate_DataValidation(t *testing.T) {
 	}
 }
 
-// TestWorkbookExport_DryRun checks the 2-or-3 step plan depending on
-// --output-path. The order should be: POST → GET (poll) → optional GET
-// (download).
+// TestWorkbookExport_DryRun verifies the export dry-run now delegates to the
+// shared drive export core: a single create-task POST (poll + download are
+// described inline rather than as separate api entries).
 func TestWorkbookExport_DryRun(t *testing.T) {
 	t.Parallel()
 
-	t.Run("xlsx without --output-path → 2 steps", func(t *testing.T) {
+	t.Run("xlsx create-task body pins type=sheet", func(t *testing.T) {
 		t.Parallel()
 		calls := parseDryRunAPI(t, WorkbookExport, []string{"--url", testURL, "--file-extension", "xlsx"})
-		if len(calls) != 2 {
-			t.Fatalf("api calls = %d, want 2 (create + poll)", len(calls))
+		if len(calls) != 1 {
+			t.Fatalf("api calls = %d, want 1 (create export task)", len(calls))
 		}
 		create := calls[0].(map[string]interface{})
 		if create["url"] != "/open-apis/drive/v1/export_tasks" {
-			t.Errorf("first url = %v", create["url"])
+			t.Errorf("url = %v", create["url"])
 		}
 		body, _ := create["body"].(map[string]interface{})
 		if body["type"] != "sheet" || body["file_extension"] != "xlsx" || body["token"] != testToken {
@@ -366,22 +474,18 @@ func TestWorkbookExport_DryRun(t *testing.T) {
 		}
 	})
 
-	t.Run("csv → 3 steps, with sub_id", func(t *testing.T) {
+	t.Run("csv includes sub_id from --sheet-id", func(t *testing.T) {
 		t.Parallel()
 		calls := parseDryRunAPI(t, WorkbookExport, []string{
 			"--url", testURL, "--file-extension", "csv", "--sheet-id", "sh1",
 			"--output-path", "/tmp/out.csv",
 		})
-		if len(calls) != 3 {
-			t.Fatalf("api calls = %d, want 3", len(calls))
+		if len(calls) != 1 {
+			t.Fatalf("api calls = %d, want 1", len(calls))
 		}
 		body, _ := calls[0].(map[string]interface{})["body"].(map[string]interface{})
-		if body["sub_id"] != "sh1" {
-			t.Errorf("csv export missing sub_id: %#v", body)
-		}
-		dl := calls[2].(map[string]interface{})
-		if !strings.Contains(dl["url"].(string), "/export_tasks/file/") {
-			t.Errorf("download url = %v", dl["url"])
+		if body["type"] != "sheet" || body["sub_id"] != "sh1" {
+			t.Errorf("csv export body = %#v (want type=sheet, sub_id=sh1)", body)
 		}
 	})
 
@@ -396,92 +500,6 @@ func TestWorkbookExport_DryRun(t *testing.T) {
 	})
 }
 
-func TestWorkbookExportDownloadErrorClassification(t *testing.T) {
-	t.Parallel()
-
-	t.Run("preserves typed request errors", func(t *testing.T) {
-		t.Parallel()
-		in := errs.NewAPIError(errs.SubtypeServerError, "typed upstream").WithCode(123)
-		got := sheetsDownloadRequestError(in)
-		if got != in {
-			t.Fatalf("typed error was not preserved: got %T %v", got, got)
-		}
-	})
-
-	t.Run("wraps raw request errors as network transport", func(t *testing.T) {
-		t.Parallel()
-		got := sheetsDownloadRequestError(errors.New("dial refused"))
-		p, ok := errs.ProblemOf(got)
-		if !ok {
-			t.Fatalf("expected typed problem, got %T %v", got, got)
-		}
-		if p.Category != errs.CategoryNetwork || p.Subtype != errs.SubtypeNetworkTransport {
-			t.Fatalf("problem = %s/%s, want %s/%s", p.Category, p.Subtype, errs.CategoryNetwork, errs.SubtypeNetworkTransport)
-		}
-	})
-
-	tests := []struct {
-		name          string
-		status        int
-		wantCategory  errs.Category
-		wantSubtype   errs.Subtype
-		wantRetryable bool
-	}{
-		{
-			name:          "5xx is retryable network server error",
-			status:        http.StatusBadGateway,
-			wantCategory:  errs.CategoryNetwork,
-			wantSubtype:   errs.SubtypeNetworkServer,
-			wantRetryable: true,
-		},
-		{
-			name:         "404 is API not found",
-			status:       http.StatusNotFound,
-			wantCategory: errs.CategoryAPI,
-			wantSubtype:  errs.SubtypeNotFound,
-		},
-		{
-			name:          "429 is retryable API rate limit",
-			status:        http.StatusTooManyRequests,
-			wantCategory:  errs.CategoryAPI,
-			wantSubtype:   errs.SubtypeRateLimit,
-			wantRetryable: true,
-		},
-		{
-			name:         "other 4xx is API unknown",
-			status:       http.StatusForbidden,
-			wantCategory: errs.CategoryAPI,
-			wantSubtype:  errs.SubtypeUnknown,
-		},
-	}
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-			got := sheetsDownloadHTTPStatusError(&larkcore.ApiResp{
-				StatusCode: tt.status,
-				RawBody:    []byte("body"),
-				Header:     http.Header{larkcore.HttpHeaderKeyLogId: []string{"log123"}},
-			})
-			p, ok := errs.ProblemOf(got)
-			if !ok {
-				t.Fatalf("expected typed problem, got %T %v", got, got)
-			}
-			if p.Category != tt.wantCategory || p.Subtype != tt.wantSubtype {
-				t.Fatalf("problem = %s/%s, want %s/%s", p.Category, p.Subtype, tt.wantCategory, tt.wantSubtype)
-			}
-			if p.Code != tt.status {
-				t.Fatalf("code = %d, want %d", p.Code, tt.status)
-			}
-			if p.LogID != "log123" {
-				t.Fatalf("log_id = %q, want log123", p.LogID)
-			}
-			if p.Retryable != tt.wantRetryable {
-				t.Fatalf("retryable = %v, want %v", p.Retryable, tt.wantRetryable)
-			}
-		})
-	}
-}
-
 // assertInputEquals compares the decoded tool input map against the wanted
 // fields. Extra fields in `got` are allowed (defaults, optional fields);
 // every key in `want` must match exactly.

shortcuts/sheets/lark_sheet_write_cells.go

@@ -197,12 +197,12 @@ func cellsSetStyleInput(runtime flagView, token, sheetID, sheetName string) (map
 	return input, nil
 }
 
-// CsvPut wraps set_range_from_csv: dump a CSV blob into a sheet, only writing
-// plain values. Use +cells-set for anything richer (formula / style / note).
+// CsvPut wraps set_range_from_csv: dump a CSV blob into a sheet. A cell whose
+// text starts with = is evaluated as a formula; use +cells-set for styles / notes / images.
 var CsvPut = common.Shortcut{
 	Service:     "sheets",
 	Command:     "+csv-put",
-	Description: "Paste RFC-4180 CSV into a sheet at --start-cell (plain values only, auto-expands sheet if needed).",
+	Description: "Paste RFC-4180 CSV into a sheet at --start-cell (values or formulas: a leading = is evaluated as a formula; no styles / comments; auto-expands sheet if needed).",
 	Risk:        "write",
 	Scopes:      []string{"sheets:spreadsheet:write_only"},
 	AuthTypes:   []string{"user", "bot"},

shortcuts/sheets/lark_sheet_write_cells_test.go

@@ -4,6 +4,7 @@
 package sheets
 
 import (
+	"fmt"
 	"strings"
 	"testing"
 
@@ -427,6 +428,44 @@ func TestCellsSet_RequiresJSONArray(t *testing.T) {
 	}
 }
 
+// TestCellsSet_RejectsUnsupportedMentionType pins the mention_type enum in
+// data/flag-schemas.json (synced from the upstream tool schema): a rich_text
+// mention whose mention_type is outside MENTION_FILE_TYPE (here 6 = cloud
+// shared folder) is rejected by the schema validator at flag-parse time,
+// before it can reach the server and blow up pb serialization
+// ("mentionFileInfo.fileType: enum value expected").
+func TestCellsSet_RejectsUnsupportedMentionType(t *testing.T) {
+	t.Parallel()
+	cells := `[[{"rich_text":[{"type":"mention","text":"x","mention_type":6,"mention_token":"t"}]}]]`
+	stdout, stderr, err := runShortcutCapturingErr(t, CellsSet, []string{
+		"--url", testURL, "--sheet-id", testSheetID,
+		"--range", "A1", "--cells", cells, "--dry-run",
+	})
+	if err == nil {
+		t.Fatalf("expected validation error; stdout=%s stderr=%s", stdout, stderr)
+	}
+	combined := stdout + stderr + err.Error()
+	if !strings.Contains(combined, "mention_type") || !strings.Contains(combined, "not in enum") {
+		t.Errorf("expected mention_type enum guard; got=%s|%s|%v", stdout, stderr, err)
+	}
+}
+
+// TestCellsSet_AllowsValidMentionTypes confirms the guard lets through a
+// user @mention (mention_type 0) and a render-supported file type (22 = DOCX).
+func TestCellsSet_AllowsValidMentionTypes(t *testing.T) {
+	t.Parallel()
+	for _, mt := range []int{0, 22} {
+		cells := fmt.Sprintf(`[[{"rich_text":[{"type":"mention","text":"x","mention_type":%d,"mention_token":"t"}]}]]`, mt)
+		stdout, stderr, err := runShortcutCapturingErr(t, CellsSet, []string{
+			"--url", testURL, "--sheet-id", testSheetID,
+			"--range", "A1", "--cells", cells, "--dry-run",
+		})
+		if err != nil {
+			t.Errorf("mention_type %d: unexpected error: stdout=%s stderr=%s err=%v", mt, stdout, stderr, err)
+		}
+	}
+}
+
 // TestCellsSetImage_DryRun verifies the 2-step plan (upload + embed) is
 // rendered, including the parent_type=sheet_image upload metadata.
 func TestCellsSetImage_DryRun(t *testing.T) {

shortcuts/sheets/shortcuts.go

@@ -38,8 +38,11 @@ func shortcutList() []common.Shortcut {
 		SheetHide,
 		SheetUnhide,
 		SheetSetTabColor,
+		SheetShowGridline,
+		SheetHideGridline,
 		WorkbookCreate,
 		WorkbookExport,
+		WorkbookImport,
 
 		// lark_sheet_sheet_structure
 		SheetInfo,
@@ -56,6 +59,7 @@ func shortcutList() []common.Shortcut {
 		CellsGet,
 		CsvGet,
 		DropdownGet,
+		TableGet,
 
 		// lark_sheet_search_replace
 		CellsSearch,
@@ -67,6 +71,7 @@ func shortcutList() []common.Shortcut {
 		CellsSetImage,
 		CsvPut,
 		DropdownSet,
+		TablePut,
 
 		// lark_sheet_range_operations
 		CellsClear,
@@ -103,5 +108,10 @@ func shortcutList() []common.Shortcut {
 		CellsBatchClear,
 		DropdownUpdate,
 		DropdownDelete,
+
+		// lark_sheet_history
+		HistoryList,
+		HistoryRevert,
+		HistoryRevertStatus,
 	}
 }

skills/lark-sheets/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: lark-sheets
-version: 2.0.0
-description: "飞书电子表格：创建和操作电子表格。支持创建表格、管理工作表与行列结构（增删/合并/调整尺寸/隐藏/冻结）、读写单元格（值/公式/样式/批注/单元格图片）、查找替换、多操作原子批量更新，以及图表、透视表、条件格式、筛选器、迷你图、浮动图片等对象的创建与维护。当用户需要创建电子表格、管理工作表、批量读写或编辑数据、统计汇总与可视化、表格美化、公式计算（含 Excel 公式迁移）等任务时使用。若用户是想按名称或关键词搜索云空间（云盘/云存储）里的表格文件，请改用 lark-drive 的 drive +search 先定位资源。当用户给出 doubao.com 的 /sheets/ URL/token 时，也应直接使用本 skill，不要因为域名不是飞书而回退到 WebFetch；路由依据是 URL 路径模式和 token，而不是域名。仅针对飞书在线电子表格，不适用于本地 Excel 文件。"
+version: 3.0.0
+description: "飞书电子表格：创建和操作电子表格。支持创建表格、管理工作表与行列结构（增删/合并/调整尺寸/隐藏/冻结）、读写单元格（值/公式/样式/批注/单元格图片）、查找替换、多操作原子批量更新，以及图表、透视表、条件格式、筛选器、迷你图、浮动图片等对象的创建与维护。当用户需要创建电子表格、管理工作表、批量读写或编辑数据、统计汇总与可视化、表格美化、公式计算（含 Excel 公式迁移）、金融/财务建模（DCF、三张表、预算、Sensitivity 等）等任务时使用。若用户是想按名称或关键词搜索云空间（云盘/云存储）里的表格文件，请改用 lark-drive 的 drive +search 先定位资源。当用户给出 doubao.com 的 /sheets/ URL/token 时，也应直接使用本 skill，不要因为域名不是飞书而回退到 WebFetch；路由依据是 URL 路径模式和 token，而不是域名。"
 metadata:
   requires:
     bins: ["lark-cli"]
@@ -40,18 +40,25 @@ metadata:
 | --- | --- | --- |
 | 读数据（纯值 / CSV） | `+csv-get`（范围用 `--range`） | — |
 | 读值 + 公式 / 样式 / 批注 | `+cells-get --include value,formula,style,comment,data_validation` | `--with-styles`、`--with-merges`、`--include-merged-cells` |
-| 写纯值（整块 CSV 平铺） | `+csv-put`（定位用 `--start-cell`，单个左上角锚点格；也接受 `--range` 别名，区间自动取左上角） | — |
+| 写纯文本值（整块 CSV 平铺，列里没有需保留的数值 / 日期语义） | `+csv-put`（定位用 `--start-cell`，单个左上角锚点格；也接受 `--range` 别名，区间自动取左上角） | — |
+| 写带类型的数据到**已有**表（列里有数字 / 金额 / 百分比 / 日期 / 计数，要可排序 / 求和 / 入图表 / 透视） | `+table-put`（列显式声明 `type` + `format`，类型保真；来源不限 DataFrame——Counter / dict / list 同理，详见 write-cells） | 在本地把数字拼成 `"$1,234"` / `"30.5%"` 字符串再 `+csv-put`（会落成文本、丢失计算能力） |
+| **新建**电子表格并写带类型的数据（类型保真需求同上，但目标表还不存在） | `+workbook-create --sheets`（协议与 `+table-put` 同构、一步建表 + typed 写入，无需先建空表再 `+table-put`；date / number 不丢，详见 workbook） | 用 `--values` 灌日期 / 数字（会落成文本、丢类型） |
 | 写值 / 公式 / 样式 | `+cells-set`（定位用 `--range`） | — |
+| 插图：图片**绑定到某条记录**、随行走（凭证 / 证件照 / 商品图 / 头像 / 二维码 / 每行配图） | `+cells-set-image`（单格 `--range`，嵌入单元格内） | — |
+| 插图：**自由摆放、不绑数据**的装饰 / 标识（logo / 水印 / 封面大图 / banner） | `+float-image-create`（浮动图片，自由定位 + 尺寸 + 层级） | — |
 | 查找单元格 | `+cells-search`（关键字用 `--find`） | `+cells-find`、`+find`、`--query` |
 | 查找并替换 | `+cells-replace` | — |
 | 看子表结构（合并 / 行高列宽 / 冻结 / 隐藏） | `+sheet-info` | `+sheet-get`、`+structure-get`、`+sheet-structure-get` |
 | 看工作簿 / 子表清单 | `+workbook-info` | — |
 | 导出 xlsx / 单表 csv | `+workbook-export` | — |
+| 导入本地 xlsx/xls/csv 文件为新表格 | `+workbook-import --file ./x.xlsx`（仅导入为电子表格；要导成多维表格走 `drive +import --type bitable`） | 把 .xlsx 在本地读成数据再 `+workbook-create` 重灌（丢原格式、低效） |
 | 清除内容 / 格式 | `+cells-clear`（范围维度用 `--scope`，取值 content / formats / all） | `--type` |
 | 批量清除多区域 | `+cells-batch-clear`（`--scope`） | `--target` |
 | 调整列宽 / 行高 | `+cols-resize` / `+rows-resize`（行、列是两个独立命令） | `--dimension`（无此 flag） |
 | 分组汇总 / 透视 | `+pivot-create`（默认不传落点 flag → 自动新建子表，零覆盖） | 用 SUMIF / 本地脚本拼一张假透视表 |
 
+> ⚠️ **两种图片别选错**：图若**绑定某条记录、要随行排序 / 筛选 / 增删**（凭证 / 证件照 / 每行配图，话里带「对应 / 每行 / 这列」等绑定词）→ 单元格图片 `+cells-set-image`；只是自由摆放的装饰（logo / 水印 / 封面）→ 浮动图片 `+float-image-create`。别因「浮动图更好控制 / 更熟」默认选浮动图。
+> ⚠️ **纯文本还是数值语义**：要写的列里有数字 / 金额 / 百分比 / 日期 / 计数 → `+table-put`（写入已有表；声明 `type` + `format`，保留排序 / 求和 / 图表 / 透视能力；**目标表还不存在就用 `+workbook-create --sheets`**，同 typed 协议、一步建表 + 写入，别先建空表再 `+table-put`）；只有纯文本才用 `+csv-put`。两者写完显示可以完全相同，但 `+csv-put` 落的是文本、不能参与计算——别把数值在本地拼成带 `$` / `%` 的字符串再走 `+csv-put`。
 > ⚠️ **定位 flag**：`+cells-get` / `+cells-set` / `+csv-get` 用 `--range`；`+csv-put` 规范用 `--start-cell`（单个左上角锚点格），也接受 `--range` 别名（区间自动取左上角），二者择一即可。
 > ⚠️ **读取附加信息**一律走 `+cells-get --include …`，**没有** `--with-styles` 这类 flag；**看合并单元格**用 `+sheet-info` 的 `merged_cells`，不要在 `+cells-get` 里找 merge flag。
 
@@ -63,28 +70,28 @@ metadata:
 
 | Reference | 描述 |
 | --- | --- |
-| [飞书表格核心操作：分析、编辑与可视化](references/lark-sheets-core-operations.md) | 飞书表格核心操作工作流。当用户需要对已有的飞书表格进行查看、分析、编辑或可视化时使用。适用场景：数据查询与统计、公式计算、表格美化、创建图表/透视表、筛选排序、批量修改数据、调整表格结构等。即使用户没有明确说"飞书表格"，只要操作对象是已有的在线表格，都应触发此工作流。不适用于本地 Excel 文件操作。 |
-| [飞书表格样式与配色规范](references/lark-sheets-visual-standards.md) | 飞书表格样式与配色规范：表头/数据区/汇总行的颜色、字号、对齐、边框等取值标准，以及新增汇总行、追加行列继承原表风格、已有区域美化等典型场景的决策流程与样式要点。工具调用参数细节请参考对应的 lark-sheets-write-cells / lark-sheets-range-operations / lark-sheets-batch-update。条件格式（高亮、标红、数据条、色阶）请使用 lark-sheets-conditional-format。仅针对飞书表格，不适用于本地 Excel 文件。 |
-| [飞书表格公式生成规则](references/lark-sheets-formula-translation.md) | Excel 公式到飞书表格公式的迁移与生成规则。核心目标不是保留 Excel 原语法，而是按飞书表格可执行规则重写公式，并在结果上尽量对齐 Excel。当用户要求把 Excel 公式改写成飞书表格公式，或需要生成飞书公式（尤其涉及 ARRAYFORMULA、原生数组函数、INDEX/OFFSET、MAP/LAMBDA、日期差、多层范围结果与二次展开）时使用。仅针对飞书在线表格，不适用于本地 Excel 文件执行。 |
+| [飞书表格核心操作：分析、编辑与可视化](references/lark-sheets-core-operations.md) | 飞书表格核心操作工作流。当用户需要对已有的飞书表格进行查看、分析、编辑或可视化时使用。适用场景：数据查询与统计、公式计算、表格美化、创建图表/透视表、筛选排序、批量修改数据、调整表格结构等。即使用户没有明确说"飞书表格"，只要操作对象是已有的在线表格，都应触发此工作流。 |
+| [飞书表格样式与配色规范](references/lark-sheets-visual-standards.md) | 飞书表格样式与配色规范：表头/数据区/汇总行的颜色、字号、对齐、边框等取值标准，以及新增汇总行、追加行列继承原表风格、已有区域美化等典型场景的决策流程与样式要点。工具调用参数细节请参考对应的 lark-sheets-write-cells / lark-sheets-range-operations / lark-sheets-batch-update。条件格式（高亮、标红、数据条、色阶）请使用 lark-sheets-conditional-format。 |
+| [飞书表格公式生成规则](references/lark-sheets-formula-translation.md) | Excel 公式到飞书表格公式的迁移与生成规则。核心目标不是保留 Excel 原语法，而是按飞书表格可执行规则重写公式，并在结果上尽量对齐 Excel。当用户要求把 Excel 公式改写成飞书表格公式，或需要生成飞书公式（尤其涉及 ARRAYFORMULA、原生数组函数、INDEX/OFFSET、MAP/LAMBDA、日期差、多层范围结果与二次展开）时使用。 |
 
 ### 按对象的工具参考（含 shortcut）
 
 | Reference | 描述 |
 | --- | --- |
-| [Lark Sheet Workbook](references/lark-sheets-workbook.md) | 管理飞书表格的工作簿结构（子表列表及元数据）。当用户提到"看看这个表格有什么"、"表格结构"、"有哪些 sheet"、"新建一个 sheet"、"删除这个工作表"、"重命名"、"复制一份"、"移动到前面"时使用。仅针对飞书表格。 |
-| [Lark Sheet Sheet Structure](references/lark-sheets-sheet-structure.md) | 管理飞书表格的子表结构与布局。适用场景：查看行高、列宽、隐藏行列、合并单元格等布局信息，以及"插入一行"、"删除这列"、"隐藏行"、"冻结表头"、行列分组（大纲折叠/展开）等操作。行列大纲仅在用户明确提到"行分组"、"列分组"、"大纲"、"outline"时才触发，"按XXX分组"等数据分组场景请使用 lark-sheets-pivot-table。如需在表尾追加数据，应先通过此 skill 插入行，再通过 lark-sheets-write-cells 写入。仅针对飞书表格。 |
-| [Lark Sheet Read Data](references/lark-sheets-read-data.md) | 读取飞书表格中的单元格数据。当用户需要"看看数据"、"分析数据"、"统计/汇总"时使用；也适用于需要查看公式、样式、批注等详细信息的场景。仅针对飞书表格。 |
-| [Lark Sheet Search & Replace](references/lark-sheets-search-replace.md) | 在飞书表格中搜索和替换文本，支持限定范围、大小写匹配、精确匹配、正则表达式。当用户需要"查找"、"搜索"、"定位"某个值，或"替换"、"批量修改文本"、"把 A 改成 B"时使用。不要用于理解表格结构（应读取数据）、不要用于数据分析（应读取数据后计算）、不要把用户操作动作中的关键词（如"汇总金额""统计数量"）当作搜索词。仅针对飞书表格。 |
-| [Lark Sheet Write Cells](references/lark-sheets-write-cells.md) | 向飞书表格的指定区域批量写入值、公式、样式、批注或单元格图片。适用场景：填写数据、设置公式、修改格式、添加批注、嵌入单元格图片（如需操作浮动图片，请使用 lark-sheets-float-image）；若只需把一块 CSV 纯值批量铺到表格上（不带公式/样式），直接使用 `+csv-put` 更短更快。追加数据需先通过 lark-sheets-sheet-structure 插入行列。仅针对飞书表格。 |
-| [Lark Sheet Range Operations](references/lark-sheets-range-operations.md) | 对飞书表格中指定区域执行结构性操作（不涉及写入单元格数据值）。适用场景：清除内容或格式（"清空"、"删除内容"、"去掉格式"）、合并/取消合并单元格、调整行高列宽（"加宽列"、"自适应列宽"）、移动/复制/填充/排序数据（"移动数据"、"复制到"、"自动填充"、"按某列排序"）。写入单元格数据请使用 lark-sheets-write-cells。仅针对飞书表格。 |
-| [Lark Sheet Batch Update](references/lark-sheets-batch-update.md) | 将多个飞书表格写入操作合并为一次批量执行，按顺序依次完成。适合需要连续执行多个写入操作的场景（如先修改结构再写入数据）。仅针对飞书表格。 |
-| [Lark Sheet Chart](references/lark-sheets-chart.md) | 管理飞书表格中的图表（柱形图、折线图、饼图、条形图、面积图、散点图、组合图、雷达图等）。当用户需要创建图表、修改图表样式或数据源、查看已有图表配置、删除图表时使用。也适用于用户提到"数据可视化"、"画个图"、"趋势分析"、"对比图"、"占比分析"、"做个图表"等数据可视化相关场景。仅针对飞书表格。 |
-| [Lark Sheet Pivot Table](references/lark-sheets-pivot-table.md) | 管理飞书表格中的数据透视表。当用户需要创建透视表、修改透视表的行列字段/聚合方式/筛选条件、查看已有透视表配置、删除透视表时使用。也适用于用户提到"分组汇总"、"交叉分析"、"按XXX统计"、"按字段分组"、"再分下组"、"多维分析"、"数据透视"等场景。仅针对飞书表格。 |
-| [Lark Sheet Conditional Format](references/lark-sheets-conditional-format.md) | 管理飞书表格中的条件格式规则（重复值高亮、单元格值比较、数据条、色阶、排名、自定义公式等）。当用户需要创建条件格式、修改已有规则的范围或样式、查看当前条件格式配置、删除规则时使用。也适用于用户提到"高亮"、"标红"、"颜色标记"、"数据条"、"色阶"、"条件样式"等场景。仅针对飞书表格。 |
-| [Lark Sheet Filter](references/lark-sheets-filter.md) | 管理飞书表格中的筛选器（filter）。当用户需要筛选数据（按文本/数值/颜色/日期条件过滤行）、查看已有筛选配置、修改或删除筛选器时使用。也适用于"只看"、"筛选出"、"仅保留符合条件的"等场景。仅针对飞书表格。 |
-| [Lark Sheet Filter View](references/lark-sheets-filter-view.md) | 管理飞书表格中的筛选视图（filter view）。当用户需要"建一个 XX 视图"、"保存这个筛选状态"、"切换不同筛选"、维护一个 sheet 上多份独立筛选配置时使用。视图与筛选器（filter）相互独立，可在同一 sheet 共存；视图的隐藏行仅在用户进入该视图时本地生效，不影响其他协作者。仅针对飞书表格。 |
-| [Lark Sheet Sparkline](references/lark-sheets-sparkline.md) | 管理飞书表格中的迷你图（折线迷你图、柱形迷你图、胜负迷你图）。当用户需要在单元格内嵌入小型图表来展示数据趋势时使用。也适用于"趋势线"、"单元格内图表"、"迷你图"等场景。注意：不等同于被禁用的 SPARKLINE() 公式函数。仅针对飞书表格。 |
-| [Lark Sheet Float Image](references/lark-sheets-float-image.md) | 管理飞书表格中的浮动图片。当用户需要在表格中插入浮动图片、调整图片位置和大小、查看已有浮动图片、删除图片时使用。也适用于"插入图片"、"添加 logo"、"放一张图"等场景。注意：如果用户需要将图片嵌入到某个单元格内部（单元格图片），请阅读 lark-sheets-write-cells。仅针对飞书表格。 |
+| [Lark Sheet Workbook](references/lark-sheets-workbook.md) | 管理飞书表格的工作簿结构（子表列表及元数据）。当用户提到"看看这个表格有什么"、"表格结构"、"有哪些 sheet"、"新建一个 sheet"、"删除这个工作表"、"重命名"、"复制一份"、"移动到前面"时使用。 |
+| [Lark Sheet Sheet Structure](references/lark-sheets-sheet-structure.md) | 管理飞书表格的子表结构与布局。适用场景：查看行高、列宽、隐藏行列、合并单元格等布局信息，以及"插入一行"、"删除这列"、"隐藏行"、"冻结表头"、行列分组（大纲折叠/展开）等操作。行列大纲仅在用户明确提到"行分组"、"列分组"、"大纲"、"outline"时才触发，"按XXX分组"等数据分组场景请使用 lark-sheets-pivot-table。如需在表尾追加数据，应先通过此 skill 插入行，再通过 lark-sheets-write-cells 写入。 |
+| [Lark Sheet Read Data](references/lark-sheets-read-data.md) | 读取飞书表格中的单元格数据。当用户需要"看看数据"、"分析数据"、"统计/汇总"时使用；也适用于需要查看公式、样式、批注等详细信息的场景。 |
+| [Lark Sheet Search & Replace](references/lark-sheets-search-replace.md) | 在飞书表格中搜索和替换文本，支持限定范围、大小写匹配、精确匹配、正则表达式。当用户需要"查找"、"搜索"、"定位"某个值，或"替换"、"批量修改文本"、"把 A 改成 B"时使用。不要用于理解表格结构（应读取数据）、不要用于数据分析（应读取数据后计算）、不要把用户操作动作中的关键词（如"汇总金额""统计数量"）当作搜索词。 |
+| [Lark Sheet Write Cells](references/lark-sheets-write-cells.md) | 向飞书表格的指定区域批量写入值、公式、样式、批注或单元格图片。适用场景：填写数据、设置公式、修改格式、添加批注、嵌入单元格图片（如需操作浮动图片，请使用 lark-sheets-float-image）；若只需把一块 CSV 批量铺到表格上（值或公式，不带样式/批注），直接使用 `+csv-put` 更短更快。追加数据需先通过 lark-sheets-sheet-structure 插入行列。 |
+| [Lark Sheet Range Operations](references/lark-sheets-range-operations.md) | 对飞书表格中指定区域执行结构性操作（不涉及写入单元格数据值）。适用场景：清除内容或格式（"清空"、"删除内容"、"去掉格式"）、合并/取消合并单元格、调整行高列宽（"加宽列"、"自适应列宽"）、移动/复制/填充/排序数据（"移动数据"、"复制到"、"自动填充"、"按某列排序"）。写入单元格数据请使用 lark-sheets-write-cells。 |
+| [Lark Sheet Batch Update](references/lark-sheets-batch-update.md) | 将多个飞书表格写入操作合并为一次批量执行，按顺序依次完成。适合需要连续执行多个写入操作的场景（如先修改结构再写入数据）。 |
+| [Lark Sheet Chart](references/lark-sheets-chart.md) | 管理飞书表格中的图表（柱形图、折线图、饼图、条形图、面积图、散点图、组合图、雷达图等）。当用户需要创建图表、修改图表样式或数据源、查看已有图表配置、删除图表时使用。也适用于用户提到"数据可视化"、"画个图"、"趋势分析"、"对比图"、"占比分析"、"做个图表"等数据可视化相关场景。 |
+| [Lark Sheet Pivot Table](references/lark-sheets-pivot-table.md) | 管理飞书表格中的数据透视表。当用户需要创建透视表、修改透视表的行列字段/聚合方式/筛选条件、查看已有透视表配置、删除透视表时使用。也适用于用户提到"分组汇总"、"交叉分析"、"按XXX统计"、"按字段分组"、"再分下组"、"多维分析"、"数据透视"等场景。 |
+| [Lark Sheet Conditional Format](references/lark-sheets-conditional-format.md) | 管理飞书表格中的条件格式规则（重复值高亮、单元格值比较、数据条、色阶、排名、自定义公式等）。当用户需要创建条件格式、修改已有规则的范围或样式、查看当前条件格式配置、删除规则时使用。也适用于用户提到"高亮"、"标红"、"颜色标记"、"数据条"、"色阶"、"条件样式"等场景。 |
+| [Lark Sheet Filter](references/lark-sheets-filter.md) | 管理飞书表格中的筛选器（filter）。当用户需要筛选数据（按文本/数值/颜色/日期条件过滤行）、查看已有筛选配置、修改或删除筛选器时使用。也适用于"只看"、"筛选出"、"仅保留符合条件的"等场景。 |
+| [Lark Sheet Filter View](references/lark-sheets-filter-view.md) | 管理飞书表格中的筛选视图（filter view）。当用户需要"建一个 XX 视图"、"保存这个筛选状态"、"切换不同筛选"、维护一个 sheet 上多份独立筛选配置时使用。视图与筛选器（filter）相互独立，可在同一 sheet 共存；视图的隐藏行仅在用户进入该视图时本地生效，不影响其他协作者。 |
+| [Lark Sheet Sparkline](references/lark-sheets-sparkline.md) | 管理飞书表格中的迷你图（折线迷你图、柱形迷你图、胜负迷你图）。当用户需要在单元格内嵌入小型图表来展示数据趋势时使用。也适用于"趋势线"、"单元格内图表"、"迷你图"等场景。注意：不等同于被禁用的 SPARKLINE() 公式函数。 |
+| [Lark Sheet Float Image](references/lark-sheets-float-image.md) | 管理飞书表格中的浮动图片。当用户需要在表格中插入浮动图片、调整图片位置和大小、查看已有浮动图片、删除图片时使用。也适用于"插入图片"、"添加 logo"、"放一张图"等场景。注意：如果用户需要将图片嵌入到某个单元格内部（单元格图片），请阅读 lark-sheets-write-cells。 |
 
 ## 公共 flag 速查
 
@@ -102,7 +109,7 @@ metadata:
 1. **spreadsheet 定位（必填）**：`--url` 与 `--spreadsheet-token` 二选一，**必须给其中之一**。两个都不给 → 校验报错 `specify at least one of --url or --spreadsheet-token`；两个都给 → 互斥冲突。
    - **`--url` 只解析 `/sheets/` 与 `/spreadsheets/` 两种链接**（从路径里抽出 token；也可以直接把裸 token 传给 `--spreadsheet-token`）。其它形态的链接不会被解析成表格 token。
    - ⚠️ **`/wiki/` 知识库链接不能直接当表格定位用**：wiki 链接背后可能是电子表格，也可能是文档 / 多维表格等其它类型，`--url` **不会**自动把 wiki token 解析成 spreadsheet token，直接传会失败。必须先把它解析成真实文档 token —— `lark-cli wiki +node-get --node-token "<wiki 链接或 token>"`，确认返回的 `obj_type` 为 `sheet` 后，取其 `obj_token` 作为 `--spreadsheet-token` 传入（解析细节见 [`../lark-wiki/SKILL.md`](../lark-wiki/SKILL.md)）。
-   - **例外**：`+workbook-create` 是新建一个还不存在的表格，**不接受任何 spreadsheet / sheet 定位 flag**（只有 `--title` / `--folder-token` / `--headers` / `--values`）。
+   - **例外**：`+workbook-create`（新建表 + 可选写入数据）与 `+workbook-import`（把本地文件导入为新表）都产出一张**还不存在**的表格，**不接受任何 spreadsheet / sheet 定位 flag**——`+workbook-create` 只有 `--title` / `--folder-token` / `--values` / `--styles` / `--sheets`，`+workbook-import` 只有 `--file`（必填）/ `--folder-token` / `--name`。
 2. **sheet 定位（公共四件套 shortcut 必填）**：`--sheet-id` 与 `--sheet-name` 二选一，**必须给其中之一**。两个都不给 → 校验报错 `specify at least one of --sheet-id or --sheet-name`。
    - ⚠️ **不确定 sheet 名时禁止直接猜 `Sheet1`**：除非用户对话明确说出 sheet 名 / id，或上下文（之前的工具调用 / URL 锚点 `?sheet=xxx`）已经出现过具体值，否则**第一步先调 `+workbook-info --url "..."`**（或 `--spreadsheet-token`）拿 `sheets[].sheet_id` / `sheets[].title` 列表再选。中文环境下子表常叫"数据" / "Sheet"（无数字）/ "工作表 1" / 业务名，猜 `Sheet1` 大概率撞 `sheet not found`，比先查多耗一次失败调用 + 重试。
    - ⚠️ **`--range` 里的 `Sheet1!` 前缀不能替代 sheet 定位**：即使写了 `--range 'Sheet1!A1:B2'`，仍**必须**额外传 `--sheet-id` 或 `--sheet-name`，否则照样报上面的错。

skills/lark-sheets/references/lark-sheets-batch-update.md

@@ -22,7 +22,7 @@
 
 当同一工具需要对多个区域重复调用时，**必须**改用 `+batch-update` 合并为单次请求——`+batch-update` 是原子提交（要么全成功要么整批回滚）；逐个调用非原子，中途失败会留下半成品。
 
-**`+dropdown-update` 的选项模式（`--options` / `--source-range` 二选一）+ 配色规则**（`--colors` 长度可短不能长、必须配 `--highlight=true` 才生效、不传按内置 10 色色板循环补色）见 [`lark-sheets-write-cells`](./lark-sheets-write-cells.md) 的「Dropdown 选项 + 配色」节，本 skill 不重复。`+dropdown-delete` 不涉及这些 flag。
+**`+dropdown-update` 的选项模式（`--options` / `--source-range` 二选一）+ 配色规则**（`--colors` 长度可短不能长、必须配 `--highlight=true` 才生效、不传按内置 10 色色板循环补色）见 [`lark-sheets-write-cells`](./lark-sheets-write-cells.md) 的「Dropdown 选项 + 配色」节，本文不重复。`+dropdown-delete` 不涉及这些 flag。
 
 ## Shortcuts
 

skills/lark-sheets/references/lark-sheets-chart.md

@@ -36,7 +36,8 @@
   - **默认情况（inline 模式）**：`refs` 范围**应包含表头行**（首行/首列即系列名），且范围要精确覆盖目标数据，不要多选或少选。
   - **合并标题行要跳过**：如果表格在表头上方存在合并的标题行（如"员工统计表"横跨多列的大标题），`refs` 必须跳过标题行、从真正的列标题行开始。例如表头在第 3 行、数据在第 4-20 行，则 `refs` 应为 `A3:G20` 而非 `A1:G20`。包含合并标题行会导致列名识别错误、表头被当作数据参与聚合计算。
   - **数据与表头分离时必须用 detached 模式**：当 `refs` 只覆盖完整数据的一个子集（按筛选/分组只画其中一段），而真正的语义表头在该子集之外时，**必须**设置 `snapshot.data.headerMode='detached'`：refs 仅传纯数据范围，维度名/系列名通过 `snapshot.data.dim1.serie.nameRef` / `snapshot.data.dim2.series[].nameRef` 指向真正的表头单元格。详见下文"硬性规则：数据与表头分离场景必须使用 detached 模式"。
-- **axes[].label 不接受 `format` / `number_format` 字段**：想给坐标轴数值加千分位、百分号等格式化时，不要在 `axes[i].label` 里传 `format` 或 `number_format`（schema 未定义，会报 `unexpected property "format" is not defined in schema`）。数值格式化统一在源数据单元格的 `cell_styles.number_format` 里设置（写 `+cells-set` 时），图表会沿用单元格格式。
+- **axes[].label 不接受 `format` / `number_format` 字段**：想给坐标轴数值加千分位、百分号等格式化时，不要在 `axes[i].label` 里传 `format` 或 `number_format`（schema 未定义，会报 `unexpected property "format" is not defined in schema`）。数值格式化统一在源数据单元格的 `cell_styles.number_format` 里设置（写 `+cells-set` 时），图表会沿用单元格格式。**日期轴同理**：横轴显示成 `45297` 这类 Excel 序列号，是因为源日期列没设日期格式——给源列设 `number_format="yyyy-mm-dd"` 后横轴才会显示成日期（反例：折线图横轴日期显示为序列号）。大数值轴显示科学计数法同理，给源列设整数 / 千分位格式（反例：透视表数值轴显示科学计数法）。
+- **轴口径要对齐用户要的指标**：用户要"占比 / 比例"时，**纵轴应是百分比**——用饼图，或柱 / 条形图设 `stack.percentage: true` 让纵轴变 %，并把数据源指向占比列 / 让数据标签显示百分比；不要交付纵轴仍是原始计数的图（反例：要求看各类占比，却用普通堆积柱、纵轴是 0–350 的人数而非百分比）。
 - **创建后必须验证**：图表创建后必须调用 `+chart-list` 验证配置是否正确
 
 > **⚠️ 硬性规则：当用户通过列标题名称（而非列索引）指定横轴/纵轴系列时，必须先读取表格首行（表头）来确定列名与列索引的对应关系，再设置 `dim1`/`dim2` 的 `index`。**
@@ -84,15 +85,17 @@
 
 1. **查尺寸**：`+workbook-info` 拿该 sheet 的 `row_count` / `column_count`（下文记为 rowCount / columnCount；`+sheet-info` 只返回布局，不含行列总数）。
 2. **估跨度**：默认单元格 **105 px 宽 × 27 px 高**，`needCols = ceil(width/105)`，`needRows = ceil(height/27)`。
-3. **校验**：`position.row + needRows ≤ rowCount` 且 `col_idx + needCols ≤ columnCount`（col 按 A=0、B=1、…、Z=25、AA=26… 换算）。
+3. **校验**：`position.row + needRows ≤ rowCount` 且 `col_idx + needCols ≤ columnCount`（`position.row` 为 **0-based**：首行 = `row:0`，与 A1 区间 / `+dim-insert --position` 的 1-based 行号不同；col 按 A=0、B=1、…、Z=25、AA=26… 换算）。
 4. **不够就先扩表**，二选一，禁止硬塞越界位置：
    - **优先**放数据下方空区：`position = {row: data_end_row + 2, col: "A"}`；
    - 否则先调 `+dim-insert`（`lark-sheets-sheet-structure`）扩行/列，再 create。
 
+⚠️ **图表落点禁止压在已有数据矩形内**——必须落在数据区**右侧或下方的空白**，否则图表浮层会遮挡原始数据被判失败（反例：折线图落在数据区中间，遮挡了下方原始数据）。
+
 **示例**：21 列 sheet 放 600×400 图 → `needCols=6, needRows=15`
 - ❌ `{row: 0, col: "W"}` — col=22 越界
 - ✅ `{row: 42, col: "A"}` — 放数据下方
-- ✅ 先 `+dim-insert --dimension column --start 21 --end 27`（在 U 列后插 6 列；U=index 20，after 即从 21 起），再放图到 `{row: 0, col: "V"}`
+- ✅ 先 `+dim-insert --position V --count 6`（在 V 列前插 6 列，即 U 列之后），再放图到 `{row: 0, col: "V"}`
 
 ## Shortcuts
 
@@ -147,9 +150,9 @@ _公共四件套 · 系统：`--yes`、`--dry-run`_
 _创建/更新的图表属性_
 
 **顶层字段**：
-- `position` (object) — 必填 { row: number, col: string }
+- `position` (object?) — 必填 { row: number, col: string }
 - `offset` (object?) — 可选 { row_offset?: number, col_offset?: number }
-- `size` (object) — 必填 { width: number, height: number }
+- `size` (object?) — 必填 { width: number, height: number }
 - `snapshot` (object?) — 图表快照配置 { title?: object, subTitle?: object, style?: object, legend?: oneOf, plotArea: object, …共 6 项 }
 
 ## Examples
@@ -164,24 +167,28 @@ _创建/更新的图表属性_
 
 > **`snapshot.data` 必填 `dim1.serie.index` 或 `dim2.series[].index` 之一**（1-based，对应 `refs.value` 范围内的列序）。schema 允许传空 `{}` 但 server 运行时强制：缺则被拒为 `snapshot.data.dim1.serie.index and dim2.series[].index are both missing; at least one must be set`，即便侥幸通过也只会渲染空图。
 
+> ⚠️ **含 `'Sheet'!` 前缀的 `--properties` 必须走 stdin 或 `@file`，不要用 inline 单引号**。`refs` / `nameRef` 里的 sheet 前缀带单引号（`'Sheet1'!A1`），若塞进 inline 的 `--properties '{...}'`，bash 会把内层那对单引号吃掉（sheet 名带空格还会被拆成多个词），JSON 直接被破坏。下面示例统一用 `--properties - <<'JSON' … JSON`（heredoc 定界符加引号 = 不做 shell 替换），或 `--properties @file.json`（`@` 只接 cwd 下相对路径）。
+
 最小可用列图（inline 模式：refs 含表头行）：
 
 ```bash
 lark-cli sheets +chart-create --url "https://example.feishu.cn/sheets/shtXXX" \
-  --sheet-name "Sheet1" --properties '{
-    "position":{"row":42,"col":"A"},
-    "size":{"width":600,"height":400},
-    "snapshot":{
-      "data":{
-        "refs":[{"value":"'Sheet1'!A1:B10"}],
-        "dim1":{"serie":{"index":1}},
-        "dim2":{"series":[{"index":2}]}
-      },
-      "plotArea":{"plot":{"type":"column"}}
-    }
-  }'
+  --sheet-name "Sheet1" --properties - <<'JSON'
+{
+  "position":{"row":42,"col":"A"},
+  "size":{"width":600,"height":400},
+  "snapshot":{
+    "data":{
+      "refs":[{"value":"'Sheet1'!A1:B10"}],
+      "dim1":{"serie":{"index":1}},
+      "dim2":{"series":[{"index":2}]}
+    },
+    "plotArea":{"plot":{"type":"column"}}
+  }
+}
+JSON
 
-# 走文件（推荐配置较多时）
+# 或落到 cwd 下相对路径文件再用 @file
 lark-cli sheets +chart-create --url "..." --sheet-name "Sheet1" --properties @chart-config.json
 ```
 
@@ -190,7 +197,8 @@ lark-cli sheets +chart-create --url "..." --sheet-name "Sheet1" --properties @ch
 饼图比 column / bar 更复杂：`sectors` 是 object，里面再包一个**单数** `sector` 数组——CLI 不替你 normalize，写错路径会被 server schema 直接拒。
 
 ```bash
-lark-cli sheets +chart-create --url "..." --sheet-name "Sheet1" --properties '{
+lark-cli sheets +chart-create --url "..." --sheet-name "Sheet1" --properties - <<'JSON'
+{
   "position":{"row":24,"col":"F"},
   "size":{"width":600,"height":450},
   "snapshot":{
@@ -208,7 +216,8 @@ lark-cli sheets +chart-create --url "..." --sheet-name "Sheet1" --properties '{
       "dim2":{"series":[{"index":2,"aggregateType":"sum"}]}
     }
   }
-}'
+}
+JSON
 ```
 
 **数据与表头分离（必须用 `detached` + `nameRef`）**：
@@ -216,7 +225,8 @@ lark-cli sheets +chart-create --url "..." --sheet-name "Sheet1" --properties '{
 场景：周度销量明细表，真实表头在第 1 行（A1=周次、C1=订单量、D1=退款量），数据按 B 列"店铺"分段；用户只要"3 号店"那一段（第 11–17 行）。
 
 ```bash
-lark-cli sheets +chart-create --url "..." --sheet-name "Sheet2" --properties '{
+lark-cli sheets +chart-create --url "..." --sheet-name "Sheet2" --properties - <<'JSON'
+{
   "position":{"row":7,"col":"F"},
   "size":{"width":600,"height":360},
   "snapshot":{
@@ -233,7 +243,8 @@ lark-cli sheets +chart-create --url "..." --sheet-name "Sheet2" --properties '{
       ]}
     }
   }
-}'
+}
+JSON
 ```
 
 约束：

skills/lark-sheets/references/lark-sheets-core-operations.md

@@ -2,7 +2,7 @@
 
 ## 概览
 
-面向"已有飞书表格"的核心工作流，核心原则：**先了解，再分析或写入，最后验证**。本文是方法论总纲；具体工具的参数细节、边界陷阱在对应子 skill，本文用指针引到那里，不重复展开。
+面向"已有飞书表格"的核心工作流，核心原则：**先了解，再分析或写入，最后验证**。本文是方法论总纲；具体工具的参数细节、边界陷阱在对应 reference，本文用指针引到那里，不重复展开。
 
 **三份「通用方法与规范」如何分工**（都不含 shortcut，按主题单一归属）：
 
@@ -12,21 +12,21 @@
 
 > **下面的铁律对所有任务一律生效**，即使你是被索引直接路由进 visual 或 formula 而没经过本文——编辑类任务务必先回到这里过一遍铁律。
 
-## 铁律（所有编辑类任务必须满足，子 skill 不得放宽）
+## 铁律（所有编辑类任务必须满足，各 reference 不得放宽）
 
-1. **最小改动**：除用户明示要改的单元格 / 列外，原表其它单元格、行列结构、Sheet 名、合并区、格式必须 1:1 保持。中间结果优先放原数据**右侧**；会与原数据混淆或要承载透视表 / 图表时才**新建空白 Sheet**。**禁止**擅自删 / 改名 / 隐藏 / 移动**已存在**的 Sheet（新建允许，节制使用）。
-2. **真实写回 + 回读校验**：交付必须是对在线表格的真实写入，并 `+csv-get` / `+cells-get` / `+<对象>-list` 回读校验。**严禁**只在文本里描述"已完成"、用普通公式 / 文本假装结构化对象、或只给占位而无真实写入。
+1. **最小改动**：除用户明示要改的单元格 / 列外，原表其它单元格、行列结构、Sheet 名、合并区、格式必须 1:1 保持。中间结果优先放原数据**右侧**；会与原数据混淆或要承载透视表 / 图表时才**新建空白 Sheet**。**禁止**擅自删 / 改名 / 隐藏 / 移动**已存在**的 Sheet（新建允许，节制使用）。**改写 / 转换类任务要精确圈定适用行列**：只对任务真正要求的对象做变换，**不该转的行 / 列保持原值 1:1**（典型反例：要求"统一翻译"时把本就是中文、应原样保留的评论也重新翻译；要求"改写某列格式"时连原始测量值也一并改动 → 应保留的原文被篡改）。
+2. **真实写回 + 回读校验**：交付必须是对在线表格的真实写入，并 `+csv-get` / `+cells-get` / `+<对象>-list` 回读校验。**严禁**只在文本里描述"已完成"、用普通公式 / 文本假装结构化对象、或只给占位而无真实写入。**收尾前必须确认产物文件真实存在 / 可导出**——别在没真正生成产物时只凭文本"已完成"就结束（反例：文本称已完成，实际没生成产物文件，等于没交付）。
 3. **读全再写，禁止只探前 N 行**：批量填充 / 补齐 / 修正类任务必须先确认**真实数据末行**再写，否则会漏写表尾（高频致命错误）。完整的"按表格形态分流读取 + `current_region` / `has_more` 兜底 + 真实末行确认"流程见 `lark-sheets-read-data` 的「确定数据范围的正确流程」。
-4. **公式优先于硬编码**：能用飞书公式表达的计算（总计 / 占比 / 增长率 / 提取 / 查找等）一律写公式而非静态值，源数据变化才能自动重算。用户口头的"分列 / 排序 / 求和 / 提取"也要落地为公式或原生工具（SORT / `TEXTBEFORE` / `MID` / 透视表 等）。Excel 公式迁移、数组语义、不支持函数清单一律以 `lark-sheets-formula-translation` 为唯一权威。
+4. **公式优先于硬编码**：能用飞书公式表达的计算（总计 / 占比 / 增长率 / 提取 / 查找等）一律写公式而非静态值，源数据变化才能自动重算。用户口头的"分列 / 排序 / 求和 / 提取"也要落地为公式或原生工具（SORT / `TEXTBEFORE` / `MID` / 透视表 等）。Excel 公式迁移、数组语义、不支持函数清单一律以 `lark-sheets-formula-translation` 为唯一权威。**即使用户没说"联动 / 自动更新"，凡是可由表内其它单元格推导的派生值（年龄=当年-出生年、占比=本类数/总数、达标=阈值判断、排名、各类分组汇总）默认就必须用公式**——用户默认期望派生列能随源数据重算，**离线 Python / 脚本算完写静态值，即便当前数值正确，改了源数据也不会自动更新，等于没满足"派生"的本意**（反例：年龄、月度汇总、占比、分组求和等派生列写死值，源数据一改结果就过时）。
 5. **续写 / 扩展必须继承样式**：续写、补齐、复制区块、新增行列时，**禁止**只读值只写值。必须连带 `cell_styles` + `border_styles` + 合并 + 行高一起继承。完整继承清单与做法见 `lark-sheets-write-cells` 的「新增列 / 新增行的样式继承」（`border_styles` 四边最易漏）。
 6. **多步写入优先 `+batch-update`**：多个连续写入、或同一工具对多个区域重复调用（多次 merge / resize / cells-set），必须合并为单次原子 `+batch-update`。语义与不可嵌套的限制见 `lark-sheets-batch-update`。
 7. **分组汇总必须用透视表**："按 X 统计 Y / 分组汇总 / 各部门数量金额"必须用 `+pivot-{create|update|delete}`（推荐省略 sheet_id 自动新建子表），**禁止**用 SUMIF / COUNTIF 或本地脚本覆盖原表替代。
-8. **任务拆成可验证 checklist**：落地前把指令拆成所有"独立可验证子要点"，每点一个 `assert`，全部通过才交付：多维度操作（按部门一/二/三级排序）每维一个 assert；多目标（删 N 行）每目标一个；多格式兼容（多种日期格式）每种至少一个样本；范围类（A1:H11 加边框）起 / 末行 / 末列三边界都核。只完成第一个要点（只排一级、只删 1 行）属违规。
+8. **任务拆成可验证 checklist**：落地前把指令拆成所有"独立可验证子要点"，每点一个 `assert`，全部通过才交付：多维度操作（按部门一/二/三级排序）每维一个 assert；多目标（删 N 行）每目标一个；多格式兼容（多种日期格式）每种至少一个样本；范围类（A1:H11 加边框）起 / 末行 / 末列三边界都核。只完成第一个要点（只排一级、只删 1 行）属违规。**题面 / 表头里写明的格式规范也是子要点**：表头注明"需标注某字段"就必须给对应单元格加规定前缀并逐条 assert 前缀存在（反例：漏加规定前缀，该要点即不达标）；"相同编号连续行合并"必须遍历所有相同编号组全部合并（反例：只合并了其中一部分组）。
 9. **全量处理要前置断言条数**：翻译 / 打标 / 批量公式落地等逐条任务，落地前把"预期处理条数"硬编码进代码，处理完 `assert actual == expected`。**严禁**输出"已完成前 N 条，剩余将继续"的半成品。
 
 ## 推荐工作流程
 
-1. **规划 skill 清单**：开工前一次性列出本任务要读的子 skill（避免读一个调一个），本轮已读过的不重复读。本 skill + `lark-sheets-workbook` 几乎每次都要。
+1. **规划 reference 清单**：开工前一次性列出本任务要读的 reference（避免读一个调一个），本轮已读过的不重复读。本文 + `lark-sheets-workbook` 几乎每次都要。
 2. **了解结构**：先 `+workbook-info` 拿子表列表 / 行列数 / 冻结位置（不可猜测，猜错会越界覆盖）；涉及合并 / 隐藏 / 分组 / 行高列宽再用 `lark-sheets-sheet-structure` 的 `+sheet-info`。
 3. **读取数据（按任务类型选路径，细则见 `lark-sheets-read-data`）**：
 
@@ -67,6 +67,7 @@
 - **喂给 CLI 的 CSV / JSON 用 UTF-8、不带 BOM**：BOM 会污染首格的值或触发 `invalid character` 解析错；脚本读写文件时显式指定 `encoding='utf-8'`。
 - **临时文件交给运行时的标准库**：用 `tempfile.gettempdir()` / `os.tmpdir()` 等取临时目录，不要硬编码固定路径；放在用户项目目录之外。
 - **命令失败先读错误再调整**：同一条命令失败后不要原样重发；先看 stderr 的报错（参数错误、缺依赖、解释器不可用等）定位原因，再决定换写法、补依赖或退回原生工具。
+- **写回的必须是纯单元格值，禁止把"值+样式标注"串当值写回**：本地脚本或某些 xlsx 解析库会把单元格渲染成 `甲方支行(V-Align: bottom)` 这种"值(样式)"字符串，CSV 字段还可能带包裹双引号。回写前必须**剥离括号样式标注、去掉残留引号**，只写原始值——否则样式描述会变成单元格的字面文本污染原数据（反例：排序后单元格值里被写进 `(V-Align: bottom)` 这类样式后缀文本，末尾还多一个双引号）。**排序本身优先用 `+range-sort` 原生工具**，不要"读出来本地排完再整列写回"，从根上避免这类回写污染。
 
 ## 公式策略
 

skills/lark-sheets/references/lark-sheets-filter-view.md

@@ -109,6 +109,17 @@ lark-cli sheets +filter-view-create --url "..." --sheet-id "$SID" \
   --properties '{"rules":[{"column_index":"C","conditions":[{"type":"number","compare_type":"greaterThan","values":[100]}]}]}'
 ```
 
+**`conditions[].type` × `compare_type` 取值**（`type` 决定可用的 `compare_type`；两者均必填）：
+
+| `type` | 可用 `compare_type` | `values` |
+|---|---|---|
+| `text` | `contains` / `doesNotContain` / `beginsWith` / `doesNotBeginWith` / `endsWith` / `doesNotEndWith` / `equals` / `notEquals` | 字符串数组 |
+| `number` | `equal` / `notEqual` / `greaterThan` / `greaterThanOrEqual` / `lessThan` / `lessThanOrEqual` / `between` / `notBetween` | 数值（或数值字符串）数组；`between` / `notBetween` 传两个边界 |
+| `multiValue` | `equal` / `notEqual` | 字符串数组（精确匹配其中任一值） |
+| `color` | `backgroundColor` / `foregroundColor` | 不传 `values`（按单元格颜色筛选） |
+
+> ⚠️ `text` 用 `equals` / `notEquals`（**带 s**），`number` / `multiValue` 用 `equal` / `notEqual`（**不带 s**）——别混。完整 schema 跑 `+filter-view-create --print-schema --flag-name properties`。
+
 > `--range` **必须覆盖表头行**（如 `A1:F1000`），不能只包含数据行；`--view-name` 重名时服务端自动改名。
 
 ### `+filter-view-update`

skills/lark-sheets/references/lark-sheets-filter.md

@@ -102,6 +102,17 @@ lark-cli sheets +filter-create --url "..." --sheet-id "$SID" \
   --properties '{"rules":[{"column_index":"B","conditions":[{"type":"multiValue","compare_type":"equal","values":["北京","上海"]}]}]}'
 ```
 
+**`conditions[].type` × `compare_type` 取值**（`type` 决定可用的 `compare_type`；两者均必填）：
+
+| `type` | 可用 `compare_type` | `values` |
+|---|---|---|
+| `text` | `contains` / `doesNotContain` / `beginsWith` / `doesNotBeginWith` / `endsWith` / `doesNotEndWith` / `equals` / `notEquals` | 字符串数组 |
+| `number` | `equal` / `notEqual` / `greaterThan` / `greaterThanOrEqual` / `lessThan` / `lessThanOrEqual` / `between` / `notBetween` | 数值（或数值字符串）数组；`between` / `notBetween` 传两个边界 |
+| `multiValue` | `equal` / `notEqual` | 字符串数组（精确匹配其中任一值） |
+| `color` | `backgroundColor` / `foregroundColor` | 不传 `values`（按单元格颜色筛选） |
+
+> ⚠️ `text` 用 `equals` / `notEquals`（**带 s**），`number` / `multiValue` 用 `equal` / `notEqual`（**不带 s**）——别混。完整 schema 跑 `+filter-create --print-schema --flag-name properties`。
+
 ### `+filter-update`
 
 > ⚠️ update 是覆盖式：`--properties` 中传新 `rules` 会替换旧组。如只想加一条，要带上已有的全部条件再追加。必填 `--range`。

skills/lark-sheets/references/lark-sheets-float-image.md

@@ -1,12 +1,13 @@
 # Lark Sheet Float Image
 
-> **单元格图片 vs 浮动图片**：飞书表格有两种图片类型，请根据需求选择正确的工具：
-> - **单元格图片**：图片嵌入在单元格内部，随单元格移动，属于单元格内容的一部分。→ 使用 `+cells-set`，在 `rich_text` 中设置 `type: "embed-image"`（见 lark-sheets-write-cells）。
-> - **浮动图片**（本 Skill）：图片悬浮在单元格上方，可自由指定位置、大小和层级，不属于任何单元格的内容。→ 使用本 Skill 的 `+float-image-{create|update|delete}`。
+> **选浮动图还是单元格图？只看一条**：这张图是不是**属于某条记录、要随那行一起排序 / 筛选 / 增删**？
+> - **是 → 单元格图片**（不在本 reference）：嵌进单元格、随行走。用 `+cells-set-image`（或 `+cells-set` 的 `rich_text` + `type: "embed-image"`，见 lark-sheets-write-cells）。典型：凭证 / 证件照 / 商品图 / 头像 / 二维码 / 每行配图；话里带「对应 / 每行 / 每条 / 这列」等绑定词即属此类。
+> - **否 → 浮动图片**（本 reference）：自由摆放、不绑数据的装饰 / 标识（logo / 水印 / 封面大图 / banner）。
+> - ⚠️ 别凭"浮动图位置尺寸更好控制 / 更熟"就选它——那是按操作便利选，不是按场景选；用浮动图承载"对应某记录"的图会在增删行 / 排序后错位。
 
 ## 真对象硬约束
 
-当用户要求"插入图片 / 添加 logo / 放一张图"时，**必须**通过 `+float-image-{create|update|delete}`（浮动图片）或 `+cells-set` 的 `embed-image`（单元格图片）创建真实的图片对象。**禁止**只在文本回复中给出图片链接 / 描述图片内容代替插入。判断标准：交付后 `+float-image-list` 或单元格 `rich_text` 必须能读到该图片对象。
+当用户要求"插入图片 / 添加 logo / 放一张图"时，**必须**通过 `+float-image-{create|update|delete}`（浮动图片）或 `+cells-set-image` / `+cells-set` 的 `embed-image`（单元格图片）创建真实的图片对象。**禁止**只在文本回复中给出图片链接 / 描述图片内容代替插入。判断标准：交付后 `+float-image-list` 或单元格 `rich_text` 必须能读到该图片对象。
 
 ## 使用场景
 
@@ -20,7 +21,7 @@
 典型工作流：先读取现有浮动图片了解配置 → 执行创建/更新/删除 → **必须再次读取验证结果**。
 
 **常见配置错误（必须注意）**：
-- **单元格图片 vs 浮动图片选择错误**：如果用户希望图片嵌入单元格内部（随单元格移动），应使用 `+cells-set` 的 `rich_text` + `embed-image`，而非本 Skill
+- **单元格图片 vs 浮动图片选择错误（最高频）**：图与某条记录一一对应、要随行排序 / 筛选 / 增删时，应走 `+cells-set-image`（见顶部判别），用浮动图会错位。
 - **图片位置参数要精确**：锚点单元格的行列索引和偏移量决定了图片位置，设置不当会导致图片遮挡数据
 - **创建后必须验证**：调用 `+float-image-list` 确认图片位置和大小正确
 

skills/lark-sheets/references/lark-sheets-range-operations.md

@@ -40,6 +40,10 @@
 
 **典型反例**：默认列宽 11 但内容含 12+ 字符的中文 / 含单位的数值（如 `109.10μmol/L`）/ 长数字未设 `number_format` 显示为科学计数法 —— 用户在结果表里看不到完整原值。
 
+**打印场景控制总宽（用户说"适合打印 / A4 / 打印范围"时必做）**：扩单列宽防截断的同时，**所有列宽之和要落在纸张可打印宽度内**——A4 横向约 ≤ 102 个半角字符（约 1000px），纵向约 ≤ 70 个字符。超宽时不要无限加宽，改用 `cell_styles.word_wrap="auto-wrap"` + 调高行高，或缩窄非关键列，让整表在一页内（反例：总列宽远超 A4 可打印宽度，且长文本行高不够被截断）。
+
+**只加宽承载新内容的列，不改动原有列的列宽**：列宽自适应**只针对新增 / 真正放不下新内容的列**；原表已有列的列宽**禁止重新计算、禁止缩小**——即便你估算的"理想宽度"与原值不同，只要原内容没被截断就不要动它。无差别地把所有列重设一遍宽度（哪怕只 ±1）都属于破坏原文件视觉格式（反例：填完数据后顺手把原有列的列宽从 16 改成 17，与原附件不一致，破坏了原视觉格式）。
+
 **⚠️ 合并单元格安全操作规则**（`+cells-{merge|unmerge}` 必读）：
 
 1. **先读后写**：操作前必须用 `+sheet-info --include merges` 或 `+cells-get` 识别已有合并区域（特征：多个连续单元格中只有左上角有值，其余为空）。
@@ -192,7 +196,7 @@ _排序条件列表（仅 sort 操作）_
 
 ## Examples
 
-> ⚠️ 本 skill 派生的 shortcut 跨 3 个分组：`+rows-resize` / `+cols-resize` → 工作表，`+cells-*` → 单元格，`+range-*` → 区域。skill 视角统一在这里讲解。
+> ⚠️ 本 reference 派生的 shortcut 跨 3 个分组：`+rows-resize` / `+cols-resize` → 工作表，`+cells-*` → 单元格，`+range-*` → 区域。这里统一从区域操作视角讲解。
 
 公共四件套：所有 shortcut 顶部排列 `--url` / `--spreadsheet-token` / `--sheet-id` / `--sheet-name`（XOR）。
 

skills/lark-sheets/references/lark-sheets-read-data.md

@@ -13,19 +13,22 @@
 
 预探后必须在公式 / 筛选条件里用 `IFERROR` / `IFS` / 提取数值的辅助列处理所有变体；不能为了通过 head(10) 的样本就直接落地。一旦设计的逻辑只覆盖 sample 中出现的格式，就属于违规。
 
+⚠️ **大数字（15 位以上的身份证 / 参考号 / 流水号）做去重 / 比较时禁止用 `+csv-get` 的显示值**：`+csv-get` 返回的是**格式化显示值**，15 位以上数字会被显示成 `1.04E+14` 这类科学计数法——多个本不相同的号在显示层全变成同一个 `1.04E+14`，拿去判重会**整列误判为重复**。比较 / 去重 / 匹配大数字时必须改用 `+cells-get`（取原始精确值）或把该列读为文本，禁止用 csv-get 的科学计数显示值（反例：大批长参考号被显示成科学计数后，互不相同的号全变成同一个值，被当成整列重复并错误高亮）。
+
 ## 使用场景
 
-读取。从飞书表格中读取单元格数据。本 reference 覆盖 3 个 shortcut，按读取目的选择：
+读取。从飞书表格中读取单元格数据。本 reference 覆盖 4 个 shortcut，按读取目的选择：
 
 | 读取目的 | 用这个 shortcut | 数据去向 | 说明 |
 |---------|----------------|---------|------|
-| 快速查看纯值数据、批量处理 | `+csv-get` | 对话上下文 | 返回 CSV 文本（加 `--rows-json` 改为结构化 rows `{row_number, values:{列字母→值}}`）；大表请按 `--range` 行窗口分批读（截断时看 `has_more`） |
+| 快速查看纯值数据、批量处理 | `+csv-get` | 对话上下文 | 返回 CSV 文本（每行带 `[row=N]` 前缀）；大表请按 `--range` 行窗口分批读（截断时看 `has_more`） |
+| 按列类型结构化读出（喂 DataFrame / round-trip 回 `+table-put`） | `+table-get` | 对话上下文 | 返回 typed 协议（`columns:[列名]` + `data` + `dtypes`/`formats`），输出形状对齐 pandas split；可一行 `pd.DataFrame(sheet["data"], columns=sheet["columns"]).astype(sheet["dtypes"])` 还原 DataFrame，或直接 round-trip 回 `+table-put` |
 | 查看公式、样式、批注、数据验证 | `+cells-get` | 对话上下文 | 返回单元格完整信息，token 开销较大 |
 | 查看某区域的下拉框（数据验证）选项 | `+dropdown-get` | 对话上下文 | 返回该 A1 范围已配置的下拉列表选项 |
 
 **选择原则**：
 - 只看值或做数据处理 → `+csv-get`；大表分批读取，避免一次拉全表撑爆上下文
-- 要结构化、按 `row_number` / 列字母定位的输出 → `+csv-get --rows-json`（默认 CSV 串更省 token，超大表批量仍用默认）
+- 要按列类型结构化读出（喂 DataFrame / round-trip 回 `+table-put`）→ `+table-get`
 - 需要公式/样式/批注 → `+cells-get`
 - 只想知道某区域下拉框有哪些选项 → `+dropdown-get`
 
@@ -40,7 +43,7 @@
 注意：
 
 - `+csv-get` 和 `+cells-get` 支持分页/截断，注意检查 `has_more` / `truncated` 标志；使用 `+cells-get` 时，在读取 `cells` 之前还必须先看 `warning_message`，并用每个 range 的 `actual_range` / `row_indices` / `col_indices` 判断真实位置
-- 隐藏行列默认包含在返回结果中（`--skip-hidden=false`），如需只看可见数据设为 `true`
+- 隐藏行列默认包含在返回结果中（`--skip-hidden=false`），如需只看可见数据设为 `true`。读取原语本身不标注哪些行列被隐藏：若要识别隐藏区间（以决定是否过滤、或如何解读混入的隐藏数据），用 `+sheet-info --include hidden_rows,hidden_cols` 取隐藏行列集合，再结合 `+csv-get` / `+cells-get` 返回的 `row_indices` / `col_indices` 判断每行 / 每列是否隐藏
 
 **常见配置错误（必须注意）**：
 - **全量读取导致上下文溢出（高频致命错误）**：不要对大表（数百行以上）直接用 `+csv-get` 或 `+cells-get` 读取全部数据到上下文。大表场景必须分批读取：用 `--range` 切行窗口逐块读（`+csv-get` / `+cells-get` 单次返回量由 `--max-chars` 自动兜底，截断时返回 `has_more`）；过大时考虑导出到本地文件后用脚本处理再分批回写
@@ -83,6 +86,7 @@
 | `+cells-get` | read | 单元格 |
 | `+dropdown-get` | read | 对象 |
 | `+csv-get` | read | 单元格 |
+| `+table-get` | read | 单元格 |
 
 ## Flags
 
@@ -115,7 +119,18 @@ _公共四件套 · 系统：`--dry-run`_
 | `--max-chars` | int | optional | 防爆，默认 200000（隐藏 flag：不在 `--help` 列出，但可正常传入） |
 | `--include-row-prefix` | bool | optional | 是否在每行前加 `[row=N]` 前缀，默认 `true` |
 | `--skip-hidden` | bool | optional | 跳过隐藏行列，默认 `false` |
-| `--rows-json` | bool | optional | 返回结构化 rows（`{row_number, values:{列字母→值}}`）而非 CSV 文本，默认 `false` |
+
+### `+table-get`
+
+_公共：URL/token（无 sheet 定位） · 系统：`--dry-run`_
+
+| Flag | Type | 必填 | 说明 |
+| --- | --- | --- | --- |
+| `--sheet-id` | string | optional | 只读该子表（按 id）；省略则读所有子表 |
+| `--sheet-name` | string | optional | 只读该子表（按名）；省略则读所有子表 |
+| `--range` | string | optional | 读取的 A1 范围；省略则读每个子表的当前数据区 |
+| `--no-header` | bool | optional | 把第一行当数据而非表头（列名取 col1/col2 …） |
+| `--dataframe-out` | string | optional | 以一份 Arrow IPC 文件（Feather v2）格式输出 typed 表格，替代默认的 JSON 输出。用 `@<path>` 传文件或 `-` 写二进制 stdout（同其他 binary I/O flag 的约定）。是 `+table-put` / `+workbook-create` 入口 `--dataframe` 的镜像 —— pandas 端 `pd.read_feather("x.arrow")` 或 `pd.read_feather(io.BytesIO(stdout))` 一行读回。仅支持单 sheet：必须给 `--sheet-id` 或 `--sheet-name`；读整本 workbook 仍走默认 JSON。列类型沿用 typed 读回（string/number/date/bool）；`number_format` 以 Arrow Field metadata 保留，Arrow 文件可直接喂回 `+table-put --dataframe`。 |
 
 ## Examples
 
@@ -140,17 +155,7 @@ lark-cli sheets +csv-get --spreadsheet-token shtXXX --sheet-name "销售明细"
 - `current_region` — 自动扩展到非空连续区域的 A1 范围。它是**真实数据边界**，**优先于 `+workbook-info` 的 `row_count`**（`row_count` 是网格物理行数，常是 200 / 1000 等默认值、远大于实际数据；按它盲读会拉回大片空行）
 - `has_more` — 是否截断；截断后续读用 `--range` 接着读
 
-**加 `--rows-json`：返回结构化 rows（而非 CSV 字符串）**
-
-```bash
-lark-cli sheets +csv-get --url "https://example.feishu.cn/sheets/shtXXX" --sheet-name "Sheet1" --range "A1:G20" --rows-json
-```
-
-`--rows-json` 下的输出契约（替换 `annotated_csv` / `col_indices` / `row_indices`）：
-
-- `rows` — 数组，每元素 `{row_number, values}`。`row_number` 是真实表格行号（整数，下游需要行号的操作直接取它）；`values` 按**列字母** key（如 `values["D"]`，绝对列字母）。**所有逻辑行都在 `rows` 里**。引号内换行已解析进单元格值，无需自己按 RFC-4180 拆行。
-- `data_not_fully_read` — **仅当没读全时出现**：`{read_through_row, data_extends_through_row, unread_rows, reread_range}`。出现即表示真实数据超出本次读取范围；批量写入前必须按 `reread_range` 重读全区，否则漏行。
-- 其余字段（`current_region` / `actual_range` / `has_more`）同上。
+> 要按列类型结构化读出（喂 DataFrame、或 round-trip 回 `+table-put`）用 `+table-get`（见下）；`+csv-get` 给的是带 `[row=N]` 前缀的纯值快照，下游需要行号/列坐标时直接从前缀与 `col_indices` 取。
 
 ### `+cells-get`
 
@@ -164,6 +169,89 @@ lark-cli sheets +cells-get --url "https://example.feishu.cn/sheets/shtXXX" --she
 
 > ⚠️ 调用方在 `cells[i][j]` 中**不能**用下标推真实行列：必须读 `ranges[n].row_indices[i]` / `ranges[n].col_indices[j]`。
 
+### `+table-get`（飞书 → DataFrame，类型保真读出）
+
+`+table-put`（写入侧，见 write-cells reference）的镜像：把表格读回与 `--sheets` 完全同构的 typed 协议（`sheets[]` + `columns:[列名]` + `data:[[行]]` + `dtypes:{列名:pandas_dtype}` + `formats?:{列名:number_format}`），可直接喂回 `+table-put` 或一行还原 DataFrame。
+
+列类型从每列 `number_format` 推断（日期格式→`date`/`datetime64[ns]`、数值→`number`/`float64`、bool→`bool`），`date` 列的序列号转回 ISO `yyyy-mm-dd`——日期、数字往返不丢类型。**列类型只在该列所有非空值一致时才定（`number` / `date` / `bool`）；一列混了类型（如数字列混入「暂无」、日期列混入裸数字）会降为 `string`（dtypes 输出 `object`），让 `dtypes` 与 `data` 里每个值自洽——能 round-trip 回 `+table-put`、不让 pandas `astype` 崩。降级是无损的（脏值原样保留为文本）；若要把零星脏值转成数值列，交给调用方在 pandas 侧做（`to_numeric(errors='coerce')`），那里原始值仍在、可追溯。** 底层复用 `get_cell_ranges` / `get_range_as_csv`。默认读所有子表、第一行当表头（`--no-header` 把首行当数据、列名取 `col1` / `col2` …）。
+
+```bash
+# 默认读所有子表 → sheets[]（与 +table-put 的 --sheets 同构，可喂回或转 DataFrame）
+lark-cli sheets +table-get --url "<表URL>"
+# 可选：--sheet-name / --sheet-id 限定只读某一个子表（不给则读全部）
+lark-cli sheets +table-get --url "<表URL>" --sheet-name "销售"
+```
+
+#### 输出 → DataFrame（2 行 helper）
+
+输出形状对齐 pandas split：`columns` 是列名数组、`data` 是二维数据、`dtypes` 是 `{列名: pandas_dtype_str}` 映射。直接喂给 `pd.DataFrame(...).astype(...)` 就能一次性还原所有列类型（不必逐列 `to_datetime` / `to_numeric`），写入侧 `df_to_sheet` 的镜像 helper：
+
+```python
+import pandas as pd
+def sheet_to_df(sheet):
+    return pd.DataFrame(sheet["data"], columns=sheet["columns"]).astype(sheet["dtypes"])
+
+# 单 sheet
+df = sheet_to_df(out["data"]["sheets"][0])
+
+# 多 sheet——按名字取
+sheets = {s["name"]: sheet_to_df(s) for s in out["data"]["sheets"]}
+df_sales = sheets["销售"]
+```
+
+> 显示格式（千分位、百分比、自定义日期）在 `sheet["formats"]`，pandas 不消费；改完数据 round-trip 回去时透传给 `+table-put` 即可，飞书侧显示不变。
+
+#### `--dataframe-out`（Arrow IPC / Feather v2 二进制读出）
+
+`--dataframe-out` 是 `+table-put` 入口 `--dataframe` 的镜像：把 typed 读回直接编码成 Arrow IPC 文件，pandas 端一行 `pd.read_feather()` 读回——省掉 JSON 解析 + `astype(dtypes)`，列类型 / `number_format` 走 Arrow schema + Field metadata 保真。**仅支持单 sheet**（Arrow 文件一 schema 容器），必须给 `--sheet-id` 或 `--sheet-name`；读整本 workbook 仍走默认 JSON。
+
+```bash
+# 文件
+lark-cli sheets +table-get --url "<表URL>" --sheet-name "销售" --dataframe-out @./out.arrow
+# binary stdout（不落盘）
+lark-cli sheets +table-get --url "<表URL>" --sheet-name "销售" --dataframe-out -
+```
+
+```python
+import io, pandas as pd, subprocess
+
+# 1) 文件
+subprocess.run(["lark-cli","sheets","+table-get","--url",URL,
+                "--sheet-name","销售","--dataframe-out","@./out.arrow"], check=True)
+df = pd.read_feather("./out.arrow")
+
+# 2) stdin/stdout 管道（不落盘）—— 跟 --dataframe 写入侧对称的一行
+res = subprocess.run(["lark-cli","sheets","+table-get","--url",URL,
+                      "--sheet-name","销售","--dataframe-out","-"],
+                     capture_output=True, check=True)
+df = pd.read_feather(io.BytesIO(res.stdout))
+```
+
+> `number_format` 进 Arrow Field metadata（key=`number_format`），Arrow 文件可以直接喂回 `+table-put --dataframe` round-trip 写回，types / formats 一路保真。
+
+#### round-trip：读 → 改 → 写回（写读对偶）
+
+`sheet_to_df` 和 write-cells reference 里的 `df_to_sheet` 是一对镜像 helper，round-trip 三段读 / 改 / 写各一行：
+
+```python
+import json, subprocess
+# 1. 读
+out = json.loads(subprocess.check_output(
+    ["lark-cli","sheets","+table-get","--url",URL,"--sheet-name","销售"]))
+sheet = out["data"]["sheets"][0]
+df = sheet_to_df(sheet)
+
+# 2. 改（pandas 操作）
+df["营收"] = df["营收"] * 1.1
+
+# 3. 写回（formats 是飞书侧显示格式，pandas 不消费，透传保留显示）
+payload = {"sheets": [df_to_sheet(df, sheet["name"], formats=sheet.get("formats"))]}
+subprocess.run(["lark-cli","sheets","+table-put","--url",URL,"--sheets","-"],
+               input=json.dumps(payload).encode(), check=True)
+```
+
+`sheet_to_df(sheet)` 消费 `(columns, data, dtypes)`，`df_to_sheet(df, name, formats=...)` 重新生成同样三个字段——读 / 写完全对偶，只有 `formats` 需要手工透传一次。
+
 ### Validate / DryRun / Execute 约束
 
 - `Validate` 阶段只做 XOR 检查、Enum 合法性、防爆参数上限校验；**禁止**联网（如不能用 `--sheet-name` 提前去查 `sheet-id`）。

skills/lark-sheets/references/lark-sheets-sparkline.md

@@ -17,7 +17,7 @@
 
 **常见配置错误（必须注意）**：
 - **数据源范围要精确**：迷你图的数据源范围必须与实际数据行列精确对应，范围偏移会导致图形展示错误
-- **不要与 SPARKLINE() 公式混淆**：飞书表格的 `SPARKLINE()` 公式函数已被禁用，迷你图只能通过本 Skill 的对象方式创建
+- **不要与 SPARKLINE() 公式混淆**：飞书表格的 `SPARKLINE()` 公式函数已被禁用，迷你图只能通过 `+sparkline-{create|update|delete}` 的对象方式创建
 - **创建后必须验证**：调用 `+sparkline-list` 确认迷你图配置正确
 
 ## Shortcuts