feat(skills): 同步 chat/bot skill 参考 — 补齐消息管理与机器人新能力 (#148 )

对齐 dws-service-endpoints.json 新增的 16 个 group-chat toolOverrides 和 2 个 bot toolOverrides，参考 dws-wukong 实现补齐以下命令文档： - message: send/list/list-all/list-topic-replies/list-by-sender/ list-mentions/list-focused/list-unread-conversations/ search/info/send-personal/list-top-conversations - group: create-org/search-common - bot: create/search-groups 同步更新命令总览、意图判断、工作流与上下文传递表。
fix(auth): restore device_code polling when flowId is absent (#145 )
2026-04-22 16:31:56 +08:00 · 2026-04-22 15:13:40 +08:00 · 2026-04-22 14:56:11 +08:00 · 2026-04-22 12:19:22 +08:00 · 2026-04-22 12:05:30 +08:00
14 changed files with 1364 additions and 407 deletions
@@ -1 +1 @@
-<svg xmlns="http://www.w3.org/2000/svg" width="108" height="20" role="img" aria-label="coverage: 48.7%"><title>coverage: 48.7%</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="108" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="61" height="20" fill="#555"/><rect x="61" width="47" height="20" fill="#e05d44"/><rect width="108" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="315" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="510">coverage</text><text x="315" y="140" transform="scale(.1)" fill="#fff" textLength="510">coverage</text><text aria-hidden="true" x="835" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="370">48.7%</text><text x="835" y="140" transform="scale(.1)" fill="#fff" textLength="370">48.7%</text></g></svg>
+<svg xmlns="http://www.w3.org/2000/svg" width="108" height="20" role="img" aria-label="coverage: 49.8%"><title>coverage: 49.8%</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="108" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="61" height="20" fill="#555"/><rect x="61" width="47" height="20" fill="#e05d44"/><rect width="108" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="315" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="510">coverage</text><text x="315" y="140" transform="scale(.1)" fill="#fff" textLength="510">coverage</text><text aria-hidden="true" x="835" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="370">49.8%</text><text x="835" y="140" transform="scale(.1)" fill="#fff" textLength="370">49.8%</text></g></svg>
@@ -4,6 +4,260 @@ All notable changes to this project will be documented in this file.

 The format is inspired by [Keep a Changelog](https://keepachangelog.com/) and this project follows [Semantic Versioning](https://semver.org/).

+## [1.0.12] - 2026-04-21
+
+Product-surface expansion: first-class `doc` (DingTalk Docs) and `minutes` (AI Minutes) skill references, refreshed `aitable` guide aligned with the shipped binary (including dashboard / chart / export), and a README sync that brings the full command catalog to **141 commands across 14 products**.
+
+### Added
+
+- **`doc` skill reference** (`skills/references/products/doc.md`) — 16-command coverage of DingTalk Docs:
+  - Discovery: `search`, `list`, `info`, `read`
+  - Authoring: `create`, `update`, `folder create`
+  - Files: `upload`, `download`
+  - Block-level editing: block `query`, `insert`, `update`, `delete`
+  - Comments: `comment list`, `create`, `reply`
+  - URL → `doc_id` extraction rules and nodeId dual-format notes
+- **`minutes` skill reference** (`skills/references/products/minutes.md`) — coverage of AI Minutes:
+  - Lists: personal / shared-with-me / all-accessible
+  - Content: basic info, AI summary, keywords, transcription, extracted todos, batch detail
+  - Editing: title update
+  - Recording control: start, pause, resume, stop
+- **SKILL.md routing**:
+  - Product overview table rows for `doc` and `minutes`
+  - Intent decision tree routes — `钉钉文档/云文档/知识库/块级编辑/文档评论` → `doc`; `听记/AI听记/会议纪要/转写/摘要/思维导图/发言人/热词` → `minutes`
+  - Danger-op table entries: `doc delete`, `doc block delete`
+  - `aitable` description completed with the `附件` (attachment) group
+- **`aitable` skill enhancements**:
+  - `field create` single-field mode (`--name` / `--type` / `--config`) with examples
+  - `base get` URL → `baseId` quick-tip
+  - Dedicated "URL → baseId 提取" chapter
+  - "`--filters` 筛选语法排错与使用规范" chapter
+  - "相关产品" cross-link section pointing to `doc`
+  - **"复杂操作" chapter** (#141) — dashboard / chart workflow (with two-call sequencing and `chart share get` vs `dashboard share get` error semantics) and two-stage `export data` polling (`scope=all/table/view` parameter constraints)
+- **README Key Services sync** (#140):
+  - New rows: `doc` (16 commands), `minutes` (22 commands — adds `hot-word`, `mind-graph`, `replace-text`, `speaker`, `upload` subgroups)
+  - `aitable` expanded from 20 → 37 commands; surfaces `chart`, `dashboard`, `export`, `import`, `view` subgroups
+  - Total command count updated from **86 → 141 across 14 products**
+  - "Coming soon" list drops `doc` and `minutes`
+
+### Changed
+
+- `aitable record query` docs rename `--keyword` → `--query` to match the shipped binary
+- `aitable record query` docs clarify `--sort` direction semantics (avoids misuse of `order`)
+- `aitable base list` guidance strengthened — "only for recent browsing; use `base search` for lookups"; intent decision prioritizes `base search` for base discovery
+
+## [1.0.11] - 2026-04-20
+
+Plugin subsystem hardening: faster cold startup, cleaner lifecycle, stricter isolation, and polished UX for PAT / i18n / error routing.
+
+### Added
+
+- `feat: supports claw-like products` — overlay path for Claw-style embedded editions
+- `feat(plugin): inject user identity (UserID, CorpID) into stdio plugin subprocesses`
+- `feat(auth): improve login UX for terminal auth denial cases` — clearer messaging + retry affordance
+- `feat: PAT scope error visualization and auto-retry with authorization polling` (#113)
+  - Human-readable error output (lark-cli style) with type/message/hint/authorization command
+  - JSON payload also available via `--format json`
+  - Auto-retry once the user completes scope authorization
+
+### Changed
+
+- `perf(plugin): serve plugin MCP tool list from disk cache on startup` — hot path skips Initialize+ListTools when snapshot exists
+- `perf(plugin): parallelize all plugin discovery and tighten cold timeouts` — HTTP cold budget 4s → 700ms (auth) / 500ms (plain); stdio and HTTP fan out concurrently
+- `perf(plugin): share cache.Store across discovery` — single `*cache.Store` above the fan-out instead of per-goroutine instances
+- `refactor(plugin): remove default/managed plugin privileged mechanism` (#124) — third-party plugins install on an equal footing via `dws plugin install`
+- `refactor(plugin): purge removed plugin settings instead of merely disabling` — `RemovePlugin` now deletes `EnabledPlugins` and `PluginConfigs` entries
+
+### Fixed
+
+- `fix(transport): cap plugin MCP startup at ~4s when endpoints are unreachable` (#119) — eliminates the 10s `dws --help` stall caused by compounding transport timeouts
+- `fix(plugin): stop stdio child processes on exit and before removal` — no more orphaned plugin subprocesses
+- `fix(pat): avoid shared PAT command state in root registration` (#129)
+- `fix: -f json 模式下错误 JSON 从 stdout 改为输出到 stderr` (#133) — restores CI stderr-based failure assertions
+- `fix(cli): localize plugin/help command strings via i18n` (#118, #134) — zh locale now shows consistent Chinese `--help`; wraps plugin module, help command, and OAuth client-id/secret flag descriptions
+- `chore: remove workspace and bundled artifacts` (#127) — clean local-only repository leftovers
+
+## [1.0.9] - 2026-04-16
+
+Plugin system launch + execution-pipeline overhaul. This is the largest release since 1.0.0: third-party MCP servers become first-class commands, the command pipeline grows to five stages, and the edition overlay gains the hooks needed for embedded hosts.
+
+### Added
+
+#### Plugin system (new)
+
+- `plugin` command family: `install`, `list`, `info`, `enable`, `disable`, `remove`, `create`, `dev`, `config set/get/list/unset`
+- Plugin manifest parsing/validation, managed/user directory-based identity
+- MCP server conversion and injection into the dynamic routing registry
+- Pipeline hook adapter for shell-based hooks
+- Stdio transport: subprocess lifecycle, `DWS_PLUGIN_ROOT` / `DWS_PLUGIN_DATA` variable expansion
+- Stdio server tools automatically registered as CLI subcommands (e.g. `dws hello greet --name Peter`)
+- Streamable-HTTP MCP tool discovery via `registerHTTPServer`
+- Updater: managed plugin update check on CLI startup (10 s timeout, best-effort)
+- `dws plugin create` scaffold (plugin.json, SKILL.md, hooks.json); `dws plugin dev` source-dir registration without copy
+- `SyncSkills` — copies plugin skills to agent directories on startup
+- **Auth Token Registry**: per-server HTTP headers declared in `plugin.json` for third-party MCP servers (e.g. Alibaba Cloud Bailian) independent from DingTalk OAuth
+- **Persistent plugin config** (`dws plugin config ...`): values persisted to `~/.dws/settings.json`, auto-injected as env vars; `${KEY}` in `plugin.json` resolves without manual `export`
+- **Build lifecycle**: `build` field compiles stdio servers to native binaries at install time
+- **Command-name conflict protection**: reserved built-in names (`auth`, `plugin`, `cache`, …) and plugin-vs-plugin duplicate detection
+- Parallel service discovery (`sync.WaitGroup`) — startup reduced from sequential `N*10s` to parallel `max(10s)`
+
+#### Core commands & diagnostics
+
+- `dws doctor` — one-stop environment/auth/network diagnostics
+- `dws config list` — centralized view of scattered configuration
+- Structured perf tracing (upgraded from debug tool to diagnostics output)
+- `feat(skill): restore find/get for legacy skill market API` — `skill find`, `skill get`; `skill add` still uses aihub download
+
+#### Edition / overlay hooks
+
+- `edition.Hooks.SaveToken` / `LoadToken` / `DeleteToken` — delegate token persistence with keychain fallback
+- `edition.Hooks.AuthClientID` / `AuthClientFromMCP` — overlay can override the OAuth client ID and route auth through MCP endpoints
+- `edition.Hooks.AfterPersistentPreRun` — wire non-MCP clients (e.g. A2A gateway) after root setup
+- `edition.Hooks.ClassifyToolResult` — custom MCP result classification before the default business-error detection
+- Token marker file (`token.json`) for embedded hosts to detect auth state without keychain access
+- `pkg/runtimetoken.ResolveAccessToken` mirroring MCP auth resolution; MCP identity headers exported via `pkg/cli` for auxiliary HTTP transports
+- `ExitCoder` interface — edition-specific errors carry custom exit codes
+- `RawStderrError` interface — errors that bypass CLI formatting and emit raw stderr (for desktop runtimes)
+
+### Changed
+
+- **Command execution pipeline: 3 → 5 stages** (`Register → PreParse → PostParse → PreRequest → PostResponse`)
+- `feat(schema): return structured degraded errors instead of silent empty catalog` — new `CatalogDegraded` error with reasons `unauthenticated` / `market_unreachable` / `runtime_all_failed`; auth pre-check short-circuits doomed MCP connections
+- `refactor(auth): unify auxiliary token resolution with MCP cached path` — shared `resolveAccessTokenFromDir`; overlays reuse the process-level token cache
+- `feat(plugin): improve CLI overlay resolution and plugin install robustness`
+  - `plugin.json` `cli` field now accepts a file path (e.g. `"cli": "overlay.json"`) in addition to inline JSON
+  - `description` field on `CLIToolOverride` for static fallback when MCP `tools/list` is unavailable
+  - Windows install uses `cmd /C` instead of `sh -c` for build commands
+
+### Fixed
+
+- `fix(plugin): harden plugin system security boundaries`
+  - Reject `file://` / local paths in git URLs; allow only `https` / `ssh`
+  - Reject symlink entries during ZIP extraction (path-traversal defense)
+  - `build.output` must be a relative path within the plugin directory
+  - Reject absolute paths in stdio command declarations
+  - Block dangerous env var names (`PATH`, `LD_PRELOAD`, …) from plugin config injection
+- `fix(plugin): schema flag params, HTTP tool discovery, and integration tests`
+- `fix(plugin): skip min version check in dev mode`
+
+## [1.0.8] - 2026-04-07
+
+AITable command surface expansion, installer alignment with npm conventions, and execution-timeout hardening.
+
+### Added
+
+- **AITable static helper commands** (20 commands in total) replacing dynamic routing:
+  - `base`: `list`, `search`, `get`, `create`, `update`
+  - `table`: `get`, `create`, `update`
+  - `field`: `get`, `create`, `update`
+  - `record`: `query`, `create`, `update`
+  - `template`: `search`
+  - `attachment`: `upload`
+- `feat(install): align skill dirs with npm and add OpenClaw` — skill install paths follow npm conventions; OpenClaw added to supported agents
+- Label rendering optimization for AITable records (`to #73551688`)
+- README: npm install method documented
+- README: note that `dws upgrade` requires v1.0.7+
+
+### Changed
+
+- `perf: optimize command timeout handling, instrumentation, and diagnostics`
+
+## [1.0.7] - 2026-04-02
+
+Self-upgrade, edition overlay foundation, and fail-closed auth enforcement.
+
+### Added
+
+- **`dws upgrade`** — self-upgrade via GitHub Releases; atomic replace; cross-platform (macOS/Linux/Windows)
+- `feat: edition layer for Wukong overlay` — build-time edition hook lets downstream overlays customize auth UX, config dir, static server list, visible products, and extra root commands
+  - `pkg/edition` defaults + `pkg/editiontest` contract tests
+  - `Makefile` target `edition-test`; CI job `edition-tests`
+  - Static server injection skips market discovery when configured
+  - Deduplicates top-level commands so overlay wins
+  - `hideNonDirectRuntimeCommands` respects edition `VisibleProducts`
+  - Gated `auth login` subcommand + hints for embedded editions
+  - Optional token auto-purge; edition `ConfigDir` override
+- `dws version` — human-readable multi-line output plus JSON with edition, architecture, build, commit
+- Tag reporting for case suites (`to #73551688`)
+- `feat(auth): unify MCP retry constant and add retry to remaining endpoints`
+
+### Changed
+
+- `style(auth): redesign OAuth authorization pages UI`
+
+### Fixed
+
+- `fix(auth): switch CLI auth check from fail-open to fail-closed`
+  - When `/cli/cliAuthEnabled` is unreachable (network error/timeout/5xx), OAuth callback now routes to the permission request page instead of silently marking "enabled"
+  - Device Flow blocks login and asks the user to verify network connectivity
+  - `CheckCLIAuthEnabled` retries with backoff (3 attempts, 0s/1s/2s) to tolerate transient issues
+
+## [1.0.6] - 2026-04-01
+
+Error diagnostics overhaul, destructive-command confirmation, and credential auto-persistence.
+
+### Added
+
+- **Interactive confirmation for destructive dynamic commands** — prompts before delete/remove operations unless `--yes` is set
+- **Enhanced error diagnostics**
+  - `ServerDiagnostics` struct extracts `trace_id`, `server_error_code`, `technical_detail`, `server_retryable` from MCP responses
+  - Pulls diagnostics from JSON-RPC `error.data`, tool call result content, and HTTP headers (`X-Trace-Id`, `X-Request-Id`, `x-dingtalk-trace-id`)
+  - Three verbosity levels for `PrintHuman`: Normal (trace ID + server code), Verbose (+ technical detail), Debug (+ RPC code / operation / reason)
+  - Local logging now includes sanitized request body, response body on error, retry attempts, and classification events
+  - `TruncateBody` / `SanitizeArguments` / `RedactHeaders` helpers with sensitive-key substring detection
+- **Auth credential persistence**
+  - `feat(auth): enhance device flow with CLI auth check and admin guidance`
+  - `feat(auth): persist OAuth credentials for reliable token refresh`
+  - `feat(auth): persist client credentials and optimize keychain access` — auto-persist `--client-id` / `--client-secret`; keychain credential cache to avoid repeated reads; enhanced logout cleans `app.json` + keychain secrets + `token.json`
+- `add report helper with flexible date parsing and defaults`
+- `feat: to #73551688 支持消息通知`
+- README: Official App mode (recommended, direct login without creating an app) + Custom App mode; admin guide for enabling CLI access
+
+### Changed
+
+- Getting Started simplified with inline login commands; whitelist references removed from the IMPORTANT banner
+- Version bump documentation updated to v1.0.5 internal; co-creation group QR code refreshed
+
+### Fixed
+
+- `fix: resolve verbosity flag lookup, FileLogger lazy binding, and business error logging`
+  - `resolveVerbosity` uses `cmd.Flags()` instead of `PersistentFlags()` so subcommands inherit `--verbose` / `--debug`
+  - `FileLogger` lazy-binds in `executeInvocation` (after `configureLogLevel` init)
+  - Business errors (HTTP 200 + `success=false`) now written to the file logger for offline diagnosis
+- OAuth callback race condition (write response before sending code)
+- `import path for errors package in skill_command.go`
+
+## [1.0.4] - 2026-03-30
+
+Token-refresh reliability and onboarding clarity.
+
+### Added
+
+- `feat(auth): persist client credentials for token refresh` — `--client-id` / `--client-secret` are stored for automatic refresh after expiration; client secret lives in the system Keychain with a file reference
+- README onboarding flow rewrite with step-by-step first-time setup and more realistic examples
+- Agent skill reference polish: clearer examples, updated intent routing patterns, expanded `simple.md` onboarding, cross-skill reference fixes
+
+## [1.0.3] - 2026-03-29
+
+Filtering power, schema rendering, and a native `todo` command family.
+
+### Added
+
+- **Nested / array-indexed output filtering**
+  - `--fields` now accepts dot-notation (e.g. `--fields response.content`) and array index access (e.g. `response.items[0]`)
+  - New field-path parser with recursive extraction logic
+- **`schema` command enhancements**
+  - Table format output for human consumption
+  - Product-level endpoint loading in the CLI loader
+  - Schema-text rendering wired into the runner output pipeline
+- **`todo` task helper family** — static `create` / `update` / `done` / `get` / `delete` with `preferLegacyLeaf` replacing dynamic commands
+  - MCP tool alignment: `create_personal_todo`, `update_todo_task`, `update_todo_done_status`, `query_todo_detail`, `delete_todo`
+  - ISO-8601 due-time parsing
+  - Hidden title aliases and delete confirmation
+  - Priority field on `todo` helper
+  - Expanded zh / en i18n coverage (fixes `en.json` spacing/wording issues)
+- README restructured with collapsible feature sections
+
 ## [1.0.2] - 2026-03-29

 Deep workspace tooling upgrade: pipeline-based input correction, output filtering, enhanced stdin handling, and multi-endpoint routing.
@@ -284,27 +284,6 @@ func retryWithPatAuthRetry(ctx context.Context, runner executor.Runner, invocati
 	return runner.Run(ctx, invocation)
 }

-// loadMCPClientIDIfNeeded ensures we have a client ID for device flow.
-// Priority: in-memory runtime value → DWS_CLIENT_ID env → MCP remote fetch.
-func loadMCPClientIDIfNeeded(ctx context.Context, configDir string) string {
-	clientID := authpkg.ClientID()
-	if clientID != "" {
-		return clientID
-	}
-	// Fallback: read from environment variable (set by previous PAT auth or caller).
-	if envID := os.Getenv("DWS_CLIENT_ID"); envID != "" {
-		authpkg.SetClientIDFromMCP(envID)
-		return envID
-	}
-	// Last resort: fetch from MCP server.
-	mcpClientID, err := authpkg.FetchClientIDFromMCP(ctx)
-	if err == nil && mcpClientID != "" {
-		authpkg.SetClientIDFromMCP(mcpClientID)
-		return mcpClientID
-	}
-	return ""
-}
-
 // ---- handlePatAuthCheck (runner.go entry point) -----------------------------

 const (
@@ -1091,18 +1091,13 @@ func loadPlugins(engine *pipeline.Engine, runner executor.Runner) []*cobra.Comma
 		}
 	}

-	// 1. Load plugins from the legacy managed/ directory (backward compat
-	//    for plugins installed by older CLI builds).
-	managedPlugins := pluginLoader.LoadManaged()
-
-	// 2. Load user plugins (per settings.json)
+	// 1. Load user plugins (per settings.json)
 	userPlugins := pluginLoader.LoadUser()

-	// 3. Load dev plugins (registered via `dws plugin dev`)
+	// 2. Load dev plugins (registered via `dws plugin dev`)
 	devPlugins := pluginLoader.LoadDev()

-	allPlugins := append(managedPlugins, userPlugins...)
-	allPlugins = append(allPlugins, devPlugins...)
+	allPlugins := append(userPlugins, devPlugins...)

 	// 3. Discover tools from streamable-http servers and build CLI commands.
 	//    Third-party servers with auth headers are discovered in parallel
@@ -1213,7 +1208,6 @@ func loadPlugins(engine *pipeline.Engine, runner executor.Runner) []*cobra.Comma

 	if len(allPlugins) > 0 {
 		slog.Debug("plugins loaded",
-			"managed", len(managedPlugins),
 			"user", len(userPlugins),
 			"dev", len(devPlugins),
 		)
@@ -1529,17 +1523,12 @@ func buildStdioCommands(p *plugin.Plugin, sc plugin.StdioServerClient, tools []t
 	// Construct virtual endpoint and server descriptor.
 	endpoint := StdioEndpoint(p.Manifest.Name, sc.Key)

-	source := "plugin"
-	if p.IsManaged {
-		source = "plugin-managed"
-	}
-
 	descriptor := market.ServerDescriptor{
 		Key:         sc.Key,
 		DisplayName: p.Manifest.Name + "/" + sc.Key,
 		Description: p.Manifest.Description,
 		Endpoint:    endpoint,
-		Source:      source,
+		Source:      "plugin",
 		CLI:         overlay,
 		HasCLIMeta:  true,
 	}
@@ -0,0 +1,144 @@
+// Copyright 2026 Alibaba Group
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package auth
+
+import "testing"
+
+func TestClassifyDenialReason(t *testing.T) {
+	cases := []struct {
+		name           string
+		status         *CLIAuthStatus
+		currentChannel string
+		want           string
+	}{
+		{
+			name: "error CHANNEL_REQUIRED",
+			status: &CLIAuthStatus{
+				ErrorCode: "CHANNEL_REQUIRED",
+			},
+			want: "channel_required",
+		},
+		{
+			name: "error NO_AUTH",
+			status: &CLIAuthStatus{
+				ErrorCode: "NO_AUTH",
+			},
+			want: "no_auth",
+		},
+		{
+			name: "success false or nil result → unknown",
+			status: &CLIAuthStatus{
+				Success: false,
+			},
+			want: "unknown",
+		},
+		{
+			name: "cliAuthEnabled true → no denial",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result:  &CLIAuthResult{CLIAuthEnabled: true},
+			},
+			want: "",
+		},
+		{
+			name: "userScope forbidden wins over channel",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result: &CLIAuthResult{
+					CLIAuthEnabled:  false,
+					UserScope:       "forbidden",
+					ChannelScope:    "specified",
+					AllowedChannels: []string{"channel-a"},
+				},
+			},
+			currentChannel: "channel-b",
+			want:           "user_forbidden",
+		},
+		{
+			// Real-world case reported: user is in allowedUsers but the current
+			// DWS_CHANNEL is not in allowedChannels. Reason must be channel,
+			// NOT user.
+			name: "user allowed but channel not in allowedChannels → channel_not_allowed",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result: &CLIAuthResult{
+					CLIAuthEnabled:  false,
+					UserScope:       "specified",
+					AllowedUsers:    []string{"014566033934857460"},
+					ChannelScope:    "specified",
+					AllowedChannels: []string{"2a4a658e467998befb7fa333c19ba2b3a3bacfa4"},
+				},
+			},
+			currentChannel: "different-channel",
+			want:           "channel_not_allowed",
+		},
+		{
+			name: "channelScope specified but current channel empty → channel_required",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result: &CLIAuthResult{
+					CLIAuthEnabled:  false,
+					UserScope:       "specified",
+					ChannelScope:    "specified",
+					AllowedChannels: []string{"channel-a"},
+				},
+			},
+			currentChannel: "",
+			want:           "channel_required",
+		},
+		{
+			name: "channel matches allowedChannels → fall back to user denial",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result: &CLIAuthResult{
+					CLIAuthEnabled:  false,
+					UserScope:       "specified",
+					ChannelScope:    "specified",
+					AllowedChannels: []string{"channel-a"},
+				},
+			},
+			currentChannel: "channel-a",
+			want:           "user_not_allowed",
+		},
+		{
+			name: "only userScope=specified, no channel restriction → user_not_allowed",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result: &CLIAuthResult{
+					CLIAuthEnabled: false,
+					UserScope:      "specified",
+				},
+			},
+			currentChannel: "",
+			want:           "user_not_allowed",
+		},
+		{
+			name: "no user or channel restriction → cli_not_enabled",
+			status: &CLIAuthStatus{
+				Success: true,
+				Result:  &CLIAuthResult{CLIAuthEnabled: false},
+			},
+			want: "cli_not_enabled",
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			got := classifyDenialReason(tc.status, tc.currentChannel)
+			if got != tc.want {
+				t.Fatalf("classifyDenialReason() = %q, want %q", got, tc.want)
+			}
+		})
+	}
+}
@@ -185,11 +185,6 @@ func (p *DeviceFlowProvider) loginOnce(ctx context.Context, attempt int) (*Token
 	if err != nil {
 		return nil, err
 	}
-	if tokenResult == nil {
-		// FlowID was empty — no polling happened; authorization URL was already
-		// printed, so the user can handle it manually.
-		return nil, nil
-	}

 	_, _ = fmt.Fprintln(p.output(), "")
 	dfPrintStep(p.output(), 3, i18n.T("使用授权码换取 Access Token..."), 0)
@@ -377,15 +372,15 @@ func (p *DeviceFlowProvider) pollDeviceStatus(ctx context.Context, flowID string
 }

 func (p *DeviceFlowProvider) waitForAuthorization(ctx context.Context, auth *DeviceAuthResponse) (*DeviceTokenResponse, error) {
-	// No FlowID from server — cannot poll status; return immediately so the
-	// user can still see the authorization URL printed earlier and handle it
-	// manually (same pattern as pat_auth_retry.go L451).
 	if auth.FlowID == "" {
-		dfPrintDim(p.output(), i18n.T("  服务端未返回 flowId，跳过轮询，请在浏览器中手动完成授权后重试"))
-		_, _ = fmt.Fprintln(p.output(), "")
-		return nil, nil
+		// Keep the pre-flowId device-code polling path for regular device flow
+		// login responses that do not include terminal polling metadata.
+		return p.waitForAuthorizationByDeviceCode(ctx, auth)
 	}
+	return p.waitForAuthorizationByFlowID(ctx, auth)
+}

+func (p *DeviceFlowProvider) waitForAuthorizationByFlowID(ctx context.Context, auth *DeviceAuthResponse) (*DeviceTokenResponse, error) {
 	startTime := time.Now()
 	interval := time.Duration(auth.Interval) * time.Second
 	deadline := time.Duration(auth.ExpiresIn) * time.Second
@@ -435,6 +430,63 @@ func (p *DeviceFlowProvider) waitForAuthorization(ctx context.Context, auth *Dev
 	}
 }

+func (p *DeviceFlowProvider) waitForAuthorizationByDeviceCode(ctx context.Context, auth *DeviceAuthResponse) (*DeviceTokenResponse, error) {
+	startTime := time.Now()
+	interval := time.Duration(auth.Interval) * time.Second
+	deadline := time.Duration(auth.ExpiresIn) * time.Second
+	pollCount := 0
+
+	for {
+		elapsed := time.Since(startTime)
+		if elapsed >= maxPollTotalWait || elapsed >= deadline {
+			_, _ = fmt.Fprintln(p.output(), "")
+			return nil, fmt.Errorf("%s", i18n.Tf("设备授权码已过期（%d 秒），请重试", auth.ExpiresIn))
+		}
+
+		select {
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		case <-time.After(interval):
+		}
+
+		pollCount++
+		elapsedSec := int(time.Since(startTime).Seconds())
+		dfPrintPollStatus(p.output(), pollCount, elapsedSec)
+
+		resp, err := p.pollDeviceToken(ctx, auth.DeviceCode)
+		if err != nil {
+			dfPrintPollResult(p.output(), "network_error", i18n.T("网络错误，继续重试..."))
+			if p.logger != nil {
+				p.logger.Debug("poll error", "error", err)
+			}
+			continue
+		}
+
+		if resp.Error == "" {
+			dfPrintPollResult(p.output(), "authorized", i18n.T("授权成功!"))
+			return resp, nil
+		}
+		switch resp.Error {
+		case "authorization_pending":
+			dfPrintPollResult(p.output(), "pending", i18n.T("等待用户授权..."))
+		case "slow_down":
+			interval += 5 * time.Second
+			if interval > maxPollInterval*time.Second {
+				interval = maxPollInterval * time.Second
+			}
+			dfPrintPollResult(p.output(), "slow_down", fmt.Sprintf(i18n.T("轮询过快，间隔增加至 %ds"), int(interval.Seconds())))
+		case "access_denied":
+			_, _ = fmt.Fprintln(p.output(), "")
+			return nil, errors.New(i18n.T("用户拒绝了授权请求"))
+		case "expired_token":
+			_, _ = fmt.Fprintln(p.output(), "")
+			return nil, errors.New(i18n.T("设备授权码已过期"))
+		default:
+			dfPrintPollResult(p.output(), "unknown", fmt.Sprintf(i18n.T("未知错误: %s"), resp.Error))
+		}
+	}
+}
+
 func (p *DeviceFlowProvider) postForm(ctx context.Context, endpoint string, params url.Values) ([]byte, error) {
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, strings.NewReader(params.Encode()))
 	if err != nil {
@@ -14,8 +14,11 @@
 package auth

 import (
+	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
+	"fmt"
 	"io"
 	"log/slog"
 	"net/http"
@@ -24,6 +27,8 @@ import (
 	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DingTalk-Real-AI/dingtalk-workspace-cli/internal/i18n"
 )

 func newDeviceFlowTestLogger() *slog.Logger {
@@ -138,6 +143,55 @@ func TestWaitForAuthorizationSucceedsAfterPending(t *testing.T) {
 	}
 }

+func TestWaitForAuthorizationFallsBackToDeviceCodeWhenFlowIDMissing(t *testing.T) {
+	t.Parallel()
+
+	var calls atomic.Int32
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.Method != http.MethodPost {
+			t.Fatalf("method = %s, want POST", r.Method)
+		}
+		if err := r.ParseForm(); err != nil {
+			t.Fatalf("ParseForm() error = %v", err)
+		}
+		if got := r.FormValue("device_code"); got != "legacy-device-code" {
+			t.Fatalf("device_code = %q, want legacy-device-code", got)
+		}
+		if calls.Add(1) <= 2 {
+			writeServiceResult(w, true, DeviceTokenResponse{Error: "authorization_pending"}, "", "")
+			return
+		}
+		writeServiceResult(w, true, DeviceTokenResponse{AuthCode: "legacy-auth-code"}, "", "")
+	}))
+	defer server.Close()
+
+	provider := NewDeviceFlowProvider(t.TempDir(), newDeviceFlowTestLogger())
+	var output bytes.Buffer
+	provider.Output = &output
+	provider.SetBaseURL(server.URL)
+
+	resp, err := provider.waitForAuthorization(context.Background(), &DeviceAuthResponse{
+		DeviceCode: "legacy-device-code",
+		ExpiresIn:  10,
+		Interval:   1,
+	})
+	if err != nil {
+		t.Fatalf("waitForAuthorization() error = %v", err)
+	}
+	if resp.AuthCode != "legacy-auth-code" {
+		t.Fatalf("auth code = %q, want legacy-auth-code", resp.AuthCode)
+	}
+	if calls.Load() != 3 {
+		t.Fatalf("poll calls = %d, want 3", calls.Load())
+	}
+	if !strings.Contains(output.String(), i18n.T("等待用户授权...")) {
+		t.Fatalf("expected device-code path to emit pending output, got %q", output.String())
+	}
+	if !strings.Contains(output.String(), i18n.T("授权成功!")) {
+		t.Fatalf("expected device-code path to emit success output, got %q", output.String())
+	}
+}
+
 func TestWaitForAuthorizationHonorsContextCancellation(t *testing.T) {
 	t.Parallel()

@@ -166,3 +220,105 @@ func TestWaitForAuthorizationHonorsContextCancellation(t *testing.T) {
 		t.Fatal("waitForAuthorization() error = nil, want context cancellation")
 	}
 }
+
+func TestWaitForAuthorizationByDeviceCodeHonorsContextCancellation(t *testing.T) {
+	t.Parallel()
+
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		writeServiceResult(w, true, DeviceTokenResponse{Error: "authorization_pending"}, "", "")
+	}))
+	defer server.Close()
+
+	provider := NewDeviceFlowProvider(t.TempDir(), newDeviceFlowTestLogger())
+	provider.Output = io.Discard
+	provider.SetBaseURL(server.URL)
+
+	ctx, cancel := context.WithTimeout(context.Background(), 1500*time.Millisecond)
+	defer cancel()
+
+	_, err := provider.waitForAuthorization(ctx, &DeviceAuthResponse{
+		DeviceCode: "legacy-device-code",
+		ExpiresIn:  60,
+		Interval:   1,
+	})
+	if !errors.Is(err, context.DeadlineExceeded) {
+		t.Fatalf("waitForAuthorization() error = %v, want context deadline exceeded", err)
+	}
+}
+
+func TestWaitForAuthorizationByDeviceCodeErrorStates(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name       string
+		timeout    time.Duration
+		responses  []DeviceTokenResponse
+		wantErr    string
+		wantErrIs  error
+		wantOutput string
+	}{
+		{
+			name:       "slow_down_then_context_cancelled",
+			timeout:    1500 * time.Millisecond,
+			responses:  []DeviceTokenResponse{{Error: "slow_down"}},
+			wantErrIs:  context.DeadlineExceeded,
+			wantOutput: fmt.Sprintf(i18n.T("轮询过快，间隔增加至 %ds"), 6),
+		},
+		{
+			name:       "access_denied",
+			timeout:    5 * time.Second,
+			responses:  []DeviceTokenResponse{{Error: "access_denied"}},
+			wantErr:    i18n.T("用户拒绝了授权请求"),
+			wantOutput: fmt.Sprintf(i18n.T("[%d] 轮询中... (%ds)"), 1, 1),
+		},
+		{
+			name:       "expired_token",
+			timeout:    5 * time.Second,
+			responses:  []DeviceTokenResponse{{Error: "expired_token"}},
+			wantErr:    i18n.T("设备授权码已过期"),
+			wantOutput: fmt.Sprintf(i18n.T("[%d] 轮询中... (%ds)"), 1, 1),
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			var calls atomic.Int32
+			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+				idx := int(calls.Add(1)) - 1
+				if idx >= len(tt.responses) {
+					idx = len(tt.responses) - 1
+				}
+				writeServiceResult(w, true, tt.responses[idx], "", "")
+			}))
+			defer server.Close()
+
+			provider := NewDeviceFlowProvider(t.TempDir(), newDeviceFlowTestLogger())
+			var output bytes.Buffer
+			provider.Output = &output
+			provider.SetBaseURL(server.URL)
+
+			ctx, cancel := context.WithTimeout(context.Background(), tt.timeout)
+			defer cancel()
+
+			_, err := provider.waitForAuthorization(ctx, &DeviceAuthResponse{
+				DeviceCode: "legacy-device-code",
+				ExpiresIn:  60,
+				Interval:   1,
+			})
+
+			if tt.wantErrIs != nil {
+				if !errors.Is(err, tt.wantErrIs) {
+					t.Fatalf("waitForAuthorization() error = %v, want %v", err, tt.wantErrIs)
+				}
+			} else if err == nil || err.Error() != tt.wantErr {
+				t.Fatalf("waitForAuthorization() error = %v, want %q", err, tt.wantErr)
+			}
+
+			if !strings.Contains(output.String(), tt.wantOutput) {
+				t.Fatalf("expected output to contain %q, got %q", tt.wantOutput, output.String())
+			}
+		})
+	}
+}
@@ -22,6 +22,7 @@ import (
 	"net/http"
 	"net/url"
 	"os"
+	"slices"
 	"time"

 	"github.com/DingTalk-Real-AI/dingtalk-workspace-cli/pkg/config"
@@ -1129,15 +1130,26 @@ type CLIAuthStatus struct {
 // The server computes cliAuthEnabled by considering the org switch, userScope,
 // and channelScope together; the CLI uses it as-is.
 type CLIAuthResult struct {
-	CLIAuthEnabled  bool     `json:"cliAuthEnabled"`
-	UserScope       string   `json:"userScope,omitempty"`       // "all" | "specified" | "forbidden"
-	AllowedUsers    []string `json:"allowedUsers,omitempty"`    // staffId list when userScope="specified"
-	ChannelScope    string   `json:"channelScope,omitempty"`    // "all" | "specified"
-	AllowedChannels []string `json:"allowedChannels,omitempty"` // channelCode list when channelScope="specified"
+	CLIAuthEnabled       bool     `json:"cliAuthEnabled"`
+	UserScope            string   `json:"userScope,omitempty"`            // "all" | "specified" | "forbidden"
+	AllowedUsers         []string `json:"allowedUsers,omitempty"`         // staffId list when userScope="specified"
+	ChannelScope         string   `json:"channelScope,omitempty"`         // "all" | "specified"
+	AllowedChannels      []string `json:"allowedChannels,omitempty"`      // channelCode list when channelScope="specified"
+	ChannelConfigEnabled bool     `json:"channelConfigEnabled,omitempty"` // whether org has any channel restriction configured
 }

 // classifyDenialReason inspects a CLIAuthStatus response and returns a machine-readable
 // denial reason string. Returns "" when access is granted.
+//
+// Priority rationale:
+//  1. Explicit org-wide ban (userScope=forbidden) always wins.
+//  2. Channel scope is evaluated BEFORE user scope because the CLI has
+//     authoritative knowledge of DWS_CHANNEL and can verify membership against
+//     allowedChannels. This avoids falsely blaming the user when the real
+//     denial cause is a channel mismatch (e.g. user is in allowedUsers but the
+//     current channel is not in allowedChannels).
+//  3. Only when the channel is unrestricted or matches do we attribute the
+//     denial to the user scope.
 func classifyDenialReason(status *CLIAuthStatus, currentChannel string) string {
 	if status.ErrorCode == "CHANNEL_REQUIRED" {
 		return "channel_required"
@@ -1148,18 +1160,26 @@ func classifyDenialReason(status *CLIAuthStatus, currentChannel string) string {
 	if status.Result == nil || !status.Success {
 		return "unknown"
 	}
-	if status.Result.CLIAuthEnabled {
+	r := status.Result
+	if r.CLIAuthEnabled {
 		return ""
 	}
-	// cliAuthEnabled=false — infer reason from auxiliary fields (same priority as server)
-	if status.Result.UserScope == "forbidden" {
+
+	if r.UserScope == "forbidden" {
 		return "user_forbidden"
 	}
-	if status.Result.UserScope == "specified" {
-		return "user_not_allowed"
+
+	if r.ChannelScope == "specified" {
+		if currentChannel == "" {
+			return "channel_required"
+		}
+		if !slices.Contains(r.AllowedChannels, currentChannel) {
+			return "channel_not_allowed"
+		}
 	}
-	if status.Result.ChannelScope == "specified" && currentChannel != "" {
-		return "channel_not_allowed"
+
+	if r.UserScope == "specified" {
+		return "user_not_allowed"
 	}
 	return "cli_not_enabled"
 }
@@ -131,9 +131,6 @@ func (p *Plugin) ToServerDescriptors() []market.ServerDescriptor {
 		}

 		source := "plugin"
-		if p.IsManaged {
-			source = "plugin-managed"
-		}

 		// Resolve headers: expand environment variable references (e.g. ${DASHSCOPE_API_KEY}).
 		var resolvedHeaders map[string]string
@@ -56,13 +56,6 @@ type Settings struct {
 	DevPlugins       map[string]string         `json:"devPlugins,omitempty"` // name → absolute path
 }

-// LoadManaged scans ~/.dws/plugins/managed/ and returns all valid
-// official plugins. Managed plugins are always enabled.
-func (l *Loader) LoadManaged() []*Plugin {
-	managedDir := filepath.Join(l.PluginsDir, "managed")
-	return l.scanDir(managedDir, true)
-}
-
 // LoadUser scans ~/.dws/plugins/user/ and returns enabled user plugins.
 func (l *Loader) LoadUser() []*Plugin {
 	userDir := filepath.Join(l.PluginsDir, "user")
@@ -86,7 +79,7 @@ func (l *Loader) LoadUser() []*Plugin {

 		// Check if this is a direct plugin directory (has plugin.json)
 		if _, err := os.Stat(filepath.Join(entryPath, "plugin.json")); err == nil {
-			p := l.loadPlugin(entryPath, false)
+			p := l.loadPlugin(entryPath)
 			if p != nil && isPluginEnabled(settings, p.Manifest.Name) {
 				plugins = append(plugins, p)
 			}
@@ -103,7 +96,7 @@ func (l *Loader) LoadUser() []*Plugin {
 				continue
 			}
 			subPath := filepath.Join(entryPath, sub.Name())
-			p := l.loadPlugin(subPath, false)
+			p := l.loadPlugin(subPath)
 			if p != nil {
 				qualifiedName := entry.Name() + "/" + p.Manifest.Name
 				if isPluginEnabled(settings, qualifiedName) {
@@ -115,39 +108,15 @@ func (l *Loader) LoadUser() []*Plugin {
 	return plugins
 }

-// LoadAll loads both managed and user plugins.
+// LoadAll loads user + dev plugins.
 func (l *Loader) LoadAll() []*Plugin {
-	managed := l.LoadManaged()
 	user := l.LoadUser()
-	return append(managed, user...)
-}
-
-// scanDir reads a directory of plugin subdirectories and loads each one.
-func (l *Loader) scanDir(dir string, isManaged bool) []*Plugin {
-	entries, err := os.ReadDir(dir)
-	if err != nil {
-		if !os.IsNotExist(err) {
-			slog.Debug("plugin: cannot read dir", "path", dir, "error", err)
-		}
-		return nil
-	}
-
-	var plugins []*Plugin
-	for _, entry := range entries {
-		if !entry.IsDir() {
-			continue
-		}
-		pluginDir := filepath.Join(dir, entry.Name())
-		p := l.loadPlugin(pluginDir, isManaged)
-		if p != nil {
-			plugins = append(plugins, p)
-		}
-	}
-	return plugins
+	dev := l.LoadDev()
+	return append(user, dev...)
 }

 // loadPlugin reads and validates a single plugin directory.
-func (l *Loader) loadPlugin(dir string, isManaged bool) *Plugin {
+func (l *Loader) loadPlugin(dir string) *Plugin {
 	manifestPath := filepath.Join(dir, "plugin.json")
 	manifest, err := ParseManifest(manifestPath)
 	if err != nil {
@@ -163,9 +132,8 @@ func (l *Loader) loadPlugin(dir string, isManaged bool) *Plugin {
 	}

 	return &Plugin{
-		Manifest:  *manifest,
-		Root:      dir,
-		IsManaged: isManaged,
+		Manifest: *manifest,
+		Root:     dir,
 	}
 }

@@ -211,7 +179,7 @@ func isPluginEnabled(s *Settings, name string) bool {
 type PluginInfo struct {
 	Name        string `json:"name"`
 	Version     string `json:"version"`
-	Type        string `json:"type"` // "managed" or "user"
+	Type        string `json:"type"` // "user" or "dev"
 	Enabled     bool   `json:"enabled"`
 	Path        string `json:"path"`
 	Description string `json:"description,omitempty"`
@@ -222,29 +190,6 @@ func (l *Loader) ListInstalled() []PluginInfo {
 	var result []PluginInfo
 	settings := l.loadSettings()

-	// Managed plugins
-	managedDir := filepath.Join(l.PluginsDir, "managed")
-	if entries, err := os.ReadDir(managedDir); err == nil {
-		for _, entry := range entries {
-			if !entry.IsDir() {
-				continue
-			}
-			dir := filepath.Join(managedDir, entry.Name())
-			m, err := ParseManifest(filepath.Join(dir, "plugin.json"))
-			if err != nil {
-				continue
-			}
-			result = append(result, PluginInfo{
-				Name:        m.Name,
-				Version:     m.Version,
-				Type:        "managed",
-				Enabled:     true, // managed plugins always enabled
-				Path:        dir,
-				Description: m.Description,
-			})
-		}
-	}
-
 	// User plugins
 	userDir := filepath.Join(l.PluginsDir, "user")
 	if entries, err := os.ReadDir(userDir); err == nil {
@@ -348,9 +293,8 @@ func (l *Loader) InstallFromDir(srcDir string) (*Plugin, error) {
 	l.setPluginEnabled(manifest.Name, true)

 	return &Plugin{
-		Manifest:  *manifest,
-		Root:      destDir,
-		IsManaged: false,
+		Manifest: *manifest,
+		Root:     destDir,
 	}, nil
 }

@@ -411,9 +355,8 @@ func (l *Loader) InstallFromGit(gitURL string) (*Plugin, error) {
 	l.setPluginEnabled(qualifiedName, true)

 	return &Plugin{
-		Manifest:  *manifest,
-		Root:      destDir,
-		IsManaged: false,
+		Manifest: *manifest,
+		Root:     destDir,
 	}, nil
 }

@@ -463,19 +406,9 @@ func parseGitURL(gitURL string) (workspace, repoName string, err error) {
 	return segments[len(segments)-2], segments[len(segments)-1], nil
 }

-// RemovePlugin removes an installed plugin by name. It searches both the
-// user and the legacy managed directories; all plugins are equally
-// removable.
+// RemovePlugin removes an installed plugin by name.
 func (l *Loader) RemovePlugin(name string, keepData bool) error {
 	pluginDir := l.findUserPluginDir(name)
-	if pluginDir == "" {
-		// Fall back to legacy managed/ directory (for plugins installed
-		// by older CLI builds that wrote under ~/.dws/plugins/managed/).
-		legacyDir := filepath.Join(l.PluginsDir, config.PluginManagedDir, name)
-		if _, err := os.Stat(filepath.Join(legacyDir, "plugin.json")); err == nil {
-			pluginDir = legacyDir
-		}
-	}
 	if pluginDir == "" {
 		return fmt.Errorf("plugin %q not found", name)
 	}
@@ -516,12 +449,8 @@ func (l *Loader) purgePluginFromSettings(name string) {

 // SetEnabled enables or disables a plugin in settings.json.
 func (l *Loader) SetEnabled(name string, enabled bool) error {
-	// Verify plugin exists
 	if l.findUserPluginDir(name) == "" {
-		managedDir := filepath.Join(l.PluginsDir, "managed", name)
-		if _, err := os.Stat(managedDir); err != nil {
-			return fmt.Errorf("plugin %q not found", name)
-		}
+		return fmt.Errorf("plugin %q not found", name)
 	}
 	l.setPluginEnabled(name, enabled)
 	return nil
@@ -695,7 +624,7 @@ func (l *Loader) LoadDev() []*Plugin {
 				"name", name, "dir", dir)
 			continue
 		}
-		p := l.loadPlugin(dir, false)
+		p := l.loadPlugin(dir)
 		if p != nil {
 			plugins = append(plugins, p)
 			slog.Debug("plugin: loaded dev plugin", "name", name, "dir", dir)
@@ -93,9 +93,8 @@ type HookEntry struct {

 // Plugin is a loaded, validated plugin ready for injection.
 type Plugin struct {
-	Manifest  Manifest
-	Root      string // absolute path to plugin directory
-	IsManaged bool   // true for official (DingTalk-Real-AI) plugins
+	Manifest Manifest
+	Root     string // absolute path to plugin directory
 }

 // ParseManifest reads and parses a plugin.json file.
@@ -199,8 +199,7 @@ func TestPluginToServerDescriptors(t *testing.T) {
 				},
 			},
 		},
-		Root:      "/tmp/plugins/conference",
-		IsManaged: true,
+		Root: "/tmp/plugins/conference",
 	}

 	descriptors := p.ToServerDescriptors()
@@ -217,8 +216,8 @@ func TestPluginToServerDescriptors(t *testing.T) {
 	if d.Endpoint != "https://mcp.conference.dingtalk.com" {
 		t.Errorf("endpoint = %q", d.Endpoint)
 	}
-	if d.Source != "plugin-managed" {
-		t.Errorf("source = %q, want plugin-managed", d.Source)
+	if d.Source != "plugin" {
+		t.Errorf("source = %q, want plugin", d.Source)
 	}
 	if d.CLI.ID != "conference" {
 		t.Errorf("cli.id = %q, want conference", d.CLI.ID)
@@ -273,7 +272,7 @@ func TestPluginToServerDescriptorsWithHeaders(t *testing.T) {
 		t.Errorf("AuthHeaders[X-Custom] = %q, want static-value", d.AuthHeaders["X-Custom"])
 	}
 	if d.Source != "plugin" {
-		t.Errorf("source = %q, want plugin (non-managed)", d.Source)
+		t.Errorf("source = %q, want plugin", d.Source)
 	}
 }

@@ -354,137 +353,57 @@ func TestLoaderScanEmpty(t *testing.T) {
 		CLIVersion: "1.0.0",
 	}

-	managed := loader.LoadManaged()
-	if len(managed) != 0 {
-		t.Errorf("expected 0 managed plugins, got %d", len(managed))
-	}
-
 	user := loader.LoadUser()
 	if len(user) != 0 {
 		t.Errorf("expected 0 user plugins, got %d", len(user))
 	}
 }

-func TestLoaderLoadManaged(t *testing.T) {
-	dir := t.TempDir()
-	managedDir := filepath.Join(dir, "managed", "conference")
-	if err := os.MkdirAll(managedDir, 0o755); err != nil {
-		t.Fatal(err)
-	}
-
-	manifest := `{
-		"name": "conference",
-		"version": "1.0.0",
-		"type": "managed",
-		"mcpServers": {
-			"conference": {
-				"type": "streamable-http",
-				"endpoint": "https://example.com"
-			}
-		}
-	}`
-	if err := os.WriteFile(filepath.Join(managedDir, "plugin.json"), []byte(manifest), 0o644); err != nil {
-		t.Fatal(err)
-	}
-
-	loader := &Loader{PluginsDir: dir, CLIVersion: "1.0.0"}
-	plugins := loader.LoadManaged()
-
-	if len(plugins) != 1 {
-		t.Fatalf("expected 1 managed plugin, got %d", len(plugins))
-	}
-	if plugins[0].Manifest.Name != "conference" {
-		t.Errorf("name = %q, want conference", plugins[0].Manifest.Name)
-	}
-	if !plugins[0].IsManaged {
-		t.Error("expected IsManaged = true")
-	}
-}
-
-// TestRemoveLegacyManagedPlugin ensures plugins that were installed under
-// the legacy ~/.dws/plugins/managed/ directory are now freely removable
-// — the old "cannot be removed" privilege has been dropped.
-func TestRemoveLegacyManagedPlugin(t *testing.T) {
-	dir := t.TempDir()
-	managedDir := filepath.Join(dir, "managed", "conference")
-	if err := os.MkdirAll(managedDir, 0o755); err != nil {
-		t.Fatal(err)
-	}
-	if err := os.WriteFile(filepath.Join(managedDir, "plugin.json"), []byte(`{"name":"conference","version":"1.0.0"}`), 0o644); err != nil {
-		t.Fatal(err)
-	}
-
-	loader := &Loader{PluginsDir: dir, CLIVersion: "1.0.0"}
-	if err := loader.RemovePlugin("conference", false); err != nil {
-		t.Fatalf("unexpected error removing legacy managed plugin: %v", err)
-	}
-	if _, err := os.Stat(managedDir); !os.IsNotExist(err) {
-		t.Errorf("managed plugin dir should be removed, stat err = %v", err)
-	}
-}
-
 // TestRemovePluginPurgesSettings verifies RemovePlugin fully purges the
 // plugin's settings — both its enabled flag and any pluginConfigs entry —
 // so settings.json does not retain dangling state for a plugin that no
-// longer exists on disk. Covers both the user and legacy managed paths.
+// longer exists on disk.
 func TestRemovePluginPurgesSettings(t *testing.T) {
-	cases := []struct {
-		name    string
-		layout  string // "user" or "legacy"
-		pkgName string
-	}{
-		{name: "user plugin", layout: "user", pkgName: "my-plugin"},
-		{name: "legacy managed plugin", layout: "legacy", pkgName: "conference"},
+	const pkgName = "my-plugin"
+	dir := t.TempDir()
+	pluginDir := filepath.Join(dir, "user", pkgName)
+	if err := os.MkdirAll(pluginDir, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(pluginDir, "plugin.json"),
+		[]byte(`{"name":"`+pkgName+`","version":"1.0.0"}`), 0o644); err != nil {
+		t.Fatal(err)
 	}
-	for _, tc := range cases {
-		t.Run(tc.name, func(t *testing.T) {
-			dir := t.TempDir()
-			var pluginDir string
-			switch tc.layout {
-			case "user":
-				pluginDir = filepath.Join(dir, "user", tc.pkgName)
-			case "legacy":
-				pluginDir = filepath.Join(dir, "managed", tc.pkgName)
-			}
-			if err := os.MkdirAll(pluginDir, 0o755); err != nil {
-				t.Fatal(err)
-			}
-			if err := os.WriteFile(filepath.Join(pluginDir, "plugin.json"),
-				[]byte(`{"name":"`+tc.pkgName+`","version":"1.0.0"}`), 0o644); err != nil {
-				t.Fatal(err)
-			}

-			loader := &Loader{PluginsDir: dir, CLIVersion: "1.0.0"}
+	loader := &Loader{PluginsDir: dir, CLIVersion: "1.0.0"}

-			// Seed settings.json with an explicit enabled flag and a
-			// pluginConfigs entry to verify both get purged.
-			settings := &Settings{
-				EnabledPlugins: map[string]bool{tc.pkgName: true, "other-plugin": true},
-				PluginConfigs: map[string]map[string]any{
-					tc.pkgName:     {"API_KEY": "secret"},
-					"other-plugin": {"TOKEN": "keep-me"},
-				},
-			}
-			loader.saveSettings(settings)
+	// Seed settings.json with an explicit enabled flag and a
+	// pluginConfigs entry to verify both get purged.
+	settings := &Settings{
+		EnabledPlugins: map[string]bool{pkgName: true, "other-plugin": true},
+		PluginConfigs: map[string]map[string]any{
+			pkgName:        {"API_KEY": "secret"},
+			"other-plugin": {"TOKEN": "keep-me"},
+		},
+	}
+	loader.saveSettings(settings)

-			if err := loader.RemovePlugin(tc.pkgName, false); err != nil {
-				t.Fatalf("RemovePlugin: %v", err)
-			}
+	if err := loader.RemovePlugin(pkgName, false); err != nil {
+		t.Fatalf("RemovePlugin: %v", err)
+	}

-			reloaded := loader.loadSettings()
-			if _, exists := reloaded.EnabledPlugins[tc.pkgName]; exists {
-				t.Errorf("EnabledPlugins should not retain removed plugin %q", tc.pkgName)
-			}
-			if _, exists := reloaded.PluginConfigs[tc.pkgName]; exists {
-				t.Errorf("PluginConfigs should not retain removed plugin %q", tc.pkgName)
-			}
-			if !reloaded.EnabledPlugins["other-plugin"] {
-				t.Error("unrelated EnabledPlugins entry should be preserved")
-			}
-			if reloaded.PluginConfigs["other-plugin"]["TOKEN"] != "keep-me" {
-				t.Error("unrelated PluginConfigs entry should be preserved")
-			}
-		})
+	reloaded := loader.loadSettings()
+	if _, exists := reloaded.EnabledPlugins[pkgName]; exists {
+		t.Errorf("EnabledPlugins should not retain removed plugin %q", pkgName)
+	}
+	if _, exists := reloaded.PluginConfigs[pkgName]; exists {
+		t.Errorf("PluginConfigs should not retain removed plugin %q", pkgName)
+	}
+	if !reloaded.EnabledPlugins["other-plugin"] {
+		t.Error("unrelated EnabledPlugins entry should be preserved")
+	}
+	if reloaded.PluginConfigs["other-plugin"]["TOKEN"] != "keep-me" {
+		t.Error("unrelated PluginConfigs entry should be preserved")
 	}
 }

@@ -112,12 +112,6 @@ const (
 // ── Plugin system ──────────────────────────────────────────────────────

 const (
-	// PluginManagedDir is a legacy subdirectory under ~/.dws/plugins/
-	// retained so plugins installed by older CLI versions remain loadable.
-	// New installs always go to PluginUserDir; there is no privilege
-	// associated with this directory anymore.
-	PluginManagedDir = "managed"
-
 	// PluginUserDir is the subdirectory under ~/.dws/plugins/ where all
 	// third-party plugins are installed. Every plugin — whether authored
 	// by the DingTalk team or anyone else — lives here with equal status.
@@ -1,77 +1,113 @@
-# 群聊与机器人 (chat) 命令参考
+# 会话与群聊 (chat) 命令参考
+
+> 命令别名: `dws im` 等价于 `dws chat`

 ## 命令总览

+### group (群组管理)
+
 | 子命令 | 用途 |
 |-------|------|
-| `search` | 搜索群聊 |
-| `group create` | 创建群 |
+| `group create` | 创建内部群 |
+| `group create-org` | 创建企业全员群 |
 | `group members list` | 查看群成员列表 |
 | `group members add` | 添加群成员 |
 | `group members remove` | 移除群成员（⚠️ 危险操作） |
 | `group members add-bot` | 添加机器人到群 |
 | `group rename` | 修改群名称 |
-| `bot search` | 搜索我的机器人 |
+| `search` | 搜索群会话 |
+| `search-common` | 搜索共同群 |
+
+### message (会话消息管理)
+
+| 子命令 | 用途 |
+|-------|------|
+| `message send` | 以当前用户身份发群消息或单聊消息 |
+| `message send-personal` | 发送个人消息（⚠️ 敏感操作） |
+| `message list` | 拉取群聊或单聊会话消息 |
+| `message list-all` | 按时间范围拉取当前用户所有会话消息 |
+| `message list-topic-replies` | 拉取群话题回复消息列表 |
+| `message list-by-sender` | 搜索指定发送者的消息 |
+| `message list-mentions` | 拉取 @我 的消息 |
+| `message list-focused` | 拉取特别关注人的消息 |
+| `message list-unread-conversations` | 获取未读会话列表 |
+| `message search` | 按关键词搜索消息 |
+| `message info` | 获取会话信息 |
 | `message send-by-bot` | 机器人发消息（群聊或批量单聊） |
-| `message recall-by-bot` | 机器人撤回消息（群聊或批量单聊） |
+| `message recall-by-bot` | 机器人撤回消息 |
 | `message send-by-webhook` | 自定义机器人 Webhook 发消息 |
+| `list-top-conversations` | 拉取置顶会话列表 |
+
+### bot (机器人管理)
+
+| 子命令 | 用途 |
+|-------|------|
+| `bot search` | 搜索我的机器人 |
+| `bot create` | 创建企业机器人 |
+| `bot search-groups` | 搜索机器人所在群 |

 ---

-## search — 搜索群聊
+## group create — 创建内部群

-```
-Usage:
-  dws chat search [flags]
-Example:
-  dws chat search --query "项目冲刺" --format json
-Flags:
-      --query string    搜索关键词 (必填)
-      --cursor string   分页游标（首页留空）
-```
-
---
-
-## group create — 创建群
+当前登录用户自动成为群主。

 ```
 Usage:
  dws chat group create [flags]
 Example:
-  dws chat group create --name "Q1 项目冲刺群" --users userId1,userId2,userId3 --format json
+  dws chat group create --name "Q1 项目冲刺群" --users userId1,userId2,userId3
 Flags:
-      --name string    群名称 (必填)
-      --users string   群成员 userId 列表，逗号分隔 (必填)
+      --users string    成员 userId 列表，用户本身会自动加入，无需包含，逗号分隔，不超过20个 (必填)
+      --name string     群名称 (必填)
 ```

-> 当前用户自动作为群主加入，无需在 --users 中重复传入。
+---
+
+## group create-org — 创建企业全员群
+
+创建面向企业组织的群，成员通过 userId 列表指定。
+
+```
+Usage:
+  dws chat group create-org [flags]
+Example:
+  dws chat group create-org --name "全员通知群" --users userId1,userId2
+Flags:
+      --name string     群名称 (必填)
+      --users string    成员 userId 列表，逗号分隔 (必填)
+```

 ---

 ## group members list — 查看群成员列表

+分页查询指定群聊的成员。
+
 ```
 Usage:
  dws chat group members list [flags]
 Example:
-  dws chat group members list --id <openConversationId> --format json
+  dws chat group members list --id <openconversation_id>
 Flags:
-      --id string       群会话 ID (必填)
-      --cursor string   分页游标
+      --cursor string   分页游标，首次从 0 开始
+      --id string       群 ID / openconversation_id (必填)
 ```

 ---

 ## group members add — 添加群成员

+向指定群聊添加成员，需传入群 ID 与用户 ID 列表。
+
 ```
 Usage:
  dws chat group members add [flags]
 Example:
-  dws chat group members add --id <openConversationId> --users userId1,userId2 --format json
+  dws chat group members add --id <openconversation_id> --users userId1,userId2
 Flags:
-      --id string      群会话 ID (必填)
-      --users string   要添加的 userId 列表，逗号分隔 (必填)
+      --id string      群 ID / openconversation_id (必填)
+      --users string   要添加的用户 userId 列表，逗号分隔 (必填)
 ```

 ---
@@ -84,24 +120,26 @@ Flags:
 Usage:
  dws chat group members remove [flags]
 Example:
-  dws chat group members remove --id <openConversationId> --users userId1,userId2 --format json
+  dws chat group members remove --id <openconversation_id> --users userId1,userId2
 Flags:
-      --id string      群会话 ID (必填)
-      --users string   要移除的 userId 列表，逗号分隔 (必填)
+      --id string      群 ID / openconversation_id (必填)
+      --users string   要移除的用户 userId 列表，逗号分隔 (必填)
 ```

 ---

 ## group members add-bot — 添加机器人到群

+将自定义机器人添加到当前用户有管理权限的群聊中，如果没有权限则会报错。
+
 ```
 Usage:
  dws chat group members add-bot [flags]
 Example:
-  dws chat group members add-bot --id <openConversationId> --robot-code <robotCode> --format json
+  dws chat group members add-bot --robot-code <robot-code> --id <openconversation_id>
 Flags:
-      --id string           群会话 ID (必填)
-      --robot-code string   机器人 code (必填)
+      --id string           群聊 openConversationId (必填)
+      --robot-code string   机器人 Code (必填)
 ```

 ---
@@ -112,10 +150,351 @@ Flags:
 Usage:
  dws chat group rename [flags]
 Example:
-  dws chat group rename --id <openConversationId> --name "新群名" --format json
+  dws chat group rename --id <openconversation_id> --name "新群名"
 Flags:
-      --id string     群会话 ID (必填)
-      --name string   新群名称 (必填)
+      --id string     群 ID / openconversation_id (必填)
+      --name string   修改后的群名称 (必填)
+```
+
+---
+
+## search — 搜索群会话
+
+根据名称搜索会话列表。
+
+```
+Usage:
+  dws chat search [flags]
+Example:
+  dws chat search --query "项目冲刺"
+Flags:
+      --cursor string   分页游标 (首页留空)
+      --query string    搜索关键词 (必填)
+```
+
+---
+
+## search-common — 搜索共同群
+
+根据昵称列表搜索共同群聊。--nicks 指定要搜索的人员昵称（逗号分隔，必填）。--match-mode 控制匹配模式：AND 表示所有人都在群里，OR 表示任一人在群里（默认 AND）。
+
+```
+Usage:
+  dws chat search-common [flags]
+Example:
+  dws chat search-common --nicks "风雷,山乔" --limit 20 --cursor 0
+  dws chat search-common --nicks "天鸡,乐函" --match-mode OR --limit 20 --cursor 0
+  dws chat search-common --nicks "风雷,山乔,天鸡" --limit 10 --cursor <nextCursor>
+Flags:
+      --nicks string        要搜索的昵称列表，逗号分隔 (必填)
+      --match-mode string   匹配模式：AND=所有人都在群里，OR=任一人在群里（默认 AND）
+      --limit int           每页返回数量（默认 20）
+      --cursor string       分页游标（默认 "0"，翻页传 nextCursor）
+
+注意:
+  - --nicks 传人员昵称（花名），逗号分隔，如 "风雷,山乔"
+  - --match-mode AND 表示群里必须包含所有指定的人；OR 表示包含任意一人即可
+  - 翻页：hasMore=true 时，用返回的 nextCursor 作为下次 --cursor
+```
+
+---
+
+## message send — 以当前用户身份发消息
+
+--group 指定群聊 ID 发群消息；--user 指定用户 userId 发单聊；--open-dingtalk-id 指定用户 openDingTalkId 发单聊。三者只能选其一，不能同时指定。消息内容为位置参数（恰好 1 个），支持 Markdown。可选 --title 作为消息标题。
+--群聊时可选 --at-all @所有人，或 --at-users 指定成员（仅群聊时生效）。
+--发送图片消息：指定 --media-id（通过 dt_media_upload 工具上传获得），自动设置 msgType=image，此时不需要传文本内容。
+
+```
+Usage:
+  dws chat message send [flags] [<text>]
+Example:
+  dws chat message send --group <openconversation_id> --text "hello"
+  dws chat message send --user <userId> --text "请查收"
+  dws chat message send --open-dingtalk-id <openDingTalkId> --text "请查收"
+  dws chat message send --group <openconversation_id> "hello"
+  dws chat message send --group <openconversation_id> --title "周报提醒" --text "请大家本周五前提交周报"
+  dws chat message send --group <openconversation_id> --at-all "<@all> 请大家注意"
+  dws chat message send --group <openconversation_id> --at-users userId1,userId2 "<@userId1> <@userId2> 请查收"
+  dws chat message send --group <openconversation_id> --media-id <mediaId>
+  dws chat message send --open-dingtalk-id <openDingTalkId> --media-id <mediaId>
+Flags:
+      --text string              消息内容（推荐使用，也可用位置参数）
+      --group string             群聊 openconversation_id（群聊时必填）
+      --user string              接收人 userId（单聊时与 --open-dingtalk-id 二选一）
+      --open-dingtalk-id string  接收人 openDingTalkId（单聊时与 --user 二选一，适用于三方应用等无法获取 userId 的场景）
+      --title string             消息标题（可选，默认「消息」）
+      --at-all                   @所有人（仅群聊时生效，可选，默认 false）
+      --at-users string          @指定成员的 userId 列表，逗号分隔（仅群聊时生效，可选）
+      --media-id string          图片 mediaId（通过 dt_media_upload 工具上传获得，需从返回链接中去除 _宽_高.格式 后缀并加上 @ 前缀），指定后发送图片消息，不需要传文本内容
+
+注意:
+  - --text 和位置参数二选一，--text 优先
+  - --group、--user、--open-dingtalk-id 三者互斥，只需指定其一：群聊用 --group，单聊用 --user 或 --open-dingtalk-id
+  - --group 的别名: --id, --chat, --conversation-id (均可替代 --group)
+  - --at-all 和 --at-users 仅在 --group 群聊时生效；当设置--at-all时，消息内容中一定要包含对应的占位符<@all>；当设置--at-users userId1,userId2时，消息内容中一定要包含对应格式的占位符<@userId1> <@userId2>
+  - --media-id 指定图片 mediaId 时自动发送图片消息（msgType=image），不需要传 --text；图片单聊仅支持 --open-dingtalk-id，不支持 --user
+```
+
+---
+
+## message send-personal — 发送个人消息
+
+> ⚠️ 敏感操作：执行前必须向用户确认，同意后才加 `--yes`。
+
+发送个人消息到指定会话或指定用户。支持 @指定人。
+
+```
+Usage:
+  dws chat message send-personal [flags]
+Example:
+  dws chat message send-personal --id <openConversationId> --content "你好" --type text
+  dws chat message send-personal --open-id <openDingTalkId> --content "消息内容" --type text
+  dws chat message send-personal --id <openConversationId> --content "内容" --at-all
+Flags:
+      --content string   消息内容 (必填)
+      --type string      消息类型，如 text、markdown (必填)
+      --id string        群聊会话 ID openConversationId（与 --open-id 二选一）
+      --open-id string   接收人 openDingTalkId（与 --id 二选一）
+      --at-all           @所有人（可选）
+      --at-users string  @指定人的 openDingTalkId 列表，逗号分隔（可选）
+
+注意:
+  - --id（群聊会话）和 --open-id（指定用户 openDingTalkId）二选一
+```
+
+---
+
+## message list — 拉取会话消息内容
+
+拉取指定群聊或单聊的会话消息内容。
+
+--group 指定群聊，--user 指定单聊用户（通过 userId），--open-dingtalk-id 指定单聊用户（通过 openDingTalkId），三者互斥。默认拉取给定时间之后的消息，--forward=false 拉之前的。hasMore=true 时用结果中的边界 createTime 作为下次 --time 翻页。
+
+```
+Usage:
+  dws chat message list [flags]
+Example:
+  dws chat message list --group <openconversation_id> --time "2025-03-01 00:00:00"
+  dws chat message list --user <userId> --time "2025-03-01 00:00:00" --limit 50
+  dws chat message list --open-dingtalk-id <openDingTalkId> --time "2025-03-01 00:00:00" --limit 50
+  dws chat message list --group <openconversation_id> --time "2025-03-01 00:00:00" --forward=false
+Flags:
+      --forward                  true=拉给定时间之后的消息，false=拉给定时间之前的消息 (default true)
+      --group string             群聊 openconversation_id（群聊时必填）
+      --limit int                返回数量，不传则不限制
+      --time string              开始时间，格式: yyyy-MM-dd HH:mm:ss (必填)
+      --user string              单聊用户 userId（单聊时与 --open-dingtalk-id 二选一）
+      --open-dingtalk-id string  单聊用户 openDingTalkId（单聊时与 --user 二选一，适用于三方应用等无法获取 userId 的场景）
+
+注意:
+  - --group、--user、--open-dingtalk-id 三者互斥，只需指定其一：群聊用 --group，单聊用 --user 或 --open-dingtalk-id
+  - --group 的别名: --id, --chat, --conversation-id (均可替代 --group)
+  - 翻页：hasMore=true 时，用结果中的边界 createTime 作为下次 --time
+  - 如果返回的会话消息中包含 openConvThreadId 字段，说明是话题类消息，需要调用 dws chat message list-topic-replies 拉取话题的回复内容列表，openConvThreadId 作为 --topic-id 参数
+```
+
+---
+
+## message list-all — 拉取指定时间范围内当前用户的所有会话消息
+
+分页拉取当前登录用户在指定时间范围内的所有会话消息。
+
+--start 和 --end 限定时间范围，--limit 指定每页数量，--cursor 传分页游标（首页传 "0"，后续从响应中的 nextCursor 获取）。
+
+```
+Usage:
+  dws chat message list-all [flags]
+Example:
+  dws chat message list-all --start "2025-03-01 00:00:00" --end "2025-03-31 23:59:59" --limit 50
+  dws chat message list-all --start "2025-03-01 00:00:00" --end "2025-03-31 23:59:59" --limit 50 --cursor "abc123token"
+Flags:
+      --start string    起始时间，格式: yyyy-MM-dd HH:mm:ss (必填)
+      --end string      结束时间，格式: yyyy-MM-dd HH:mm:ss (必填)
+      --limit int       每页返回数量（默认 50）
+      --cursor string   分页游标（首页传 "0"，后续从响应中的 nextCursor 获取）
+
+注意:
+  - 四个参数每次请求都会传递给服务端，cursor 首页传 "0"
+  - 与 chat message list 的区别：list 拉取指定单个会话（群聊或单聊）的消息，list-all 拉取当前用户所有会话的消息
+  - 翻页：hasMore=true 时，用响应中的 nextCursor 值作为下次 --cursor 参数继续翻页
+  - 时间格式统一为 yyyy-MM-dd HH:mm:ss
+```
+
+---
+
+## message list-topic-replies — 拉取群话题回复消息列表
+
+查询指定群聊中某条话题消息的全部回复。--group 指定群会话 ID，--topic-id 指定话题 ID（由 dws chat message list 返回）。
+
+```
+Usage:
+  dws chat message list-topic-replies [flags]
+Example:
+  dws chat message list-topic-replies --group <openconversation_id> --topic-id <topicId>
+  dws chat message list-topic-replies --group <openconversation_id> --topic-id <topicId> --time "2025-03-01 00:00:00" --limit 20
+Flags:
+      --group string      群会话 openconversationId (必填)
+      --topic-id string   话题 ID，由 dws chat message list 返回 (必填)
+      --time string       开始时间，格式: yyyy-MM-dd HH:mm:ss（可选）
+      --limit int         返回数量（默认 50）
+      --forward           true=从老往新，false=从新往老（默认 false）
+```
+
+---
+
+## message list-by-sender — 拉取指定发送者的消息
+
+搜索特定人发送给我的消息，返回结果包含单聊和群聊标识。--sender-user-id 指定发送者 userId，--sender-open-dingtalk-id 指定发送者 openDingTalkId，二者互斥。
+
+```
+Usage:
+  dws chat message list-by-sender [flags]
+Example:
+  dws chat message list-by-sender --sender-user-id <userId> --start "2026-03-10T00:00:00+08:00" --end "2026-03-11T00:00:00+08:00" --limit 50 --cursor 0
+  dws chat message list-by-sender --sender-open-dingtalk-id <openDingTalkId> --start "2026-03-10T00:00:00+08:00" --end "2026-03-11T00:00:00+08:00" --limit 50 --cursor 0
+Flags:
+      --sender-user-id string              发送者 userId（与 --sender-open-dingtalk-id 二选一）
+      --sender-open-dingtalk-id string     发送者 openDingTalkId（与 --sender-user-id 二选一）
+      --start string                       开始时间，ISO-8601 格式 (必填)
+      --end string                         结束时间，ISO-8601 格式 (必填)
+      --limit int                          每页返回数量（默认 50）
+      --cursor string                      分页游标（默认 "0"，翻页传 nextCursor）
+
+注意:
+  - --sender-user-id 和 --sender-open-dingtalk-id 二者互斥，必须且只能指定其一
+  - 不需要指定单聊/群聊，MCP 返回结果自带会话类型标识
+  - 时间支持多种 ISO-8601 格式，如 "2026-03-10T00:00:00+08:00"、"2026-03-10 14:00:00" 等
+  - 翻页：hasMore=true 时，用返回的 nextCursor 作为下次 --cursor
+```
+
+---
+
+## message list-mentions — 拉取 @我 的消息
+
+搜索时间范围内 @我 的消息，可选指定群聊。
+
+```
+Usage:
+  dws chat message list-mentions [flags]
+Example:
+  dws chat message list-mentions --start "2026-03-10T00:00:00+08:00" --end "2026-03-11T00:00:00+08:00" --limit 50 --cursor 0
+  dws chat message list-mentions --group <openconversation_id> --start "2026-03-10T00:00:00+08:00" --end "2026-03-11T00:00:00+08:00" --limit 50 --cursor 0
+Flags:
+      --group string    群聊 openconversation_id（可选，不传则查全部）
+      --start string    开始时间，ISO-8601 格式 (必填)
+      --end string      结束时间，ISO-8601 格式 (必填)
+      --limit int       每页返回数量（默认 50）
+      --cursor string   分页游标（默认 "0"，翻页传 nextCursor）
+
+注意:
+  - --group 可选，不传则查询所有会话中 @我 的消息；传入则只查指定群聊
+  - --group 的别名: --id, --chat, --conversation-id (均可替代 --group)
+  - 翻页：hasMore=true 时，用返回的 nextCursor 作为下次 --cursor
+```
+
+---
+
+## message list-focused — 拉取特别关注人的消息
+
+拉取当前用户特别关注人的消息。
+
+```
+Usage:
+  dws chat message list-focused [flags]
+Example:
+  dws chat message list-focused --limit 50
+  dws chat message list-focused --limit 20 --cursor <nextCursor>
+Flags:
+      --limit int       每页返回数量（默认 50）
+      --cursor int64    分页游标（首次不传或传 0，翻页传 nextCursor）
+
+注意:
+  - 首次调用不传 --cursor 或传 0，后续翻页传 nextCursor
+```
+
+---
+
+## message list-unread-conversations — 获取未读会话列表
+
+获取当前用户有未读消息的会话信息。可选通过 `--count` 限制返回条数。
+
+```
+Usage:
+  dws chat message list-unread-conversations [flags]
+Example:
+  dws chat message list-unread-conversations
+  dws chat message list-unread-conversations --count 20
+Flags:
+      --count int    返回未读会话条数（可选）
+```
+
+---
+
+## message search — 按关键词搜索消息
+
+在当前用户的会话中按关键词搜索消息。--keyword 必填，可选 --group 限定搜索某个会话。
+
+```
+Usage:
+  dws chat message search [flags]
+Example:
+  dws chat message search --keyword "changefree" --start "2026-04-01T00:00:00+08:00" --end "2026-04-15T00:00:00+08:00" --limit 50 --cursor 0
+  dws chat message search --keyword "codereview" --group <openconversation_id> --start "2026-04-01T00:00:00+08:00" --end "2026-04-15T00:00:00+08:00" --limit 100 --cursor 0
+Flags:
+      --keyword string   搜索关键词 (必填)
+      --group string     群聊 openconversation_id（可选，不传则搜索所有会话）
+      --start string     开始时间，ISO-8601 格式 (必填)
+      --end string       结束时间，ISO-8601 格式 (必填)
+      --limit int        每页返回数量（默认 100）
+      --cursor string    分页游标（默认 "0"，翻页传 nextCursor）
+
+注意:
+  - --group 可选，不传则搜索所有会话中的消息；传入则只搜索指定会话
+  - --group 的别名: --id, --chat, --conversation-id (均可替代 --group)
+  - 翻页：hasMore=true 时，用返回的 nextCursor 作为下次 --cursor
+```
+
+---
+
+## message info — 获取会话信息
+
+获取指定群聊或单聊会话的详情信息。
+
+```
+Usage:
+  dws chat message info [flags]
+Example:
+  dws chat message info --id <openConversationId>
+  dws chat message info --open-id <openDingTalkId>
+Flags:
+      --id string        群聊会话 ID openConversationId（与 --open-id 二选一）
+      --open-id string   用户 openDingTalkId（单聊时与 --id 二选一）
+
+注意:
+  - --id（群聊）和 --open-id（单聊用户 openDingTalkId）二选一
+```
+
+---
+
+## list-top-conversations — 拉取置顶会话列表
+
+拉取当前用户的置顶会话列表。
+
+```
+Usage:
+  dws chat list-top-conversations [flags]
+Example:
+  dws chat list-top-conversations --limit 1000
+  dws chat list-top-conversations --limit 1000 --cursor <nextCursor>
+Flags:
+      --limit int        每页返回数量（默认 1000）
+      --cursor int       分页游标（首次不传或传 0，翻页传 nextCursor）
+
+注意:
+  - 用户询问"置顶会话"时，直接调用此命令返回置顶会话列表即可
+  - 用户询问"置顶消息"时，需两步：先调用此命令拉取置顶会话列表获取各会话的 openConversationId，再用 `chat message list --group <openConversationId>` 分别拉取每个会话内的消息
+  - 翻页：hasMore=true 时，用返回的 nextCursor 作为下次 --cursor
 ```

 ---
@@ -126,161 +505,307 @@ Flags:
 Usage:
  dws chat bot search [flags]
 Example:
-  dws chat bot search --name "考勤" --format json
+  dws chat bot search --page 1
+  dws chat bot search --page 1 --size 10 --name "日报"
 Flags:
-      --name string   机器人名称（模糊搜索）
-      --page int      页码（默认 1）
-      --size int      每页数量（默认 50）
+      --name string   按名称搜索
+      --page int      页码，从1开始 (默认 1)
+      --size int      每页条数 (默认 50)，别名: --limit
+```
+
+---
+
+## bot create — 创建企业机器人
+
+```
+Usage:
+  dws chat bot create [flags]
+Example:
+  dws chat bot create --name "日报提醒机器人" --desc "负责每日日报提醒"
+Flags:
+      --name string   机器人名称 (必填)
+      --desc string   机器人描述（可选）
+```
+
+---
+
+## bot search-groups — 搜索机器人所在群
+
+搜索指定机器人已加入的群列表。
+
+```
+Usage:
+  dws chat bot search-groups [flags]
+Example:
+  dws chat bot search-groups --keyword "项目"
+  dws chat bot search-groups --keyword "冲刺" --cursor <nextCursor>
+Flags:
+      --keyword string   搜索关键词 (必填)
+      --cursor string    分页游标（首页留空，翻页传返回的 cursor）
 ```

 ---

 ## message send-by-bot — 机器人发消息

-支持两种模式：群聊发送 和 批量单聊发送，通过 `--group` 和 `--users` 互斥区分。
+群聊：传 --group 指定群；单聊：传 --users 指定用户列表，二者只能选其一，不能同时指定。--text 支持 Markdown。

-### 群聊发送
 ```
 Usage:
  dws chat message send-by-bot [flags]
 Example:
-  dws chat message send-by-bot --robot-code <code> --group <openConversationId> \
-    --title "日报提醒" --text "请提交今日日报" --format json
+  dws chat message send-by-bot --robot-code <robot-code> --group <openconversation_id> --title "日报" --text "## 今日完成..."
+  dws chat message send-by-bot --robot-code <robot-code> --users userId1,userId2 --title "提醒" --text "请提交周报"
 Flags:
-      --robot-code string   机器人 code (必填)
-      --group string        群会话 ID (必填，与 --users 互斥)
+      --group string        群聊 openConversationId（群聊时必填）
+      --robot-code string   机器人 Code (必填)
+      --text string         消息内容 Markdown (必填)
      --title string        消息标题 (必填)
-      --text string         消息内容，支持 Markdown (必填)
-```
+      --users string        用户 userId 列表，逗号分隔，最多20个（单聊时必填）

-### 批量单聊发送
+注意:
+  - --group 与 --users 互斥，必须且只能指定其一
+  - --group 的别名: --id, --chat, --conversation-id (均可替代 --group)
 ```
-Usage:
-  dws chat message send-by-bot [flags]
-Example:
-  dws chat message send-by-bot --robot-code <code> --users "user1,user2" \
-    --title "通知" --text "会议已取消" --format json
-Flags:
-      --robot-code string   机器人 code (必填)
-      --users string        用户 ID 列表，逗号分隔，最多 20 个 (必填，与 --group 互斥)
-      --title string        消息标题 (必填)
-      --text string         消息内容，支持 Markdown (必填)
-```
-
-> ⚠️ `--group` 和 `--users` 互斥：群聊用 `--group`，单聊用 `--users`，不能同时传。

 ---

 ## message recall-by-bot — 机器人撤回消息

-支持两种模式：群聊撤回 和 批量单聊撤回。
+群聊：传 --group 与 --keys；单聊：仅传 --keys。--keys 为发送时返回的 processQueryKey 列表，逗号分隔。

-### 群聊撤回
 ```
 Usage:
  dws chat message recall-by-bot [flags]
 Example:
-  dws chat message recall-by-bot --robot-code <code> --group <openConversationId> \
-    --keys "key1,key2" --format json
+  dws chat message recall-by-bot --robot-code <robot-code> --group <openconversation_id> --keys <process-query-key>
+  dws chat message recall-by-bot --robot-code <robot-code> --keys key1,key2
 Flags:
-      --robot-code string   机器人 code (必填)
-      --group string        群会话 ID (必填，与批量单聊互斥)
-      --keys string         消息 key 列表，逗号分隔 (必填)
+      --group string        群聊 openConversationId（群聊撤回时必填）
+      --keys string         消息 processQueryKey 列表，逗号分隔 (必填)
+      --robot-code string   机器人 Code (必填)
 ```

-### 批量单聊撤回
-```
-Usage:
-  dws chat message recall-by-bot [flags]
-Example:
-  dws chat message recall-by-bot --robot-code <code> --keys "key1,key2" --format json
-Flags:
-      --robot-code string   机器人 code (必填)
-      --keys string         消息 key 列表，逗号分隔 (必填)
-```
-
-> ⚠️ 消息 key 从 `send-by-bot` 返回结果中提取。
-
 ---

 ## message send-by-webhook — 自定义机器人 Webhook 发消息

+@ 人时需在 --text 中包含 @userId 或 @手机号，否则 @ 不生效。
+
 ```
 Usage:
  dws chat message send-by-webhook [flags]
 Example:
-  dws chat message send-by-webhook --token <robotToken> \
-    --title "告警" --text "CPU 使用率超过 90%" --format json
+  dws chat message send-by-webhook --token <webhook-token> --title "告警" --text "CPU 超 90%" --at-all
+  dws chat message send-by-webhook --token <webhook-token> --title "test" --text "hi @118785" --at-users 118785
 Flags:
-      --token string        自定义机器人 Webhook Token (必填)
-      --title string        消息标题 (必填)
+      --at-all              @ 所有人
+      --at-mobiles string   @ 指定手机号，逗号分隔
+      --at-users string     @ 指定用户，逗号分隔（需在 text 中包含 @userId）
      --text string         消息内容 (必填)
-      --at-all              @所有人
-      --at-mobiles string   @指定手机号列表，逗号分隔
-      --at-users string     @指定用户 ID 列表，逗号分隔
+      --title string        消息标题 (必填)
+      --token string        Webhook Token (必填)
 ```

 ---

 ## 意图判断

- 用户说"搜索一个群" → `search`
- 用户说"帮我建个群" → `group create`
- 用户说"看看群里有谁" → `group members list`
- 用户说"把张三拉进群" → 先 `contact user search` 获取 userId，再 `group members add`
- 用户说"把张三移出群" → 先 `contact user search` 获取 userId，再 `group members remove`（⚠️ 需确认）
- 用户说"改一下群名" → `group rename`
- 用户说"让机器人在群里发通知" → `message send-by-bot --group`
- 用户说"机器人给张三发消息" → 先 `contact user search` 获取 userId，再 `message send-by-bot --users`
- 用户说"通过 Webhook 发告警" / 用户有 Webhook Token → `message send-by-webhook`
- 用户说"撤回机器人消息" → `message recall-by-bot`
- 用户说"查一下我的机器人" → `bot search`
- 用户说"把机器人加到群里" → `group members add-bot`
+用户说"建群/创建群聊" → `chat group create`
+用户说"创建企业全员群/组织群" → `chat group create-org`
+用户说"搜索群/找群" → `chat search`
+用户说"群成员/看群里有谁" → `chat group members list`
+用户说"拉人进群/加群成员" → `chat group members add`
+用户说"踢人/移除群成员" → `chat group members remove`
+用户说"加机器人到群" → `chat group members add-bot`
+用户说"改群名" → `chat group rename`
+用户说"聊天记录/会话消息/拉取会话" → `chat message list`
+用户说"某人发给我的消息/指定发送者/某人的消息" → `chat message list-by-sender`（用户未明确说"单聊"时优先使用，跨单聊/群聊）
+用户说"拉取和某人的单聊记录/单聊消息" → `chat message list --user`（用户明确说"单聊"时使用）
+用户说"@我的消息/at我的/提及我的" → `chat message list-mentions`
+用户说"未读消息会话/未读会话列表/我的未读会话" → `chat message list-unread-conversations`
+用户说"发群消息(以个人身份)" → `chat message send --group`
+用户说"发单聊消息(以个人身份)" → `chat message send --user`（有 userId 时）或 `chat message send --open-dingtalk-id`（有 openDingTalkId 时）
+用户说"发个人消息/个人通知" → `chat message send-personal`（⚠️ 敏感操作，需确认）
+用户说"机器人发消息/机器人群发" → `chat message send-by-bot`
+用户说"机器人撤回消息" → `chat message recall-by-bot`
+用户说"Webhook 发消息/告警消息" → `chat message send-by-webhook`
+用户说"话题回复/群话题消息回复/拉取话题回复" → `chat message list-topic-replies`
+用户说"所有消息/全部会话消息/拉取全部消息/时间范围内消息/我的消息/我今天的消息/查我的钉钉消息/最近的消息" → `chat message list-all`
+用户说"特别关注人的消息/关注的人的消息/星标联系人的消息" → `chat message list-focused`
+用户说"查看我的机器人" → `chat bot search`
+用户说"创建机器人" → `chat bot create`
+用户说"搜索消息/查找关键词/搜一下消息里的XX" → `chat message search`
+用户说"我和XX的共同群/我们都在哪些群/查共同群" → `chat search-common`
+用户说"置顶会话/置顶消息/我的置顶/查看置顶" → `chat list-top-conversations`
+用户说"获取会话信息/会话详情" → `chat message info`
+用户说"机器人在哪些群/机器人的群" → `chat bot search-groups`

-**关键区分**: `send-by-bot`(企业内部机器人，需 robotCode) vs `send-by-webhook`(自定义机器人 Webhook，需 token)
+关键区分:
+- `chat message list` — 拉取指定会话的消息（需指定 --group 或 --user），按时间点 + 方向翻页
+- `chat message list --user` — list 的单聊模式，拉取与指定用户的单聊记录（用户明确说"单聊""私聊"时使用）
+- `chat message list-by-sender` — 搜索指定发送者发给我的消息，跨所有会话（单聊+群聊均包含，用户只说"某人发的消息"时优先使用）
+- `chat message list-mentions` — 拉取 @我 的消息（跨单聊/群聊，可选指定群）
+- `chat message list-unread-conversations` — 拉取当前用户存在未读消息的会话列表（可选 `--count`）
+- `chat message list-all` — 拉取当前用户所有会话的消息，按时间范围 + cursor 分页。只要用户没有指定某个具体的会话（如某个群名、某个人名），即使提到"单聊消息""群聊消息"等笼统范围，也应路由到此命令
+- `chat message list-topic-replies` — 拉取群话题的回复消息列表
+- `chat message list-focused` — 拉取特别关注人的消息，cursor 分页
+- `chat list-top-conversations` — 拉取置顶会话列表（用户询问"置顶会话"或"置顶消息"时路由到此），cursor 分页
+- `chat message send` — 以**当前用户**身份发消息（群聊或单聊），text 为位置参数；支持 --media-id 发送图片消息
+- `chat message send-personal` — 发送个人消息，支持通过 openConversationId 或 openDingTalkId 指定目标（⚠️ 敏感操作）
+- `chat message search` — 按关键词搜索消息内容（跨所有会话，可选指定群）
+- `chat search-common` — 搜索共同群，查询指定人共同所在的群聊（AND=所有人都在，OR=任一人在）
+- `chat message send-by-bot` — 以**机器人**身份发消息（群聊或单聊），text 为 --text flag
+- `chat message send-by-webhook` — 通过**自定义机器人 Webhook** 发群消息
+- `chat message recall-by-bot` — 通过机器人撤回已发送的消息
+- `chat message info` — 获取指定会话的详情信息
+- `chat bot create` — 创建新的企业机器人
+- `chat bot search-groups` — 搜索机器人所在群列表

 ## 核心工作流

 ```bash
-# ── 工作流: 建群并添加机器人 ──
+# 1. 搜索群 — 提取 openconversation_id
+dws chat search --query "项目冲刺" --format json

-# 1. 搜索同事 userId
-dws contact user search --keyword "张三" --format json
+# 2. 拉取群消息
+dws chat message list --group <openconversation_id> --time "2025-03-01 00:00:00" --format json

-# 2. 创建群
-dws chat group create --name "项目群" --users <userId1>,<userId2> --format json
+# 2b. 拉取未读会话列表
+dws chat message list-unread-conversations --count 20 --format json

-# 3. 搜索机器人
-dws chat bot search --format json
+# 3. 以个人身份发送群消息
+dws chat message send --group <openconversation_id> --title "周报提醒" "请大家本周五前提交周报" --format json

-# 4. 添加机器人到群
-dws chat group members add-bot --id <openConversationId> --robot-code <code> --format json
+# 4. 以个人身份单聊（通过 userId）
+dws chat message send --user <userId> "你好" --format json
+
+# 4b. 以个人身份单聊（通过 openDingTalkId，三方应用等无法获取 userId 时使用）
+dws chat message send --open-dingtalk-id <openDingTalkId> "你好" --format json
+
+# 5. 机器人发群消息（Markdown）
+dws chat message send-by-bot --robot-code <robot-code> \
+  --group <openconversation_id> --title "日报" --text "## 今日完成..." --format json
+
+# 6. 机器人单聊发消息
+dws chat message send-by-bot --robot-code <robot-code> \
+  --users userId1,userId2 --title "提醒" --text "请提交周报" --format json
+
+# 7. Webhook 发告警
+dws chat message send-by-webhook --token <webhook-token> \
+  --title "告警" --text "CPU 超 90%" --at-all --format json
 ```

-```bash
-# ── 工作流: 机器人群发消息 ──
+## 复合工作流

-# 1. 搜索可用机器人
+### 机器人发消息后撤回（完整流程）
+
+撤回只能用于 `send-by-bot` 发出的消息。个人身份 (`chat message send`) 发出的消息**无法通过 API 撤回**。
+
+```bash
+# Step 1: 查我的机器人 — 提取 robot-code
 dws chat bot search --format json

-# 2. 发送群消息
-dws chat message send-by-bot --robot-code <code> --group <groupId> \
+# Step 2: 用机器人发消息 — 提取返回中的 processQueryKey
+dws chat message send-by-bot --robot-code <robot-code> --group <openconversation_id> \
  --title "通知" --text "内容" --format json
+
+# Step 3: 用同一个 robot-code + processQueryKey 撤回
+dws chat message recall-by-bot --robot-code <robot-code> --group <openconversation_id> \
+  --keys <processQueryKey> --format json
 ```

-```bash
-# ── 工作流: Webhook 告警 ──
+### 创建并使用机器人（完整流程）

-# 直接通过 Webhook Token 发送
-dws chat message send-by-webhook --token <token> \
-  --title "告警" --text "服务异常" --at-all --format json
+```bash
+# Step 1: 创建机器人
+dws chat bot create --name "项目提醒机器人" --desc "项目状态提醒" --format json
+
+# Step 2: 搜索群 — 提取 openConversationId
+dws chat search --query "项目群" --format json
+
+# Step 3: 将机器人添加到群
+dws chat group members add-bot --id <openConversationId> --robot-code <robotCode> --format json
+
+# Step 4: 机器人发消息
+dws chat message send-by-bot --robot-code <robotCode> --group <openConversationId> \
+  --title "提醒" --text "请及时更新项目状态" --format json
+```
+
+### 机器人 @指定人发群消息
+
+`--text` 中**必须**包含 `<@userId>` 占位符，否则 @ 不生效。
+
+```bash
+# Step 1: 搜人获取 userId
+dws aisearch person --keyword "张三" --dimension name --format json
+
+# Step 2: 取 userId 发送（注意 text 中的占位符）
+dws chat message send-by-bot --robot-code <robot-code> --group <openconversation_id> \
+  --title "提醒" --text "<@userId1> <@userId2> 请查收本周报告" --format json
+```
+
+### 发送图片/文件消息（跨产品: drive → chat）
+
+```bash
+# Step 1: 上传文件到钉盘 — 获取 uploadId 和凭证
+dws drive upload-info --file-name "截图.png" --file-size <字节数> --format json
+
+# Step 2: HTTP PUT 上传文件到 OSS
+curl -X PUT -T "截图.png" "<upload-info 返回的上传 URL>"
+
+# Step 3: 提交上传 — 获取 dentryUuid
+dws drive commit --file-name "截图.png" --file-size <字节数> --upload-id <uploadId> --format json
+
+# Step 4: 获取下载链接
+dws drive download --file-id <dentryUuid> --format json
+
+# Step 5: 用 Markdown 图片语法发送
+dws chat message send --group <openconversation_id> \
+  --text "![截图](下载链接)" --format json
 ```

 ## 上下文传递表

 | 操作 | 从返回中提取 | 用于 |
 |------|-------------|------|
-| `search` | openConversationId | `group members` / `group rename` / `group members add` / `send-by-bot --group` |
-| `group create` | openConversationId | 同上 |
-| `bot search` | robotCode | `send-by-bot` / `recall-by-bot` / `add-bot` |
-| `message send-by-bot` | processQueryKey | `recall-by-bot --keys` |
+| `chat search` | `openConversationId` | message send/list、group members 等的 --group |
+| `chat group create` | `openConversationId` | 同上 |
+| `chat message list-all` | `nextCursor` | 下次 list-all 的 --cursor |
+| `aisearch person` | `userId` | message send 的 --user、--at-users、send-by-bot 的 --users、list-by-sender 的 --sender-user-id |
+| `aisearch person` → `contact user get` | `openDingTalkId` | list-by-sender 的 --sender-open-dingtalk-id、message send/list 的 --open-dingtalk-id |
+| `chat bot search` | `robotCode` | send-by-bot / recall-by-bot 的 --robot-code、group members add-bot 的 --robot-code |
+| `chat bot create` | `robotCode` | send-by-bot / recall-by-bot 的 --robot-code |
+| `chat message send-by-bot` | `processQueryKey` | recall-by-bot 的 --keys |
+| `chat message search` | `nextCursor` | 下次 message search 的 --cursor |
+| `chat search-common` | `openConversationId` | message send/list 等的 --group |
+| `drive download` | 下载链接 | message send 的 Markdown 图片/链接语法 |
+
+## 注意事项
+
+- `--group` 为群聊会话 ID (openconversation_id)，可从群搜索或群聊信息中获取
+- `chat message send` 的 text 是位置参数（恰好 1 个），非 flag；群聊用 `--group`，单聊用 `--user`（userId）或 `--open-dingtalk-id`（openDingTalkId），三者互斥；`--at-all`、`--at-users` 仅在 `--group` 群聊时生效；发送图片消息用 `--media-id`
+- `chat message send-personal` 为敏感操作（isSensitive），执行前需用户明确确认
+- `chat message list-all` 的四个参数（--start、--end、--limit、--cursor）每次请求都必须传递；翻页时用响应中的 nextCursor 值作为下次 --cursor
+- `chat message list` 的 `--group`、`--user`、`--open-dingtalk-id` 三者互斥，必须且只能指定其一
+- `chat message list-by-sender` 不需要指定单聊/群聊，返回结果自带会话类型标识
+- `chat message list-mentions` 可选 `--group` 指定群聊，不传则查全部
+- `chat message list-unread-conversations` 获取当前用户未读会话列表，可选 `--count` 指定返回条数
+- `chat message search` 按关键词搜索消息内容，`--keyword` 必填，可选 `--group` 限定搜索某个会话
+- `chat search-common` 搜索共同群，`--nicks` 传人员昵称（逗号分隔），`--match-mode` AND/OR 控制匹配逻辑
+- `chat list-top-conversations` 拉取置顶会话列表，分页用 `--limit`（默认 1000）/`--cursor`
+- `send-by-bot` 群聊传 `--group`，单聊传 `--users`，二者互斥且必选其一
+- `recall-by-bot` 群聊传 `--group` + `--keys`，单聊仅传 `--keys`（不传 `--group` 即为单聊撤回）
+- `send-by-webhook` 支持 `--at-all`、`--at-mobiles`、`--at-users` 进行 @ 操作，但需在 `--text` 中包含 `@userId` 或 `@手机号` 才能生效
+
+## 自动化脚本
+
+| 脚本 | 场景 | 用法 |
+|------|------|------|
+| [chat_export_messages.py](../../scripts/chat_export_messages.py) | 导出群聊消息到 JSON 文件 | `python chat_export_messages.py --query "项目冲刺" --time "2026-03-10 00:00:00"` |
+| [chat_history_with_user.py](../../scripts/chat_history_with_user.py) | 查询与某人的单聊聊天记录 | `python chat_history_with_user.py --name "张三" --time "2026-03-10 00:00:00"` |
+
+## 相关产品
+
+- [contact](./contact.md) — 搜索同事/好友，获取 userId 用于 --user、--at-users、send-by-bot --users、list-by-sender --sender-user-id；获取 openDingTalkId 用于 list-by-sender --sender-open-dingtalk-id、--open-dingtalk-id
+- [drive](./drive.md) — 上传文件获取下载链接，用于 Markdown 图片/文件消息
Author	SHA1	Message	Date
修雨	f19a3ccfa5	feat(skills): 同步 chat/bot skill 参考 — 补齐消息管理与机器人新能力 (#148 ) 对齐 dws-service-endpoints.json 新增的 16 个 group-chat toolOverrides 和 2 个 bot toolOverrides，参考 dws-wukong 实现补齐以下命令文档： - message: send/list/list-all/list-topic-replies/list-by-sender/ list-mentions/list-focused/list-unread-conversations/ search/info/send-personal/list-top-conversations - group: create-org/search-common - bot: create/search-groups 同步更新命令总览、意图判断、工作流与上下文传递表。	2026-04-22 16:31:56 +08:00
xuan	89c5038446	fix(auth): restore device_code polling when flowId is absent (#145 ) * fix(auth): fallback to device_code when flowId is missing * chore(auth): remove dead helper and harden fallback tests	2026-04-22 15:13:40 +08:00
修雨	4f915e4e2c	refactor(plugin): drop managed/ compat layer left by PR #126 (#135 ) PR #126 removed the privileged managed-plugin mechanism but kept LoadManaged, IsManaged, and ~/.dws/plugins/managed/ fallbacks so that plugins installed by pre-#126 builds would keep loading. The window for that migration was ~4 days (2026-04-15 → 2026-04-19), and the original install path mostly failed anyway (issue #124 / GitHub Pages HTML response), so there is no real user base to preserve. Scope of deletion: - plugin.Loader.LoadManaged, scanDir's isManaged param, loadPlugin's isManaged param. LoadAll now combines user + dev only. - Plugin.IsManaged field. Converter/root.go no longer branch on it; stdio and HTTP descriptors both report source="plugin". - ~/.dws/plugins/managed/ branches in ListInstalled, SetEnabled, RemovePlugin. - config.PluginManagedDir constant. - TestLoaderLoadManaged, TestRemoveLegacyManagedPlugin, and the "legacy managed" table case in TestRemovePluginPurgesSettings. Kept on purpose: - Manifest.Type public schema field ("managed"\|"user") — it appears in existing plugin.json files and is only validated, never behaviorally consumed. Users with an orphaned ~/.dws/plugins/managed/ directory can `rm -rf ~/.dws/plugins/managed/` — the CLI no longer reads it.	2026-04-22 14:56:11 +08:00
Ari	ec03b7cca3	docs(changelog): backfill 1.0.3 through 1.0.12 release notes (#146 ) CHANGELOG.md had been stuck at 1.0.2; releases 1.0.3, 1.0.4, 1.0.6, 1.0.7, 1.0.8, 1.0.9, 1.0.11, and 1.0.12 are now documented in full. Versions 1.0.5 and 1.0.10 were not shipped as git tags and are intentionally skipped. Highlights: - 1.0.3: nested/array output filtering, schema table rendering, todo static command family - 1.0.4: client credential keychain persistence for token refresh - 1.0.6: error diagnostics overhaul (trace_id/server_code/verbosity levels), destructive-command confirmation, OAuth credential persistence - 1.0.7: dws upgrade self-updater, Wukong edition overlay, auth check flipped to fail-closed - 1.0.8: aitable static helper commands (20 commands) - 1.0.9: plugin system launch, dws doctor/config, 3->5 stage command pipeline, edition hooks - 1.0.11: plugin cold-start perf, PAT scope error UX, i18n fixes - 1.0.12: doc + minutes skill references, aitable dashboard/chart/export sections, README catalog sync to 141 commands / 14 products	2026-04-22 12:19:22 +08:00
fantiuandgithub-actions[bot]	e0544579d2	fix(auth): attribute CLI auth denial to channel before user scope (#144 ) * chore: update coverage badge [skip ci] * chore: update coverage badge [skip ci] * fix(auth): attribute CLI auth denial to channel before user scope --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>	2026-04-22 12:05:30 +08:00