diff --git a/AGENTS.md b/AGENTS.md
new file mode 100644
index 0000000..bbcfd78
--- /dev/null
+++ b/AGENTS.md
@@ -0,0 +1,749 @@
+# AGENTS.md — Feishu Integration Master Agent Reference
+
+**Audience**: AI agents (Claude Code, OpenClaw, Cursor, etc.)
+**Purpose**: Complete reference for all Feishu operations — tool inventory, decision tree, auth verification, command examples.
+**Quick reference**: Consult SKILL.md for routing decisions. This file is the full manual.
+
+All instructions in this file are imperative. Execute them literally.
+
+---
+
+## Architecture Overview
+
+The Feishu integration uses a two-layer architecture:
+
+```
+User <-> AI Agent <-> feishu-mcp Integration Layer
+ |
+ +---------+---------+
+ | |
+ Feishu CLI Community MCP
+ (@larksuite/cli) (feishu-mcp)
+ 200+ commands 15 doc editing tools
+ 19 AI skills Block-level precision
+ 11 business domains
+ OAuth port: 3000 OAuth port: 3333
+```
+
+- **Feishu CLI** (`@larksuite/cli`): Breadth layer. 19 AI skills covering 11 business domains with 200+ commands. Auth stored in OS keychain. OAuth callback on port 3000. Shortcut syntax uses `+` prefix (e.g., `+agenda`, `+send`, `+create`).
+- **Community MCP** (`feishu-mcp`): Depth layer. 15 document editing tools with block-level precision, batch operations, and formatting control (8-color palette, alignment). Auth via env vars (`FEISHU_APP_ID`, `FEISHU_APP_SECRET`). OAuth callback on port 3333.
+- **Official MCP** (`@larksuiteoapi/lark-mcp`): Optional. Advanced bitable, doc search, and wiki tools. Not installed by default.
+
+Key principle: Use CLI for breadth (anything beyond documents). Use MCP for depth (precise block manipulation, formatting, batch operations).
+
+---
+
+## Complete Tool Inventory
+
+### CLI Skills (19 total, 11 business domains)
+
+#### lark-shared — 认证与身份
+应用配置初始化、认证登录、身份切换、权限管理。
+```bash
+lark-cli auth status # Check current auth state
+lark-cli auth login --recommend # OAuth browser login
+lark-cli config init --new # Initialize app configuration
+```
+
+#### lark-calendar — 日历
+日历与日程管理：查看/搜索日程、创建/更新日程、管理参会人、查询忙闲状态。
+```bash
+lark-cli calendar +agenda # View today's schedule
+lark-cli calendar +create # Create calendar event
+lark-cli calendar +freebusy # Check free/busy status
+lark-cli calendar +suggestion # Get meeting time suggestions
+```
+
+#### lark-im — 即时通讯
+收发消息、搜索聊天记录、管理群聊。
+```bash
+lark-cli im +send # Send message
+lark-cli im +search # Search chat history
+```
+
+#### lark-doc — 文档
+从 Markdown 创建文档、获取内容、更新、搜索。
+```bash
+lark-cli doc +create # Create document from Markdown
+lark-cli doc +search # Search documents in cloud space
+lark-cli docs +update # Update document (append/overwrite/replace/insert/delete)
+lark-cli docs +media-insert # Insert image/file into document
+```
+
+#### lark-drive — 云空间
+上传/下载文件、管理文件夹、管理权限和评论。
+```bash
+lark-cli drive upload # Upload file to cloud space
+lark-cli drive download # Download file from cloud space
+```
+
+#### lark-sheets — 电子表格
+创建、读写、搜索、导出电子表格。
+```bash
+lark-cli sheets +create # Create spreadsheet with headers
+lark-cli sheets +write # Write cell data
+lark-cli sheets +read # Read cell range
+```
+
+#### lark-base — 多维表格
+表、字段、记录、视图、仪表盘管理。
+```bash
+lark-cli base +query # Query records (filter/sort)
+lark-cli base +create # Create bitable
+```
+
+#### lark-task — 任务
+创建/更新任务、子任务、清单、分配成员。
+```bash
+lark-cli task +create # Create task
+lark-cli task +get-my-tasks # Get my task list
+```
+
+#### lark-mail — 邮件
+起草、发送、回复、转发、搜索邮件。
+```bash
+lark-cli mail +compose # Compose and send email
+lark-cli mail +inbox # View inbox
+lark-cli mail +search # Search emails
+```
+
+#### lark-contact — 通讯录
+查询组织架构、搜索员工信息。
+```bash
+lark-cli contact +search # Search contacts (name/email/phone)
+```
+
+#### lark-wiki — 知识库
+管理知识空间、文档节点、层级结构。
+```bash
+lark-cli wiki +list # List knowledge spaces
+lark-cli wiki +create # Create wiki document
+```
+
+#### lark-event — 事件订阅
+WebSocket 实时监听飞书事件。
+```bash
+lark-cli event # Subscribe to events (NDJSON output)
+```
+
+#### lark-vc — 视频会议
+查询会议记录、获取会议纪要。
+```bash
+lark-cli vc # Query meeting records
+```
+
+#### lark-minutes — 妙记
+获取妙记基础信息和 AI 产物。
+```bash
+lark-cli minutes # Get minutes info (summary, todos, chapters)
+```
+
+#### lark-whiteboard — 画板
+Mermaid 图表、流程图、架构图、思维导图。
+```bash
+lark-cli whiteboard +create # Create new whiteboard
+lark-cli docs +whiteboard-update # Edit whiteboard content
+```
+
+#### lark-openapi-explorer — API 探索
+从官方文档库挖掘未封装的原生 OpenAPI 接口。
+```bash
+lark-cli openapi-explorer # Explore unwrapped OpenAPI endpoints
+```
+
+#### lark-skill-maker — 技能创建
+创建自定义 lark-cli Skill。
+```bash
+lark-cli skill-maker # Create custom CLI skill
+```
+
+#### lark-workflow-meeting-summary — 会议纪要工作流
+汇总会议纪要并生成结构化报告。
+```bash
+lark-cli workflow-meeting-summary # Aggregate meeting minutes into report
+```
+
+#### lark-workflow-standup-report — 站会报告工作流
+生成日程 + 未完成任务摘要。
+```bash
+lark-cli workflow-standup-report # Generate schedule + task summary
+```
+
+---
+
+### Community MCP Tools (17 total)
+
+Each tool is invoked via direct tool call or mcporter:
+
+#### create_feishu_document
+Create a document in a folder.
+- Parameters: `title` (string, required), `folderToken` (string, optional)
+```bash
+mcporter call feishu.create_feishu_document --args '{"title":"新文档"}'
+```
+
+#### get_feishu_document_info
+Get document metadata (title, revision, create/update time).
+- Parameters: `document_id` (string, required)
+```bash
+mcporter call feishu.get_feishu_document_info --args '{"document_id":"doxcnXXX"}'
+```
+
+#### get_feishu_document_blocks
+Get the block structure tree of a document.
+- Parameters: `document_id` (string, required)
+```bash
+mcporter call feishu.get_feishu_document_blocks --args '{"document_id":"doxcnXXX"}'
+```
+
+#### get_feishu_document_text
+Extract plain text from a document.
+- Parameters: `document_id` (string, required)
+```bash
+mcporter call feishu.get_feishu_document_text --args '{"document_id":"doxcnXXX"}'
+```
+
+#### batch_create_feishu_blocks
+Batch create blocks (text, code, heading, list) atomically.
+- Parameters: `document_id` (string, required), `blocks` (array, required)
+```bash
+mcporter call feishu.batch_create_feishu_blocks --args '{"document_id":"doxcnXXX","blocks":[{"type":"heading2","content":"标题"},{"type":"text","content":"正文"}]}'
+```
+
+#### update_feishu_block_text
+Update text/code/heading block with formatting (8 colors, alignment).
+- Parameters: `document_id` (string, required), `block_id` (string, required), `content` (string, required), `text_color` (string, optional), `align` (string, optional)
+```bash
+mcporter call feishu.update_feishu_block_text --args '{"document_id":"doxcnXXX","block_id":"blkXXX","content":"更新内容","text_color":"red"}'
+```
+
+#### delete_feishu_document_blocks
+Delete blocks by block_id.
+- Parameters: `document_id` (string, required), `block_ids` (array, required)
+```bash
+mcporter call feishu.delete_feishu_document_blocks --args '{"document_id":"doxcnXXX","block_ids":["blkXXX"]}'
+```
+
+#### upload_and_bind_image_to_block
+Upload an image and bind it to a block in one step.
+- Parameters: `document_id` (string, required), `image_path` or `image_url` (string, required)
+```bash
+mcporter call feishu.upload_and_bind_image_to_block --args '{"document_id":"doxcnXXX","image_url":"https://example.com/img.png"}'
+```
+
+#### create_feishu_table
+Create a table in a document.
+- Parameters: `document_id` (string, required), `rows` (number, required), `cols` (number, required)
+```bash
+mcporter call feishu.create_feishu_table --args '{"document_id":"doxcnXXX","rows":3,"cols":3}'
+```
+
+#### create_feishu_folder
+Create a folder in cloud space.
+- Parameters: `name` (string, required), `folderToken` (string, optional)
+```bash
+mcporter call feishu.create_feishu_folder --args '{"name":"项目文档"}'
+```
+
+#### list_feishu_folder_contents
+List contents of a folder.
+- Parameters: `folderToken` (string, required)
+```bash
+mcporter call feishu.list_feishu_folder_contents --args '{"folderToken":"fldcnXXX"}'
+```
+
+#### get_feishu_folder_files
+Get file listing from a folder.
+- Parameters: `folderToken` (string, required)
+```bash
+mcporter call feishu.get_feishu_folder_files --args '{"folderToken":"fldcnXXX"}'
+```
+
+#### search_feishu_documents
+Search documents by keyword.
+- Parameters: `query` (string, required), `count` (number, optional)
+```bash
+mcporter call feishu.search_feishu_documents --args '{"query":"会议记录","count":10}'
+```
+
+#### get_feishu_wiki_node_info
+Get wiki node metadata.
+- Parameters: `token` (string, required)
+```bash
+mcporter call feishu.get_feishu_wiki_node_info --args '{"token":"wikcnXXX"}'
+```
+
+#### get_feishu_whiteboard_content
+Read whiteboard content (read-only).
+- Parameters: `whiteboard_id` (string, required)
+```bash
+mcporter call feishu.get_feishu_whiteboard_content --args '{"whiteboard_id":"wbXXX"}'
+```
+
+---
+
+### Official MCP Tools (Optional)
+
+If `@larksuiteoapi/lark-mcp` is installed, these additional tools become available. This layer is optional and not part of the default installation.
+
+| Tool | Description | Example |
+|---|---|---|
+| `bitable_v1_app_create` | Create bitable app | `mcporter call lark.bitable_v1_app_create --args '{"data":{"name":"表"}}'` |
+| `bitable_v1_appTableRecord_create` | Create bitable record | `mcporter call lark.bitable_v1_appTableRecord_create --args '...'` |
+| `bitable_v1_appTableRecord_search` | Search bitable records | `mcporter call lark.bitable_v1_appTableRecord_search --args '...'` |
+| `docx_builtin_search` | Search cloud documents | `mcporter call lark-user.docx_builtin_search --args '{"data":{"search_key":"..."}}'` |
+| `docx_v1_document_rawContent` | Get document raw text | `mcporter call lark-user.docx_v1_document_rawContent --args '...'` |
+| `wiki_v1_node_search` | Search wiki nodes | `mcporter call lark-user.wiki_v1_node_search --args '...'` |
+| `wiki_v2_space_getNode` | Get wiki space node | `mcporter call lark-user.wiki_v2_space_getNode --args '...'` |
+
+---
+
+## Decision Tree
+
+Follow this tree step-by-step. At each numbered item, evaluate the condition and follow the matching branch. Every path ends at a specific tool/command.
+
+1. **User wants to send or read messages**
+ 1. Send message → `lark-cli im +send`
+ 2. Search chat history → `lark-cli im +search`
+
+2. **User wants to create a document**
+ 1. Has Markdown content or simple creation → `lark-cli doc +create`
+ 2. Needs to specify target folder by token → `create_feishu_document`
+ 3. Needs batch formatted blocks immediately after creation → `lark-cli doc +create` then `batch_create_feishu_blocks`
+
+3. **User wants to edit an existing document**
+ 1. Simple append/overwrite/replace → `lark-cli docs +update`
+ 2. Needs specific block color or alignment → `update_feishu_block_text`
+ 3. Needs batch creation of multiple formatted blocks → `batch_create_feishu_blocks`
+ 4. Needs to delete specific blocks by ID → `delete_feishu_document_blocks`
+ 5. Needs to insert image → `lark-cli docs +media-insert`
+ 6. Needs to insert table → `create_feishu_table`
+
+4. **User wants to search for something**
+ 1. Search cloud documents → `lark-cli doc +search`
+ 2. Search chat messages → `lark-cli im +search`
+ 3. Search emails → `lark-cli mail +search`
+ 4. Search contacts → `lark-cli contact +search`
+
+5. **User wants calendar/scheduling operations**
+ 1. View schedule → `lark-cli calendar +agenda`
+ 2. Create event → `lark-cli calendar +create`
+ 3. Check free/busy → `lark-cli calendar +freebusy`
+ 4. Get time suggestions → `lark-cli calendar +suggestion`
+
+6. **User wants to manage tasks**
+ 1. Create task → `lark-cli task +create`
+ 2. View my tasks → `lark-cli task +get-my-tasks`
+
+7. **User wants to send email**
+ 1. Compose/send → `lark-cli mail +compose`
+ 2. View inbox → `lark-cli mail +inbox`
+ 3. Search → `lark-cli mail +search`
+
+8. **User wants to work with spreadsheets**
+ 1. Create spreadsheet → `lark-cli sheets +create`
+ 2. Write data → `lark-cli sheets +write`
+ 3. Read data → `lark-cli sheets +read`
+
+9. **User wants to work with bitable (multi-dimensional tables)**
+ 1. Query records → `lark-cli base +query`
+ 2. Create bitable → `lark-cli base +create`
+
+10. **User wants meeting/minutes information**
+ 1. Query past meeting records → `lark-cli vc`
+ 2. Get meeting minutes/AI summary → `lark-cli minutes`
+ 3. Query future meetings (schedule) → `lark-cli calendar +agenda`
+ 4. Generate meeting summary report → `lark-cli workflow-meeting-summary`
+
+11. **User wants to manage wiki/knowledge base**
+ 1. List knowledge spaces → `lark-cli wiki +list`
+ 2. Create wiki document → `lark-cli wiki +create`
+ 3. Get wiki node metadata → `get_feishu_wiki_node_info`
+
+12. **User wants to create diagrams/charts**
+ 1. Mermaid flowchart/sequence/mindmap → `lark-cli docs +whiteboard-update`
+ 2. Create new whiteboard → `lark-cli whiteboard +create`
+ 3. Read whiteboard content → `get_feishu_whiteboard_content`
+
+13. **User wants file/folder operations**
+ 1. Upload file → `lark-cli drive upload`
+ 2. Download file → `lark-cli drive download`
+ 3. Create folder → `create_feishu_folder`
+ 4. List folder contents → `list_feishu_folder_contents`
+
+14. **User wants contact/people information**
+ 1. Search by name/email/phone → `lark-cli contact +search`
+
+15. **User wants to subscribe to events**
+ 1. Real-time event monitoring → `lark-cli event`
+
+**Fallback**: If CLI returns "skill not found", run `npx skills add larksuite/cli -y -g` and retry the command.
+
+---
+
+## Auth Verification Procedures
+
+### Procedure 1: CLI Auth Check
+
+1. Run `lark-cli auth status`.
+2. If output shows a valid App ID, user identity, and unexpired token → auth is valid. Proceed.
+3. If output shows "Not logged in" or "Token expired" → run `lark-cli auth login --recommend`.
+4. If `lark-cli` is not found → run `npm install -g @larksuite/cli` then `npx skills add larksuite/cli -y -g`.
+
+### Procedure 2: MCP Auth Check
+
+1. Check that `FEISHU_APP_ID` env var is set and starts with `cli_`.
+2. Check that `FEISHU_APP_SECRET` env var is set and is non-empty.
+3. If either is missing → ask the user for their Feishu App ID and App Secret.
+4. Test with a minimal MCP call: `mcporter call feishu.get_feishu_document_info --args '{"document_id":"test"}'`. If the error is about document_id (not auth), auth is working.
+5. If the error mentions "unauthorized" or "invalid token" → verify env vars and restart mcporter.
+
+### Procedure 3: Token Refresh
+
+1. **CLI token refresh**: Run `lark-cli auth login --recommend`. The browser opens for OAuth. Wait for user to approve.
+2. **MCP token refresh**: If using user auth mode, the user must re-authorize. Run the community MCP startup command to trigger OAuth on port 3333. If using tenant mode, tokens are auto-refreshed via App ID + App Secret.
+3. **Verify after refresh**: Run `lark-cli auth status` for CLI. Run a test MCP call for MCP.
+
+---
+
+## Extracting IDs from Feishu URLs
+
+When a user provides a Feishu URL instead of a token/ID, extract the ID using these patterns:
+
+| URL Format | ID Type | Regex |
+|---|---|---|
+| `https://*.feishu.cn/docx/{id}` | document_id | `/docx/([A-Za-z0-9]+)` |
+| `https://*.feishu.cn/wiki/{token}` | wiki_token | `/wiki/([A-Za-z0-9]+)` |
+| `https://*.feishu.cn/sheets/{token}` | spreadsheet_token | `/sheets/([A-Za-z0-9]+)` |
+| `https://*.feishu.cn/base/{token}` | bitable app_token | `/base/([A-Za-z0-9]+)` |
+| `https://*.feishu.cn/drive/folder/{token}` | folder_token | `/folder/([A-Za-z0-9]+)` |
+| `https://*.feishu.cn/minutes/{token}` | minute_token | `/minutes/([A-Za-z0-9]+)` |
+
+Example: From `https://abc.feishu.cn/docx/doxcn1234567890`, extract `doxcn1234567890` as the document_id.
+
+---
+
+## Command Examples (30+)
+
+### Calendar Domain
+
+**1.** "查看今天的日程" (View today's schedule)
+```bash
+lark-cli calendar +agenda
+```
+Expected output: List of today's events with time, title, and attendees.
+
+**2.** "帮我创建一个明天下午 2 点的会议" (Create a meeting tomorrow at 2 PM)
+```bash
+lark-cli calendar +create
+```
+Expected output: Event created confirmation with event link.
+
+**3.** "查一下张三这周的忙闲" (Check Zhang San's availability this week)
+```bash
+lark-cli calendar +freebusy
+```
+Expected output: Free/busy time blocks for the specified user.
+
+**4.** "推荐一个三人都有空的时间" (Suggest a time when all three are free)
+```bash
+lark-cli calendar +suggestion
+```
+Expected output: Multiple time slot suggestions with availability info.
+
+### Messaging Domain
+
+**5.** "给张三发消息说明天开会" (Send Zhang San a message about tomorrow's meeting)
+```bash
+lark-cli im +send
+```
+Expected output: Message sent confirmation with message_id.
+
+**6.** "搜索关于项目上线的聊天记录" (Search chat history about project launch)
+```bash
+lark-cli im +search
+```
+Expected output: List of matching messages with sender, time, and content preview.
+
+### Documents Domain
+
+**7.** "创建一个项目周报文档" (Create a weekly report document)
+```bash
+lark-cli doc +create
+```
+Expected output: Document created with URL.
+
+**8.** "搜索上周的会议纪要" (Search for last week's meeting notes)
+```bash
+lark-cli doc +search
+```
+Expected output: List of matching documents with titles and URLs.
+
+**9.** "在文档末尾追加一段总结" (Append a summary to the document)
+```bash
+lark-cli docs +update
+```
+Expected output: Document updated confirmation.
+
+**10.** "在文档里插入一张截图" (Insert a screenshot into the document)
+```bash
+lark-cli docs +media-insert
+```
+Expected output: Image inserted confirmation with image key.
+
+**11.** "获取文档的 Block 结构" (Get block structure of a document)
+```bash
+mcporter call feishu.get_feishu_document_blocks --args '{"document_id":"doxcnXXX"}'
+```
+Expected output: JSON array of blocks with type, content, and block_id.
+
+**12.** "提取文档的纯文本" (Extract plain text from a document)
+```bash
+mcporter call feishu.get_feishu_document_text --args '{"document_id":"doxcnXXX"}'
+```
+Expected output: Plain text string of the document content.
+
+**13.** "一次性添加三个标题段落" (Batch add three heading paragraphs)
+```bash
+mcporter call feishu.batch_create_feishu_blocks --args '{"document_id":"doxcnXXX","blocks":[{"type":"heading2","content":"第一章"},{"type":"heading2","content":"第二章"},{"type":"heading2","content":"第三章"}]}'
+```
+Expected output: Array of created block_ids.
+
+**14.** "把标题改成红色" (Change heading color to red)
+```bash
+mcporter call feishu.update_feishu_block_text --args '{"document_id":"doxcnXXX","block_id":"blkXXX","content":"标题","text_color":"red"}'
+```
+Expected output: Block updated confirmation.
+
+**15.** "删除文档中的第三个 Block" (Delete the third block)
+```bash
+mcporter call feishu.delete_feishu_document_blocks --args '{"document_id":"doxcnXXX","block_ids":["blk3XXX"]}'
+```
+Expected output: Deletion confirmation.
+
+**16.** "上传图片到文档" (Upload image to document)
+```bash
+mcporter call feishu.upload_and_bind_image_to_block --args '{"document_id":"doxcnXXX","image_url":"https://example.com/img.png"}'
+```
+Expected output: Image uploaded and bound, returns block_id.
+
+**17.** "在文档中创建一个 3x4 表格" (Create a 3x4 table in the document)
+```bash
+mcporter call feishu.create_feishu_table --args '{"document_id":"doxcnXXX","rows":3,"cols":4}'
+```
+Expected output: Table created, returns table block_id.
+
+### Cloud Storage Domain
+
+**18.** "上传这个 PDF" (Upload this PDF)
+```bash
+lark-cli drive upload
+```
+Expected output: Upload confirmation with file token.
+
+**19.** "下载这个文件" (Download this file)
+```bash
+lark-cli drive download
+```
+Expected output: File downloaded to local path.
+
+**20.** "创建一个叫项目资料的文件夹" (Create a folder called Project Materials)
+```bash
+mcporter call feishu.create_feishu_folder --args '{"name":"项目资料"}'
+```
+Expected output: Folder created with folder_token.
+
+**21.** "列出项目文档文件夹的内容" (List contents of the project documents folder)
+```bash
+mcporter call feishu.list_feishu_folder_contents --args '{"folderToken":"fldcnXXX"}'
+```
+Expected output: Array of files/folders with names, types, and tokens.
+
+### Spreadsheets Domain
+
+**22.** "创建一个销售数据表" (Create a sales data spreadsheet)
+```bash
+lark-cli sheets +create
+```
+Expected output: Spreadsheet created with URL and token.
+
+**23.** "把数据写入 A1:C5" (Write data to A1:C5)
+```bash
+lark-cli sheets +write
+```
+Expected output: Cells updated confirmation.
+
+**24.** "读取 A1:D10 的数据" (Read data from A1:D10)
+```bash
+lark-cli sheets +read
+```
+Expected output: Cell values in tabular format.
+
+### Bitable Domain
+
+**25.** "查询项目表中本周的记录" (Query this week's records from the project table)
+```bash
+lark-cli base +query
+```
+Expected output: Records matching the filter with field values.
+
+**26.** "新建一个 Bug 跟踪表" (Create a new bug tracking table)
+```bash
+lark-cli base +create
+```
+Expected output: Bitable created with app_token and URL.
+
+### Tasks Domain
+
+**27.** "创建一个任务：下周五前完成设计评审" (Create a task: complete design review by next Friday)
+```bash
+lark-cli task +create
+```
+Expected output: Task created with task_id and link.
+
+**28.** "看看我的待办任务" (View my pending tasks)
+```bash
+lark-cli task +get-my-tasks
+```
+Expected output: List of tasks with status, due date, and assignee.
+
+### Email Domain
+
+**29.** "给团队发一封项目进度邮件" (Send a project progress email to the team)
+```bash
+lark-cli mail +compose
+```
+Expected output: Email sent confirmation with message_id.
+
+**30.** "搜索关于合同的邮件" (Search for emails about contracts)
+```bash
+lark-cli mail +search
+```
+Expected output: List of matching emails with subject, sender, and date.
+
+### Wiki Domain
+
+**31.** "列出所有知识空间" (List all knowledge spaces)
+```bash
+lark-cli wiki +list
+```
+Expected output: List of wiki spaces with names and tokens.
+
+**32.** "获取这个 Wiki 节点的信息" (Get wiki node information)
+```bash
+mcporter call feishu.get_feishu_wiki_node_info --args '{"token":"wikcnXXX"}'
+```
+Expected output: Node metadata including title, type, parent.
+
+### Whiteboard Domain
+
+**33.** "创建一个新画板" (Create a new whiteboard)
+```bash
+lark-cli whiteboard +create
+```
+Expected output: Whiteboard created with URL.
+
+**34.** "在画板上画一个流程图" (Draw a flowchart on the whiteboard)
+```bash
+lark-cli docs +whiteboard-update
+```
+Expected output: Whiteboard updated with Mermaid diagram.
+
+**35.** "获取画板内容" (Get whiteboard content)
+```bash
+mcporter call feishu.get_feishu_whiteboard_content --args '{"whiteboard_id":"wbXXX"}'
+```
+Expected output: JSON structure of whiteboard elements.
+
+### Contacts Domain
+
+**36.** "搜索张三的联系方式" (Search for Zhang San's contact info)
+```bash
+lark-cli contact +search
+```
+Expected output: User profile with name, email, phone, department.
+
+### Meetings Domain
+
+**37.** "看看昨天有哪些会议记录" (View yesterday's meeting records)
+```bash
+lark-cli vc
+```
+Expected output: List of meeting records with title, time, organizer.
+
+### Search Domain (cross-domain)
+
+**38.** "搜索关于 Q1 的文档" (Search for documents about Q1)
+```bash
+lark-cli doc +search
+```
+Expected output: List of documents matching the query.
+
+**39.** "搜索文档（通过 MCP）" (Search documents via MCP)
+```bash
+mcporter call feishu.search_feishu_documents --args '{"query":"Q1","count":10}'
+```
+Expected output: Array of matching documents with title and token.
+
+---
+
+## Operating in MCP-Only Mode
+
+If CLI is not installed (setup ran with `--mcp-only`), only the following capabilities are available:
+
+**Available (via Community MCP):**
+- Document creation, editing, reading, and block manipulation
+- Folder creation and listing
+- Document search
+- Wiki node metadata queries
+- Whiteboard content reading (read-only)
+
+**Not available (CLI required):**
+- Messaging (im)
+- Calendar
+- Email
+- Tasks
+- Contacts
+- Meetings / Minutes
+- Spreadsheets
+- Bitable
+- Event subscriptions
+- Whiteboard creation and editing
+- Mermaid diagrams / advanced charts
+- File upload / download
+
+If a user requests an unavailable capability in MCP-only mode, inform them: "此功能需要飞书 CLI。运行 `npm install -g @larksuite/cli && npx skills add larksuite/cli -y -g` 安装 CLI。"
+
+---
+
+## Permissions Quick Reference
+
+| Domain | Required Scope |
+|---|---|
+| Documents | `docx:document` |
+| Cloud Storage | `drive:drive` |
+| Wiki | `wiki:wiki` |
+| Bitable | `bitable:app` |
+| Messaging | `im:message` |
+| Search | `search:docs:read` |
+| Calendar | `calendar:calendar` |
+| Email | `mail:message` |
+| Contacts | `contact:user.base:readonly` |
+| Meetings | `vc:meeting:readonly` |
+| Minutes | `minutes:minute:readonly` |
+| Tasks | `task:task` |
+
+Manage permissions at: `https://open.feishu.cn/app/{APP_ID}/security`
+
+---
+
+## Error Handling Quick Reference
+
+If any command fails, consult TROUBLESHOOT-AGENT.md for diagnosis and recovery steps.
+
+Common quick fixes:
+- "command not found: lark-cli" → `npm install -g @larksuite/cli`
+- "skill not found" → `npx skills add larksuite/cli -y -g`
+- "Token expired" → `lark-cli auth login --recommend`
+- "Permission denied" → Add missing scope at `https://open.feishu.cn/app/{APP_ID}/security`
+- "FEISHU_APP_ID not set" → Set env vars and restart mcporter
diff --git a/CLAUDE.md b/CLAUDE.md
new file mode 100644
index 0000000..bf0ab32
--- /dev/null
+++ b/CLAUDE.md
@@ -0,0 +1,193 @@
+# Feishu-MCP v2 — 飞书全能集成套件
+
+## What This Project Is
+
+**feishu-mcp 是飞书全能集成套件（Meta-Skill Package）**— 让飞书 CLI 和社区 MCP 作为一个统一系统工作的集成层。
+
+它不是 MCP 服务器，不是 CLI 替代品，不是运行时组件。它是：
+1. **一条命令安装脚本** — 安装 + 配置 CLI + 社区 MCP + OpenClaw 检测
+2. **决策矩阵 Skill** (`SKILL.md`) — 告诉 AI agent 每个任务该用哪个工具
+3. **Agent 文档** — 让 AI 助手自助完成安装、诊断、错误恢复
+4. **用户引导文档** — 零技术背景用户的中文指南
+5. **平台配置模板** — OpenClaw / Claude Code / Cursor 预配置
+6. **工作流 Skill** — 展示 CLI + MCP 协同能力 + 中文友好化包装
+
+## Harness Methodology
+
+This project uses a **Planner → Builder → QA** harness for ALL development (full process).
+
+### Agent Roles
+- **@planner**: Expands goals into specs with acceptance criteria. Writes WHAT, never HOW.
+- **@builder**: Implements features according to specs. Follows existing patterns.
+- **@qa**: Independently verifies output. Scores 5 dimensions, any < 6 = FAIL.
+- **@coordinator**: Routes requests to the right agent based on change size.
+
+### Communication
+Agents communicate through files in `.harness/`:
+- `spec.md` — Overall transformation plan (ground truth)
+- `contracts/{unit}.md` — Per-unit specs (Planner output)
+- `reports/build_{unit}_r{N}.md` — Builder reports
+- `reports/qa_{unit}_r{N}.md` — QA reports
+- `progress.tsv` — Iteration history
+- `experience/` — Accumulated patterns and lessons
+
+### Workflow
+```
+Plan → Build → QA ─── PASS → git commit → next unit
+ └─ FAIL → same failure?
+ ├─ NO → fix → re-verify
+ └─ YES → replan
+```
+
+## Architecture: Two-Layer (CLI + Community MCP)
+
+```
+User <-> OpenClaw/Claude Code/Cursor <-> feishu-mcp 集成层
+ |
+ +---------+---------+
+ | |
+ Feishu CLI Community MCP
+ (@larksuite/cli) (feishu-mcp)
+ 200+ commands 15 doc editing tools
+ 19 AI skills Block-level precision
+ 11 business domains
+```
+
+**Key Principle: CLI for breadth, Community MCP for depth.**
+
+### Why Two Layers (Not Three)
+
+CLI 已覆盖官方 MCP 的大部分能力。官方 MCP（Beta）作为高级可选项，不在默认安装中。
+
+### CLI vs Community MCP: Complementary Capabilities
+
+| Capability | Feishu CLI | Community MCP | Who Wins |
+|-----------|-----------|---------------|----------|
+| Batch block creation | ⚠️ Partial (Markdown) | ✅ `batch_create` | **MCP unique advantage** |
+| Text formatting (color/align) | ✅ (Markdown syntax) | ✅ (8-color, alignment) | **MCP more precise** |
+| Mermaid diagrams | ✅ (whiteboard skill) | ❌ (planned) | **CLI unique advantage** |
+| Advanced charts/DSL | ✅ (DSL path) | ❌ (planned) | **CLI unique advantage** |
+| Whiteboard editing | ✅ (create+edit) | ⚠️ (read-only) | **CLI unique advantage** |
+| Image upload | ✅ | ✅ | CLI more robust |
+| Block editing | ✅ (selection) | ✅ (direct API) | MCP more granular |
+| Block deletion | ✅ | ✅ | Equal |
+| Messaging/Calendar/Mail/Tasks/Contacts/Meetings/Wiki/Sheets/Base | ✅ | ❌ | **CLI only** |
+
+### Decision Matrix: CLI vs MCP
+
+| Task | Use | Reason |
+|------|-----|--------|
+| Send message | CLI (`lark-cli im +send`) | Full control, attachments |
+| Calendar ops | CLI (`lark-cli calendar +agenda`) | Complete API coverage |
+| Create document | CLI (`lark-cli doc +create`) | Markdown input, simpler |
+| Edit doc blocks precisely | Community MCP (`update_feishu_block_text`) | Granular block control |
+| Batch create blocks | Community MCP (`batch_create_feishu_blocks`) | Atomic batch operation |
+| Text color/alignment | Community MCP (style properties) | 8-color palette, explicit align |
+| Insert Mermaid diagram | CLI (`lark-cli whiteboard`) | Community MCP doesn't support yet |
+| Insert charts | CLI (whiteboard DSL) | Community MCP doesn't support yet |
+| Whiteboard create/edit | CLI (`lark-cli docs +whiteboard-update`) | MCP is read-only |
+| Upload image to doc | CLI (`lark-cli docs +media-insert`) or MCP | Both work, CLI more robust |
+| Query bitable | CLI (`lark-cli base +query`) | CLI simpler |
+| Search docs | CLI (`lark-cli doc +search`) | Official, more reliable |
+| Send mail | CLI (`lark-cli mail +compose`) | Only option |
+| Manage tasks | CLI (`lark-cli task +list`) | Only option |
+| Wiki operations | CLI (`lark-cli wiki`) | Only option |
+| Contacts | CLI (`lark-cli contact`) | Only option |
+| Meetings/Minutes | CLI (`lark-cli vc`, `lark-cli minutes`) | Only option |
+
+## Conventions
+
+### Languages
+- **User-facing content**: Chinese (primary), English section headers
+- **Agent-facing docs**: English (primary), Chinese examples where helpful
+- **Code comments**: English
+- **Commit messages**: English
+
+### File Structure
+```
+feishu-mcp/
+ setup.sh ← Legacy setup (v1, kept for compatibility)
+ setup-v2.sh ← New smart setup script
+ README.md ← User-facing intro + quick start
+ GUIDE.md ← Step-by-step visual guide for non-technical users
+ SKILL.md ← Decision matrix (CLI + MCP routing)
+ AGENTS.md ← Master reference for AI agents
+ QUICKSTART-AGENT.md ← Agent setup automation guide
+ TROUBLESHOOT-AGENT.md ← Agent error recovery guide
+ CLAUDE.md ← This file (project context + harness rules)
+ configs/
+ openclaw.json ← OpenClaw config template
+ claude-code.json ← Claude Code config template
+ cursor.json ← Cursor config template
+ skills/
+ feishu-overview.md ← Overview: Chinese index of all 19 CLI skills + MCP tools
+ doc-review.md ← Workflow: read blocks (MCP) → create tasks (CLI) → send summary (CLI)
+ data-to-doc.md ← Workflow: query bitable (CLI) → generate formatted doc (MCP)
+ knowledge-sync.md ← Workflow: read Wiki (CLI) → update doc content (MCP)
+ meeting-summary-plus.md ← Enhanced: CLI built-in + MCP formatted doc output
+ .claude/
+ agents/
+ planner.md
+ builder.md
+ qa.md
+ coordinator.md
+ settings.json
+ .harness/
+ spec.md
+ progress.tsv
+ contracts/
+ reports/
+ experience/
+```
+
+### Naming
+- Shell scripts: `kebab-case.sh`
+- Top-level docs: `UPPER-CASE.md`
+- Subdirectory docs: `kebab-case.md`
+- Config files: `{tool-name}.json`
+
+### Shell Script Standards
+- Shebang: `#!/usr/bin/env bash`
+- `set -euo pipefail` at top
+- All user prompts in Chinese
+- Error messages suggest fix action
+- No `sudo` required
+- No hardcoded paths (use `$HOME`, `$XDG_CONFIG_HOME`)
+- Idempotent (safe to re-run)
+
+### Documentation Standards
+- Agent docs: imperative voice ("Do X", "Check Y", "If Z then W")
+- User docs: conversational, zero jargon, step-by-step
+- Every doc has a clear audience header
+- No placeholder values in published docs — use descriptive comments
+
+## Quality Dimensions
+
+QA scores 5 dimensions (1-10 each). Any < 6 = FAIL:
+1. **User Experience** (30%) — minimal steps, no config editing, clear errors
+2. **Agent Operability** (25%) — unambiguous tool selection, automated recovery
+3. **Completeness** (20%) — all Feishu capabilities accessible
+4. **Error Handling** (15%) — common errors have recovery paths
+5. **Documentation Clarity** (10%) — non-technical user can follow
+
+## Key Dependencies
+
+### Required (default install)
+- Node.js >= 16
+- npm
+- `@larksuite/cli` >= 1.0.0 (official Feishu CLI)
+- `feishu-mcp` (community MCP for document editing)
+- `mcporter` (MCP port multiplexer)
+- Feishu Open Platform app (user creates at open.feishu.cn)
+
+### Optional
+- `@larksuiteoapi/lark-mcp` (official MCP, advanced users only)
+- `@larksuiteoapi/feishu-openclaw-plugin` (OpenClaw channel plugin)
+
+## Auth Flow Summary
+
+1. User creates app at open.feishu.cn → gets App ID + App Secret
+2. Setup script configures redirect URL (http://localhost:3000/callback for CLI, http://localhost:3333/callback for community MCP)
+3. `lark-cli auth login --recommend` opens browser for OAuth
+4. User approves permissions → token stored in OS keychain (CLI) and env vars (MCP)
+5. Token refresh: CLI handles automatically, community MCP may need manual refresh
diff --git a/GUIDE.md b/GUIDE.md
new file mode 100644
index 0000000..87c0135
--- /dev/null
+++ b/GUIDE.md
@@ -0,0 +1,392 @@
+# 从零开始配置飞书 AI 助手
+
+本指南面向**从未使用过飞书开放平台**的用户。按照以下步骤操作，你就能让 AI 助手完整控制你的飞书。
+
+全程大约需要 10-15 分钟。
+
+---
+
+## 目录
+
+- [第一步：创建飞书应用](#第一步创建飞书应用)
+- [第二步：配置应用权限](#第二步配置应用权限)
+- [第三步：配置安全设置](#第三步配置安全设置)
+- [第四步：发布应用](#第四步发布应用)
+- [第五步：运行安装脚本](#第五步运行安装脚本)
+- [第六步：验证安装](#第六步验证安装)
+- [第七步：第一次对话 — 试试这些例子](#第七步第一次对话--试试这些例子)
+- [常见问题](#常见问题)
+
+---
+
+## 第一步：创建飞书应用
+
+你需要在飞书开放平台创建一个「自建应用」。这个应用相当于 AI 助手操作飞书时使用的「身份证」。
+
+### 1.1 打开飞书开放平台
+
+在浏览器中访问：**https://open.feishu.cn**
+
+如果还没有登录，使用你的飞书账号登录。
+
+### 1.2 进入开发者后台
+
+登录后，点击页面右上角的「开发者后台」按钮。
+
+### 1.3 创建新应用
+
+1. 点击「创建自建应用」按钮
+2. 填写应用信息：
+ - **应用名称**：填一个你能认出来的名字，比如「AI 飞书助手」或「我的飞书 MCP」
+ - **应用描述**：可以填「通过 AI 工具操作飞书」（随意填写，不影响功能）
+ - **应用图标**：可以上传一个图标，也可以使用默认图标
+3. 点击「确认创建」
+
+### 1.4 记下 App ID 和 App Secret
+
+创建完成后，你会进入应用管理页面。在「凭证与基础信息」页面中，你可以看到：
+
+- **App ID**：以 `cli_` 开头的一串字符，比如 `cli_a1b2c3d4e5f6`
+- **App Secret**：一串较长的字符
+
+**请把这两个值复制下来保存好**，后面安装的时候需要用到。
+
+> 注意：App Secret 是敏感信息，不要分享给他人或提交到公开的代码仓库。
+
+---
+
+## 第二步：配置应用权限
+
+权限决定了 AI 助手可以操作飞书的哪些功能。你可以根据自己的需要选择开通。
+
+### 2.1 进入权限管理页面
+
+在应用管理页面的左侧菜单中，找到并点击「权限管理」。
+
+### 2.2 开通基础权限
+
+在搜索框中搜索以下权限名称，找到后点击「开通」按钮：
+
+**必须开通的权限：**
+
+| 在搜索框中输入 | 权限名称 | 作用 |
+|---------------|---------|------|
+| `docx:document` | 查看、评论和编辑云文档 | AI 能创建和编辑飞书文档 |
+| `drive:drive` | 查看、管理和上传/下载云空间中的文件 | AI 能管理文件和文件夹 |
+| `wiki:wiki` | 查看、编辑和管理知识库 | AI 能访问知识库内容 |
+
+### 2.3 开通推荐权限（按需）
+
+以下权限对应不同的飞书功能，根据你的需求选择开通：
+
+**消息功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `im:message` | 获取与发送消息 | 让 AI 发送和接收飞书消息 |
+| `im:chat` | 获取群信息 | 让 AI 查看群聊信息 |
+
+**日历功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `calendar:calendar` | 查看和管理日历及日程 | 让 AI 查看日程、创建会议 |
+
+**表格功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `bitable:app` | 查看、评论和编辑多维表格 | 让 AI 操作多维表格 |
+| `sheets:spreadsheet` | 查看和管理电子表格 | 让 AI 创建和编辑电子表格 |
+
+**任务功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `task:task` | 查看和管理任务 | 让 AI 创建和管理待办任务 |
+
+**邮件功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `mail:message` | 读取和发送邮件 | 让 AI 收发邮件 |
+
+**通讯录功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `contact:user.base:readonly` | 获取用户基本信息 | 让 AI 搜索员工联系方式 |
+
+**会议和妙记功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `vc:meeting` | 查看会议记录 | 让 AI 查询已结束的会议 |
+| `minutes:minute` | 查看妙记 | 让 AI 获取会议 AI 总结 |
+
+**搜索功能：**
+
+| 搜索 | 权限名称 | 开通后可以 |
+|------|---------|-----------|
+| `search:docs:read` | 搜索云文档 | 让 AI 按关键词搜索文档 |
+
+> 不确定需要哪些？建议先开通基础权限和日历、消息功能。其他权限以后随时可以补开——当 AI 提示「权限不足」时，再回来开通对应权限即可。
+
+---
+
+## 第三步：配置安全设置
+
+安装脚本在获取授权时需要通过浏览器回调，所以需要把回调地址配置到应用中。
+
+### 3.1 进入安全设置
+
+在应用管理页面的左侧菜单中，点击「安全设置」。
+
+### 3.2 添加重定向 URL
+
+找到「重定向 URL」部分，点击「添加」按钮，依次添加以下两个地址：
+
+**第一个地址（用于飞书 CLI 授权）：**
+```
+http://localhost:3000/callback
+```
+
+**第二个地址（用于社区 MCP 授权）：**
+```
+http://localhost:3333/callback
+```
+
+添加完成后，确认列表中显示了这两个地址。
+
+> 这两个地址是本地回调地址，只在你的电脑上使用，不会暴露到互联网。
+
+---
+
+## 第四步：发布应用
+
+飞书应用需要发布后才能正常使用。
+
+### 4.1 创建版本
+
+1. 在应用管理页面的左侧菜单中，点击「版本管理与发布」
+2. 点击「创建版本」
+3. 填写版本信息：
+ - **版本号**：填 `1.0.0` 即可
+ - **更新说明**：随意填写，比如「初始版本」
+4. 点击「保存」
+
+### 4.2 申请发布
+
+1. 点击「申请发布」
+2. 如果你是企业管理员，会直接通过
+3. 如果你不是管理员，需要等待管理员审批
+
+> 如果你只是在个人环境使用（比如个人版飞书），通常创建后就可以直接使用，不需要额外审批。
+
+---
+
+## 第五步：运行安装脚本
+
+### 5.1 确认前置条件
+
+在运行脚本前，确认你的电脑上已安装 Node.js。打开终端（macOS 的「终端」应用或 Linux 的终端），输入：
+
+```bash
+node --version
+```
+
+如果显示 `v16.x.x` 或更高版本（比如 `v18.17.0`、`v20.11.0`），说明已准备好。
+
+如果提示找不到命令，请先安装 Node.js：前往 https://nodejs.org 下载并安装。
+
+### 5.2 运行安装命令
+
+在终端中执行：
+
+```bash
+curl -fsSL https://raw.githubusercontent.com/AlexAnys/feishu-mcp/main/setup-v2.sh | bash
+```
+
+### 5.3 按提示操作
+
+脚本会依次提示你：
+
+1. **输入 App ID** — 粘贴你在第一步中保存的 App ID（以 `cli_` 开头）
+2. **输入 App Secret** — 粘贴你保存的 App Secret
+3. **浏览器授权** — 脚本会自动打开浏览器，页面上会显示飞书授权确认。点击「授权」按钮
+4. **检测 AI 工具** — 脚本会自动检测你系统上安装的 AI 工具（OpenClaw、Claude Code、Cursor），并生成对应配置
+
+整个过程大约 2-3 分钟，脚本会在每一步显示进度。
+
+### 5.4 安装完成
+
+当你看到安装成功的提示信息时，说明一切就绪。
+
+---
+
+## 第六步：验证安装
+
+打开你的 AI 工具（OpenClaw、Claude Code 或 Cursor），在对话中输入：
+
+> "帮我检查一下飞书的认证状态"
+
+如果 AI 回复了认证信息（比如显示当前登录的用户名），说明安装成功。
+
+你也可以试试：
+
+> "帮我看一下今天的飞书日程"
+
+如果 AI 返回了你的日程信息（或者提示今天没有日程），说明一切正常。
+
+---
+
+## 第七步：第一次对话 — 试试这些例子
+
+以下是一些你可以立刻尝试的对话示例，涵盖不同的飞书功能：
+
+### 查看日程
+
+> "我今天有什么安排？"
+
+AI 会列出你今天的飞书日历日程。
+
+### 发送消息
+
+> "给张三发一条飞书消息，内容是：明天的周会改到下午 3 点"
+
+AI 会通过飞书给张三发送消息。（需要先确认张三在你的通讯录中。）
+
+### 创建文档
+
+> "帮我创建一个飞书文档，标题是《Q1 工作总结》，里面加三个章节：目标回顾、主要成果、下季度计划"
+
+AI 会在飞书中创建文档并添加章节标题，然后给你返回文档链接。
+
+### 搜索文档
+
+> "搜索一下飞书里关于产品上线的文档"
+
+AI 会搜索你的飞书云空间，返回匹配的文档列表。
+
+### 查看任务
+
+> "看看我在飞书里还有哪些未完成的任务"
+
+AI 会列出你的待办任务列表。
+
+### 查收件箱
+
+> "看看我最近有什么飞书邮件"
+
+AI 会显示你的最近收件箱内容。
+
+### 查询多维表格
+
+> "帮我查一下项目跟踪表里状态为「进行中」的记录"
+
+AI 会查询你的多维表格并返回符合条件的数据。
+
+### 搜索同事
+
+> "搜索一下李四的飞书联系方式"
+
+AI 会在通讯录中搜索并返回李四的信息。
+
+### 创建画板流程图
+
+> "帮我创建一个飞书画板，画一个用户注册流程图"
+
+AI 会创建画板并在上面画出流程图。
+
+### 组合操作
+
+> "帮我看一下昨天所有会议的纪要，整理成一份总结文档"
+
+AI 会先查询昨天的会议记录，获取纪要内容，然后创建一个新文档把总结写进去。
+
+---
+
+## 常见问题
+
+### 创建应用相关
+
+**Q: 我没有飞书开放平台的权限怎么办？**
+
+所有飞书用户都可以访问飞书开放平台并创建自建应用。如果你无法访问 open.feishu.cn，可能是企业管理员限制了开发者功能，请联系你的 IT 管理员。
+
+**Q: 我应该创建「自建应用」还是「商店应用」？**
+
+选择「自建应用」。商店应用是给第三方开发者上架到飞书应用市场用的，不适合个人使用。
+
+**Q: App Secret 泄露了怎么办？**
+
+在飞书开放平台的应用管理页面中，找到「凭证与基础信息」，点击 App Secret 旁边的「重置」按钮生成新的密钥。重置后需要重新运行安装脚本。
+
+### 权限相关
+
+**Q: 不知道该开通哪些权限？**
+
+建议先只开通三个基础权限（文档、云空间、知识库）。当你使用某个功能时，如果 AI 提示「权限不足」，再回来开通对应权限即可。这样最安全，也最简单。
+
+**Q: 开通权限后需要重新发布应用吗？**
+
+如果只是增加权限范围（scope），通常不需要重新发布。但如果平台提示需要发布新版本，按照第四步的操作再发布一次即可。
+
+**Q: 权限审批需要多久？**
+
+如果你是管理员，权限会立即生效。如果需要管理员审批，时间取决于你公司的审批流程。
+
+### 安装脚本相关
+
+**Q: 脚本提示 Node.js 版本过低怎么办？**
+
+请前往 https://nodejs.org 下载并安装最新的 LTS（长期支持）版本。安装完成后重新运行脚本。
+
+**Q: 浏览器没有自动打开怎么办？**
+
+脚本会在终端中显示授权链接。手动复制这个链接到浏览器中打开，完成授权后回到终端继续。
+
+**Q: 安装中断了，可以重新运行吗？**
+
+可以。安装脚本设计为可以安全地反复运行，不会重复安装已有组件或覆盖已有配置。直接重新运行即可。
+
+**Q: 授权页面显示错误怎么办？**
+
+请检查：
+1. 重定向 URL 是否正确配置（`http://localhost:3000/callback` 和 `http://localhost:3333/callback`）
+2. 应用是否已发布
+3. 是否开通了所需权限
+
+如果以上都确认无误，尝试重新运行安装脚本。
+
+### 使用相关
+
+**Q: AI 说「找不到工具」或「功能不可用」怎么办？**
+
+这通常意味着飞书 CLI 技能没有正确安装。在终端中运行以下命令检查：
+
+```bash
+lark-cli skills list
+```
+
+如果列表为空或不完整，运行以下命令重新安装技能：
+
+```bash
+npx skills add larksuite/cli -y -g
+```
+
+**Q: AI 提示「认证过期」或「token 无效」怎么办？**
+
+告诉 AI「重新登录飞书」，AI 会自动执行重新授权流程。也可以在终端手动运行：
+
+```bash
+lark-cli auth login --recommend
+```
+
+**Q: 我想在另一台电脑上也使用，怎么做？**
+
+在新电脑上重新运行安装脚本即可。使用同一个 App ID 和 App Secret，不需要重新创建应用。
+
+---
+
+> 遇到本指南未覆盖的问题？请在 [GitHub Issues](https://github.com/AlexAnys/feishu-mcp/issues) 中提出，我们会尽快回复。
diff --git a/QUICKSTART-AGENT.md b/QUICKSTART-AGENT.md
new file mode 100644
index 0000000..f252b7e
--- /dev/null
+++ b/QUICKSTART-AGENT.md
@@ -0,0 +1,308 @@
+# QUICKSTART-AGENT.md — Feishu Integration Setup Automation
+
+**Audience**: AI agents automating Feishu setup for a new user.
+**Purpose**: Step-by-step procedure to install and configure CLI + Community MCP from scratch.
+**Prerequisites**: Node.js >= 16, npm, internet access.
+
+All instructions are imperative. Execute steps in order. Prompt the user only when prefixed with `[ASK USER]` or `[USER ACTION]`.
+
+---
+
+## Pre-Setup Checks
+
+Run these checks before starting. If any fails, resolve before proceeding.
+
+### Check 1: Node.js version
+
+```bash
+node --version
+```
+- **Expected**: `v16.x.x` or higher (e.g., `v18.17.0`, `v20.11.0`).
+- **If fails**: Output "Node.js 未安装或版本过低。请安装 Node.js 16 或更高版本：https://nodejs.org/". Stop.
+
+### Check 2: npm presence
+
+```bash
+npm --version
+```
+- **Expected**: Any version number (e.g., `9.8.1`).
+- **If fails**: Output "npm 未找到。Node.js 安装通常自带 npm。请重新安装 Node.js：https://nodejs.org/". Stop.
+
+### Check 3: CLI already installed
+
+```bash
+which lark-cli
+```
+- **Expected**: A path (e.g., `/usr/local/bin/lark-cli`).
+- **If not found**: CLI not installed. Continue to setup — installation will happen in the procedure.
+
+### Check 4: MCP already configured
+
+```bash
+echo $FEISHU_APP_ID
+```
+- **Expected**: A string starting with `cli_` (e.g., `cli_a1234567890`).
+- **If empty**: MCP not configured. Continue to setup.
+- **If set**: MCP may already be configured. Verify with a test call in the smoke test section.
+
+### Check 5: mcporter presence
+
+```bash
+which mcporter
+```
+- **Expected**: A path (e.g., `/usr/local/bin/mcporter`).
+- **If not found**: mcporter not installed. Will be installed during setup.
+
+---
+
+## Setup Procedure
+
+### Step 1: Run setup script
+
+**[EXECUTE]** Run the setup script:
+
+```bash
+bash setup-v2.sh
+```
+
+If `setup-v2.sh` does not exist in the current directory, download it:
+
+```bash
+curl -fsSL https://raw.githubusercontent.com/AlexAnys/feishu-mcp/main/setup-v2.sh -o setup-v2.sh && bash setup-v2.sh
+```
+
+**If fails**: Check error output. Common issues:
+- "npm: command not found" → Node.js/npm not installed. Go back to Pre-Setup Checks.
+- Network error → Check internet connectivity. Retry after network is restored.
+- Permission error → Do not use `sudo`. Check npm global directory permissions.
+
+If the setup script is not available, proceed with manual steps (Steps 2-6 below). If the script ran successfully, skip to Step 7.
+
+---
+
+### Step 2: Install Feishu CLI
+
+**[EXECUTE]** Install the CLI globally:
+
+```bash
+npm install -g @larksuite/cli
+```
+
+**If fails**: Try with explicit registry:
+```bash
+npm install -g @larksuite/cli --registry https://registry.npmjs.org
+```
+If still fails, check npm permissions. Output the error to the user.
+
+---
+
+### Step 3: Install CLI skills
+
+**[EXECUTE]** Load all 19 AI skills:
+
+```bash
+npx skills add larksuite/cli -y -g
+```
+
+**If fails**: Retry once. If still fails, output "CLI 技能安装失败。请检查网络连接后重试。" and continue — skills can be installed later.
+
+---
+
+### Step 4: Guide user through Feishu app creation
+
+**[ASK USER]** Display:
+
+> 需要创建一个飞书自建应用来获取 API 凭证。
+>
+> 请打开飞书开放平台：https://open.feishu.cn/app
+>
+> 1. 点击「创建企业自建应用」
+> 2. 输入应用名称（如：我的 AI 助手）
+> 3. 创建完成后，在「凭证与基础信息」页面找到 **App ID** 和 **App Secret**
+>
+> 请提供你的 App ID（以 cli_ 开头）：
+
+Wait for user to provide the App ID. Store as `APP_ID`.
+
+**[ASK USER]** Display:
+
+> 请提供 App Secret：
+
+Wait for user to provide the App Secret. Store as `APP_SECRET`.
+
+**If user cannot create the app**: Direct them to https://open.feishu.cn/app with step-by-step guidance. They must be a Feishu admin or have app creation permissions.
+
+---
+
+### Step 5: Configure redirect URLs
+
+**[USER ACTION]** Display:
+
+> 在飞书开放平台的应用设置中，请添加以下重定向 URL：
+>
+> 安全设置 → 重定向 URL → 添加：
+> - `http://localhost:3000/callback`（CLI 认证用）
+> - `http://localhost:3333/callback`（MCP 认证用）
+>
+> 添加完成后请确认。
+
+Wait for user confirmation.
+
+**If user cannot find the setting**: Direct them to: `https://open.feishu.cn/app/{APP_ID}/security` → "重定向 URL" section.
+
+---
+
+### Step 6: Configure and authenticate
+
+**[EXECUTE]** Initialize CLI configuration:
+
+```bash
+lark-cli config init --new
+```
+
+When prompted, enter the App ID and App Secret provided by the user.
+
+**[EXECUTE]** Start OAuth login:
+
+```bash
+lark-cli auth login --recommend
+```
+
+**[USER ACTION]** Display:
+
+> 浏览器将打开飞书授权页面。请在浏览器中：
+> 1. 使用飞书账号登录
+> 2. 点击「授权」按钮
+> 3. 看到「授权成功」后回到这里
+
+Wait for OAuth to complete.
+
+**If fails**: Check that port 3000 is not in use (`lsof -i :3000`). If occupied, kill the process and retry.
+
+---
+
+### Step 7: Configure community MCP
+
+**[EXECUTE]** Install mcporter if not present:
+
+```bash
+npm install -g mcporter
+```
+
+**[EXECUTE]** Set MCP environment variables (use the App ID and App Secret from Step 4):
+
+```bash
+export FEISHU_APP_ID="$APP_ID"
+export FEISHU_APP_SECRET="$APP_SECRET"
+```
+
+**If setup-v2.sh was used**: These should already be configured. Verify with `echo $FEISHU_APP_ID`.
+
+---
+
+### Step 8: Verify auth status
+
+**[EXECUTE]** Check CLI auth:
+
+```bash
+lark-cli auth status
+```
+
+- **Expected**: Shows App ID, user identity, valid token.
+- **If "Not logged in"**: Re-run `lark-cli auth login --recommend`.
+- **If "Token expired"**: Re-run `lark-cli auth login --recommend`.
+
+**[EXECUTE]** Check MCP env vars:
+
+```bash
+echo "APP_ID: $FEISHU_APP_ID"
+echo "SECRET set: $([ -n "$FEISHU_APP_SECRET" ] && echo yes || echo no)"
+```
+
+- **Expected**: APP_ID shows `cli_xxx`, SECRET shows `yes`.
+- **If empty**: Re-set the environment variables from Step 7.
+
+---
+
+### Step 9: Enable Feishu permissions
+
+**[USER ACTION]** Display:
+
+> 请在飞书开放平台开通以下权限：
+>
+> 打开：https://open.feishu.cn/app/{APP_ID}/security
+>
+> 基础权限（必需）：
+> - `docx:document` — 读写文档
+> - `drive:drive` — 云空间
+> - `wiki:wiki` — 知识库
+>
+> 推荐权限：
+> - `bitable:app` — 多维表格
+> - `im:message` — 消息
+> - `search:docs:read` — 搜索
+> - `calendar:calendar` — 日历
+> - `mail:message` — 邮件
+> - `contact:user.base:readonly` — 通讯录
+> - `task:task` — 任务
+>
+> 开通后请确认。
+
+Wait for user confirmation.
+
+---
+
+## Smoke Tests
+
+After setup completes, run these tests to verify the integration works end-to-end.
+
+### Smoke Test 1: CLI Layer
+
+```bash
+lark-cli auth status
+```
+
+- **Success**: Output shows valid App ID and user identity with unexpired token.
+- **Failure**: Output shows "Not logged in" or error.
+- **Recovery**: Run `lark-cli auth login --recommend` and repeat.
+
+### Smoke Test 2: CLI Command
+
+```bash
+lark-cli calendar +agenda
+```
+
+- **Success**: Output shows today's calendar events (or empty list if no events).
+- **Failure — "skill not found"**: Run `npx skills add larksuite/cli -y -g` then retry.
+- **Failure — "permission denied"**: Add `calendar:calendar` scope at `https://open.feishu.cn/app/{APP_ID}/security`.
+- **Failure — auth error**: Run `lark-cli auth login --recommend` then retry.
+
+### Smoke Test 3: MCP Layer
+
+```bash
+mcporter call feishu.create_feishu_document --args '{"title":"测试文档"}'
+```
+
+- **Success**: Output returns a document_id and URL. Delete the test document afterwards.
+- **Failure — "unauthorized"**: Check `FEISHU_APP_ID` and `FEISHU_APP_SECRET` env vars. Re-set and restart mcporter.
+- **Failure — "permission denied"**: Add `docx:document` scope at `https://open.feishu.cn/app/{APP_ID}/security`.
+- **Failure — "mcporter: command not found"**: Run `npm install -g mcporter` and retry.
+
+---
+
+## Post-Setup Summary
+
+After all smoke tests pass, output:
+
+> 飞书集成配置完成！
+>
+> - CLI 认证：已通过 ✓
+> - MCP 连接：已通过 ✓
+> - 权限配置：已完成 ✓
+>
+> 现在可以使用以下功能：
+> - 发消息、管理日历、查看任务（CLI）
+> - 创建/编辑文档、精确 Block 操作（MCP）
+> - 搜索文档、管理知识库、上传文件
+>
+> 详细用法参考 SKILL.md（路由决策）或 AGENTS.md（完整工具手册）。
diff --git a/README.md b/README.md
index bdba612..972af3f 100644
--- a/README.md
+++ b/README.md
@@ -1,265 +1,248 @@
-# feishu-mcp
+# feishu-mcp — 飞书全能集成套件
-飞书全能 MCP — 让 AI 助手完整操作你的飞书：创建文档、编辑内容、插入图表、管理表格。
+**让 AI 助手完整操作你的飞书**
-专为 **Moltbot/Clawdbot** 用户设计，也适用于 Cursor、Claude Desktop、Windsurf 等 AI 工具。
+一条命令安装，用自然语言控制飞书的消息、日历、文档、表格、任务、邮件等全部功能。
-## ✨ 能做什么？
+---
-### 📄 文档操作
-- **创建文档** — "帮我新建一个项目周报文档"
-- **编辑内容** — "在文档开头加一段摘要"
-- **读取内容** — "总结一下这篇文档的要点"
-- **批量更新** — "把所有标题改成粗体"
+## What can AI do? — 你可以让 AI 做什么
-### 📊 表格与图表
-- **创建表格** — "插入一个三列的对比表"
-- **Mermaid 图表** — "画一个用户注册流程图"
-- **思维导图** — "把这些要点整理成思维导图"
+安装完成后，你可以在 AI 对话中直接说出以下需求，AI 会自动帮你完成：
-### 🖼️ 多媒体
-- **插入图片** — "把这张截图加到文档里"
-- **画板内容** — "获取画板里的流程图"
+### 消息与沟通
+- "给张三发条消息，提醒他明天开会"
+- "搜索聊天记录中关于项目上线的讨论"
-### 📚 知识管理
-- **搜索文档** — "找一下上周的会议记录"
-- **浏览文件夹** — "列出项目文档文件夹的内容"
-- **知识库** — "在 Wiki 里创建一个新页面"
+### 日历与日程
+- "看看我今天有什么会议"
+- "帮我约一个明天下午 3 点的会议，邀请李四"
+- "查一下王五这周有没有空"
-### 📋 多维表格
-- **创建表格** — "新建一个项目跟踪表"
-- **写入数据** — "把这些任务添加到表里"
-- **查询记录** — "查一下本周的待办事项"
+### 文档与编辑
+- "帮我新建一个项目周报文档"
+- "在文档末尾追加一段总结"
+- "把这个标题改成红色"
+- "一次性添加 5 个格式化段落"
-## 🚀 快速开始
+### 表格与数据
+- "创建一个销售数据电子表格"
+- "查询项目跟踪表中本周的记录"
+- "把这些数据写入表格"
-### 1. 准备飞书应用
+### 任务与待办
+- "创建一个任务：下周五前完成设计评审"
+- "看看我还有哪些未完成的任务"
-去 [飞书开放平台](https://open.feishu.cn/app) 创建自建应用：
+### 邮件
+- "给团队发一封项目进度邮件"
+- "查看最近的收件箱"
+- "搜索关于合同的邮件"
-1. 创建应用，拿到 **App ID** 和 **App Secret**
-2. 开通权限（见下方权限列表）
-3. 安全设置 → 重定向 URL 添加：`http://localhost:3333/callback`
-4. 发布应用
+### 知识库与搜索
+- "列出所有知识空间"
+- "搜索关于 Q1 的文档"
+- "在技术文档知识库里创建一个新页面"
-### 2. 一键配置
+### 会议与妙记
+- "看看昨天有哪些会议记录"
+- "获取这个妙记的 AI 总结"
-```bash
-curl -fsSL https://raw.githubusercontent.com/AlexAnys/feishu-mcp/main/setup.sh | bash
-```
+### 画板与图表
+- "创建一个新画板"
+- "在画板上画一个用户注册流程图"
-按提示输入 App ID 和 App Secret，脚本会自动完成所有配置。
+### 通讯录
+- "搜索张三的联系方式"
+- "查一下产品部有哪些人"
-### 3. 开始使用
-
-在 Moltbot 对话中直接说：
-> "帮我在飞书创建一个新文档，标题是《项目周报》"
+### 云空间文件管理
+- "上传这个 PDF 到飞书云空间"
+- "创建一个新文件夹叫项目资料"
---
-## 📦 包含的 MCP 服务
+## Quick Start — 一键安装
-本 Skill 整合了两个 MCP，发挥各自优势：
+### 第一步：准备飞书应用
-| MCP | 用途 | 服务名 |
-|-----|------|--------|
-| [feishu-mcp](https://github.com/cso1z/Feishu-MCP) (社区版) | 文档创建/编辑、图表、图片 | `feishu` |
-| [lark-openapi-mcp](https://github.com/larksuite/lark-openapi-mcp) (官方) | 多维表格、日历、消息、搜索 | `lark` / `lark-user` |
+你需要先在飞书开放平台创建一个自建应用，获取 App ID 和 App Secret。
-## 🔑 需要的权限
+> 如果你从未使用过飞书开放平台，请查看 **[GUIDE.md — 从零开始配置飞书 AI 助手](./GUIDE.md)**，里面有详细的每一步操作说明。
-在飞书开放平台开通以下权限：
+简要步骤：
+1. 打开 [飞书开放平台](https://open.feishu.cn/app)，登录后点击「创建自建应用」
+2. 记下应用的 **App ID** 和 **App Secret**
+3. 在应用的「安全设置」中添加两个重定向 URL：
+ - `http://localhost:3000/callback`
+ - `http://localhost:3333/callback`
+4. 在「权限管理」中开通所需权限（见下方[权限列表](#permissions--需要的权限)）
+5. 发布应用（创建版本并发布）
-### 基础权限（必需）
-| 权限 | 说明 |
-|------|------|
-| `docx:document` | 读写文档 |
-| `drive:drive` | 访问云空间 |
-| `wiki:wiki` | 访问知识库 |
+### 第二步：运行安装脚本
-### 推荐权限
-| 权限 | 说明 |
-|------|------|
-| `bitable:app` | 多维表格 |
-| `im:message` | 发送消息 |
-| `search:docs:read` | 搜索文档 |
-| `calendar:calendar` | 日历管理 |
+```bash
+curl -fsSL https://raw.githubusercontent.com/AlexAnys/feishu-mcp/main/setup-v2.sh | bash
+```
-> 💡 首次调用某功能时如果提示权限不足，按提示链接开通即可。
+脚本会引导你：
+- 输入 App ID 和 App Secret
+- 自动安装飞书 CLI 和社区 MCP
+- 打开浏览器完成 OAuth 授权
+- 检测你使用的 AI 工具并生成对应配置
-## 🛠️ 可用工具一览
+整个过程大约 2-3 分钟。
-### 文档管理（feishu-mcp）
+### 第三步：开始对话
-| 工具 | 功能 |
-|------|------|
-| `create_feishu_document` | 创建新文档 |
-| `get_feishu_document_info` | 获取文档信息 |
-| `get_feishu_document_blocks` | 获取文档结构 |
-| `batch_create_feishu_blocks` | 批量创建内容块 |
-| `update_feishu_block_text` | 更新文本内容 |
-| `delete_feishu_document_blocks` | 删除内容块 |
-| `create_feishu_table` | 创建表格 |
-| `upload_and_bind_image_to_block` | 上传图片 |
-| `get_feishu_folder_files` | 获取文件夹内容 |
-| `create_feishu_folder` | 创建文件夹 |
-| `search_feishu_documents` | 搜索文档 |
-| `get_feishu_whiteboard_content` | 获取画板内容 |
+在你的 AI 工具中直接说：
-### 多维表格（lark-mcp）
+> "帮我看一下今天的飞书日程"
-| 工具 | 功能 |
-|------|------|
-| `bitable_v1_app_create` | 创建多维表格 |
-| `bitable_v1_appTable_list` | 列出数据表 |
-| `bitable_v1_appTableRecord_create` | 创建记录 |
-| `bitable_v1_appTableRecord_search` | 搜索记录 |
-| `bitable_v1_appTableRecord_update` | 更新记录 |
+如果 AI 成功返回了你的日程安排，说明安装已完成。
-### 搜索与知识库（lark-mcp）
+---
-| 工具 | 功能 |
-|------|------|
-| `docx_builtin_search` | 搜索云文档 |
-| `docx_v1_document_rawContent` | 获取文档纯文本 |
-| `wiki_v1_node_search` | 搜索知识库 |
-| `wiki_v2_space_getNode` | 获取 Wiki 节点 |
+## Supported AI Tools — 支持的 AI 工具
-## 📝 使用示例
+| AI 工具 | 支持程度 | 说明 |
+|---------|---------|------|
+| **OpenClaw** | 推荐 | 主要适配平台，安装脚本自动检测并配置 |
+| **Claude Code** | 完全支持 | 自动生成 MCP 配置文件 |
+| **Cursor** | 完全支持 | 自动生成 MCP 配置文件 |
-### 创建并编辑文档
+安装脚本会自动检测你系统上安装了哪些 AI 工具，并生成对应的配置文件到 `configs/` 目录。
-```
-用户：帮我创建一个文档，标题是"Q1 工作总结"
+---
-AI：我来创建文档...
- [调用 create_feishu_document]
- 文档已创建：https://xxx.feishu.cn/docx/xxx
-
-用户：在里面加上三个章节：目标回顾、主要成果、下季度计划
+## Architecture — 工作原理
-AI：正在添加章节...
- [调用 batch_create_feishu_blocks]
- 已添加三个章节标题，请查看文档。
-```
+如果你好奇这个套件是怎么让 AI 操作飞书的，这里做一个简单说明。
-### 插入流程图
+feishu-mcp 整合了两个工具层，它们各有所长、互相配合：
```
-用户：在文档里画一个用户注册流程图
-
-AI：我来创建 Mermaid 流程图...
- [调用 batch_create_feishu_blocks with mermaid]
-
- 已插入流程图：
- 用户访问 → 填写信息 → 验证手机 → 注册成功
+你 <-> AI 助手 (OpenClaw / Claude Code / Cursor)
+ |
+ feishu-mcp 集成层
+ |
+ +---------+---------+
+ | |
+ 飞书 CLI 社区 MCP
+ (官方命令行工具) (文档精编工具)
+ 覆盖 11 个业务域 15 个文档编辑工具
+ 消息/日历/邮件/ 精确到每个内容块
+ 任务/表格/知识库... 颜色/对齐/批量操作
```
-### 搜索并总结
+**简单理解**：
+- **飞书 CLI** 负责「广度」— 消息、日历、邮件、任务、表格、知识库等所有非文档领域都靠它
+- **社区 MCP** 负责「深度」— 需要精确编辑文档内容（比如给标题改颜色、一次创建多个格式化段落）时用它
+- AI 助手根据你的需求自动选择用哪个工具，你不需要关心这些细节
-```
-用户：找一下上周的产品会议记录，总结出行动项
-
-AI：搜索中...
- [调用 docx_builtin_search]
- 找到 3 篇相关文档，正在读取...
- [调用 docx_v1_document_rawContent]
-
- 行动项汇总：
- 1. @张三 完成原型设计 - 本周五
- 2. @李四 准备用户访谈 - 下周一
- ...
-```
+> 想了解更多技术细节？请查看 [SKILL.md](./SKILL.md)（AI 决策矩阵）和 [AGENTS.md](./AGENTS.md)（AI 操作手册）。
-### 写入多维表格
+---
-```
-用户：把这些任务添加到项目跟踪表
+## Permissions — 需要的权限
-AI：正在写入...
- [调用 bitable_v1_appTableRecord_create]
- 已添加 5 条记录到多维表格。
-```
+在飞书开放平台的「权限管理」页面中开通以下权限。
-## 🔧 手动配置（可选）
+### 基础权限（必须开通）
-
-点击展开手动配置步骤
+| 权限标识 | 说明 | 用于 |
+|---------|------|------|
+| `docx:document` | 读写云文档 | 创建和编辑文档 |
+| `drive:drive` | 访问云空间 | 管理文件和文件夹 |
+| `wiki:wiki` | 访问知识库 | 浏览和管理知识空间 |
-### 安装依赖
+### 推荐权限（按需开通）
-```bash
-npm install -g mcporter
-```
+| 权限标识 | 说明 | 用于 |
+|---------|------|------|
+| `im:message` | 发送和接收消息 | 消息和聊天功能 |
+| `calendar:calendar` | 日历管理 | 查看日程、创建会议 |
+| `bitable:app` | 多维表格 | 查询和管理多维表格数据 |
+| `search:docs:read` | 搜索文档 | 搜索云空间内容 |
+| `task:task` | 任务管理 | 创建和查看待办任务 |
+| `mail:message` | 邮件读写 | 收发邮件 |
+| `contact:user.base:readonly` | 通讯录基础信息 | 搜索员工和查看联系方式 |
+| `vc:meeting` | 视频会议 | 查询会议记录 |
+| `minutes:minute` | 妙记 | 获取会议纪要和 AI 总结 |
+| `sheets:spreadsheet` | 电子表格 | 创建和编辑表格 |
-### 配置文件
-
-编辑 `config/mcporter.json`：
-
-```json
-{
- "mcpServers": {
- "feishu": {
- "command": "npx",
- "args": ["-y", "feishu-mcp@latest", "--stdio"],
- "env": {
- "FEISHU_APP_ID": "cli_你的AppID",
- "FEISHU_APP_SECRET": "你的AppSecret",
- "FEISHU_AUTH_TYPE": "user"
- }
- },
- "lark": {
- "command": "npx",
- "args": ["-y", "@larksuiteoapi/lark-mcp", "mcp", "-a", "cli_你的AppID", "-s", "你的AppSecret"]
- },
- "lark-user": {
- "command": "npx",
- "args": ["-y", "@larksuiteoapi/lark-mcp", "mcp", "-a", "cli_你的AppID", "-s", "你的AppSecret", "--oauth", "--token-mode", "user_access_token"]
- }
- }
-}
-```
+> 不需要一次全部开通。首次使用某个功能时如果提示权限不足，再去开通即可。
-### 用户授权
+---
-```bash
-# 社区版授权
-npx feishu-mcp@latest --feishu-app-id="cli_xxx" --feishu-app-secret="secret" --feishu-auth-type="user"
+## FAQ — 常见问题
-# 官方版授权
-npx -y @larksuiteoapi/lark-mcp login -a "cli_xxx" -s "secret"
-```
+### 安装相关
+
+**Q: 安装脚本需要什么环境？**
+
+需要 Node.js 16 或更高版本和 npm。macOS 和 Linux 系统都支持。不需要管理员权限（sudo）。
+
+**Q: 安装失败怎么办？**
-
+脚本在每一步都会显示中文错误提示和修复建议。如果遇到问题，可以重新运行脚本（它是安全的，可以反复执行），或查看 [GUIDE.md](./GUIDE.md) 中的常见问题部分。
-## ❓ 常见问题
+**Q: 可以重新运行安装脚本吗？**
-### Q: tenant 和 user 认证有什么区别？
+可以。脚本设计为可重复运行（幂等），不会重复安装已有的组件，也不会覆盖你已有的配置。
+
+### 认证相关
+
+**Q: tenant 和 user 认证有什么区别？**
| 认证类型 | 身份 | 适用场景 |
|----------|------|----------|
-| `tenant` | 应用 | 操作应用创建的资源 |
-| `user` | 用户本人 | 访问个人文档（推荐）|
+| `tenant`（应用身份） | 以应用名义操作 | 操作应用自己创建的资源 |
+| `user`（用户身份） | 以你本人名义操作 | 访问你自己的文档、日程等 |
+
+**强烈建议使用 user 认证**，这样 AI 操作的飞书内容会显示为你本人的操作，功能也更完整。安装脚本默认使用 user 认证。
+
+**Q: 授权过期了怎么办？**
+
+AI 工具通常会自动刷新授权。如果确实过期了，你可以说「重新登录飞书」，AI 会帮你重新完成授权。
-**强烈建议使用 user 认证**，功能更完整。
+### 使用相关
-### Q: 权限不足怎么办？
+**Q: 文档链接里的 token 怎么获取？**
-按错误提示的链接开通权限，或在开放平台手动添加。
+飞书文档链接格式：`https://xxx.feishu.cn/docx/AbCdEfGhIjKl`
+最后一段 `AbCdEfGhIjKl` 就是 document_token。你也可以直接把链接发给 AI，它会自动提取。
-### Q: 文档链接怎么获取 token？
+**Q: 支持哪些飞书功能？**
+
+目前覆盖 11 个业务域：消息、日历、文档、云空间、电子表格、多维表格、任务、邮件、通讯录、知识库、会议/妙记，以及画板和事件订阅。完整的功能列表见 [SKILL.md](./SKILL.md)。
+
+**Q: AI 是怎么决定用哪个工具的？**
+
+feishu-mcp 自带一个「决策矩阵」（[SKILL.md](./SKILL.md)），AI 会根据你的需求自动路由到合适的工具。你只需要用自然语言描述需求，不需要关心底层实现。
+
+---
+
+## Related Docs — 其他文档
+
+| 文档 | 说明 | 适合谁看 |
+|------|------|---------|
+| [GUIDE.md](./GUIDE.md) | 从零开始的安装配置图文指南 | 第一次使用飞书开放平台的用户 |
+| [SKILL.md](./SKILL.md) | AI 决策矩阵 — 每个操作对应哪个工具 | 好奇 AI 怎么选择工具的用户 |
+| [AGENTS.md](./AGENTS.md) | AI 操作完整手册 | AI 开发者 / 高级用户 |
+| [QUICKSTART-AGENT.md](./QUICKSTART-AGENT.md) | AI 自动安装指南 | AI 代理自动化使用 |
+| [TROUBLESHOOT-AGENT.md](./TROUBLESHOOT-AGENT.md) | 错误诊断与恢复 | 遇到问题时参考 |
+
+---
-飞书文档链接格式：`https://xxx.feishu.cn/docx/AbCdEfGhIjKl`
-最后一段 `AbCdEfGhIjKl` 就是 document_token。
+## Contributing — 参与贡献
-## 🔗 相关资源
+欢迎提交 Issue 和 Pull Request。
-- [feishu-mcp (社区版)](https://github.com/cso1z/Feishu-MCP) — 文档编辑能力
-- [lark-openapi-mcp (官方)](https://github.com/larksuite/lark-openapi-mcp) — 官方 API
-- [moltbot-feishu](https://github.com/AlexAnys/moltbot-feishu) — 飞书消息通道插件
-- [飞书开放平台](https://open.feishu.cn)
-- [Moltbot](https://github.com/moltbot/moltbot)
+- **报告问题**：在 [Issues](https://github.com/AlexAnys/feishu-mcp/issues) 中描述你遇到的问题
+- **功能建议**：同样在 Issues 中提出
+- **代码贡献**：Fork 后提交 PR，请遵循项目现有的代码风格
-## 📜 License
+## License
MIT
diff --git a/SKILL.md b/SKILL.md
index c19765a..8368734 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -1,75 +1,544 @@
---
name: feishu-mcp
-description: 飞书全能 MCP - 创建/编辑文档、插入图表、管理表格、搜索知识库。整合社区版+官方 MCP，适配 Moltbot/Clawdbot。
-metadata: {"clawdbot":{"emoji":"📘","requires":{"bins":["mcporter"]}}}
+description: 飞书全能集成套件 — CLI（19 技能、11 业务域、200+ 命令）+ 社区 MCP（15 文档编辑工具、Block 级精度）统一路由决策矩阵，让 AI Agent 精准选择每个飞书操作的最佳工具。
+metadata:
+ clawdbot:
+ emoji: "📘"
+ requires:
+ bins:
+ - mcporter
---
-# 飞书全能 MCP
+# 飞书全能集成套件 — 决策矩阵
-整合社区版 feishu-mcp（文档编辑）+ 官方 lark-mcp（多维表格/搜索），让 Clawdbot 完整操作飞书。
+整合飞书 CLI（`@larksuite/cli`，19 AI 技能，200+ 命令，覆盖 11 业务域）和社区 MCP（`feishu-mcp`，15 文档编辑工具，Block 级精度），为 AI Agent 提供统一路由决策。
-## 快速配置
+---
-```bash
-curl -fsSL https://raw.githubusercontent.com/AlexAnys/feishu-mcp/main/setup.sh | bash
-```
+## Decision Principles — 决策原则
-## 三个 MCP 服务
+按顺序应用以下规则，选择第一条匹配的规则：
-| 服务 | 来源 | 能力 |
-|------|------|------|
-| `feishu` | 社区版 | 创建/编辑文档、图表、图片 |
-| `lark` | 官方 | 多维表格、发消息 |
-| `lark-user` | 官方 | 搜索文档、读取知识库 |
+1. **非文档域一律用 CLI** — 如果任务涉及消息、日历、邮件、任务、通讯录、会议、妙记、电子表格、多维表格、Wiki 管理 或 事件订阅，使用 CLI。
+2. **精确 Block 编辑用 MCP** — 如果任务需要精确的 Block 级文档编辑（指定颜色、对齐方式、批量创建格式化 Block、更新特定 Block 的样式），使用社区 MCP。
+3. **CLI 和 MCP 都能做时优先 CLI** — 如果两层都能完成任务，优先使用 CLI，除非 MCP 在精度上有明确优势。
+4. **图表和画板编辑用 CLI** — Mermaid 图表、高级图表 DSL、画板创建/编辑仅 CLI 支持（MCP 仅能读取画板内容）。
+5. **多步工作流可组合两层** — 如果任务需要先查询数据（CLI）再生成格式化文档（MCP），按步骤组合使用。
-## 主要工具
+---
-### 文档操作 (feishu)
-```bash
-# 创建文档
-mcporter call feishu.create_feishu_document --args '{"title":"新文档"}'
+## Quick Decision Table — 快速决策表
-# 添加内容
-mcporter call feishu.batch_create_feishu_blocks --args '{"document_id":"xxx","blocks":[...]}'
+| User Intent (用户意图) | Tool Layer | Specific Tool/Command | Notes |
+|---|---|---|---|
+| 查看今天的日程 | CLI | `lark-cli calendar +agenda` | 快速概览今日行程 |
+| 创建日程/会议 | CLI | `lark-cli calendar +create` | 支持邀请参会人 |
+| 查询某人忙闲状态 | CLI | `lark-cli calendar +freebusy` | 查询主日历忙闲 |
+| 推荐空闲会议时段 | CLI | `lark-cli calendar +suggestion` | 多方案推荐 |
+| 发送飞书消息 | CLI | `lark-cli im +send` | 支持文本/富文本/附件 |
+| 搜索聊天记录 | CLI | `lark-cli im +search` | 按关键词搜索 |
+| 从 Markdown 创建文档 | CLI | `lark-cli doc +create` | Markdown 输入最简单 |
+| 搜索云空间文档 | CLI | `lark-cli doc +search` | 官方接口更可靠 |
+| 更新文档内容（追加/覆盖） | CLI | `lark-cli docs +update` | 支持 append/overwrite/replace/insert/delete 模式 |
+| 在文档中插入图片 | CLI | `lark-cli docs +media-insert` | CLI 更健壮 |
+| 创建文档（指定文件夹） | MCP | `create_feishu_document` | 可指定 folderToken |
+| 获取文档元信息 | MCP | `get_feishu_document_info` | 返回标题、修改时间等 |
+| 获取文档 Block 结构 | MCP | `get_feishu_document_blocks` | 返回完整 Block 树 |
+| 提取文档纯文本 | MCP | `get_feishu_document_text` | 纯文本提取 |
+| 批量创建格式化 Block | MCP | `batch_create_feishu_blocks` | 原子操作，支持 text/code/heading/list |
+| 更新 Block 文本及样式 | MCP | `update_feishu_block_text` | 8 色彩、对齐方式 |
+| 删除文档中的 Block | MCP | `delete_feishu_document_blocks` | 按 block_id 删除 |
+| 上传图片并绑定到 Block | MCP | `upload_and_bind_image_to_block` | 图片上传 + 绑定一步完成 |
+| 在文档中创建表格 | MCP | `create_feishu_table` | 指定行列数 |
+| 创建文件夹 | MCP | `create_feishu_folder` | 云空间文件夹 |
+| 列出文件夹内容 | MCP | `list_feishu_folder_contents` | 浏览文件夹结构 |
+| 获取文件夹文件列表 | MCP | `get_feishu_folder_files` | 文件夹内文件清单 |
+| 搜索文档（MCP 层） | MCP | `search_feishu_documents` | 社区 MCP 搜索 |
+| 获取 Wiki 节点信息 | MCP | `get_feishu_wiki_node_info` | Wiki 元数据 |
+| 获取画板内容 | MCP | `get_feishu_whiteboard_content` | 仅读取，不可编辑 |
+| 上传文件到云空间 | CLI | `lark-cli drive upload` | 支持多种文件类型 |
+| 下载云空间文件 | CLI | `lark-cli drive download` | 下载到本地 |
+| 创建电子表格 | CLI | `lark-cli sheets +create` | 创建并写入表头 |
+| 读取电子表格数据 | CLI | `lark-cli sheets +read` | 读取单元格范围 |
+| 写入电子表格数据 | CLI | `lark-cli sheets +write` | 批量写入单元格 |
+| 查询多维表格记录 | CLI | `lark-cli base +query` | 支持筛选/排序 |
+| 创建多维表格 | CLI | `lark-cli base +create` | 创建表 + 字段 |
+| 创建待办任务 | CLI | `lark-cli task +create` | 支持分配/截止日期 |
+| 查看我的任务列表 | CLI | `lark-cli task +get-my-tasks` | 个人待办 |
+| 写邮件并发送 | CLI | `lark-cli mail +compose` | 支持抄送/附件 |
+| 查看收件箱 | CLI | `lark-cli mail +inbox` | 最近邮件列表 |
+| 搜索邮件 | CLI | `lark-cli mail +search` | 按关键词搜索 |
+| 搜索员工/联系人 | CLI | `lark-cli contact +search` | 姓名/邮箱/手机号 |
+| 列出 Wiki 知识空间 | CLI | `lark-cli wiki +list` | 浏览知识空间 |
+| 在 Wiki 中创建文档 | CLI | `lark-cli wiki +create` | 知识库节点创建 |
+| 创建画板 | CLI | `lark-cli whiteboard +create` | 新建飞书画板 |
+| 编辑画板内容 | CLI | `lark-cli docs +whiteboard-update` | Mermaid/图表/视觉编辑 |
+| 查看认证状态 | CLI | `lark-cli auth status` | 检查 token 是否有效 |
+| 登录认证 | CLI | `lark-cli auth login` | OAuth 浏览器认证 |
+| Check my schedule for tomorrow | CLI | `lark-cli calendar +agenda` | English: same command |
+| Create a document from notes | CLI | `lark-cli doc +create` | English: Markdown input |
+| Search for meeting minutes | CLI | `lark-cli doc +search` | English: keyword search |
+| 生成格式化报告（查询 + 文档） | CLI+MCP | `lark-cli base +query` → `batch_create_feishu_blocks` | 多步：先查数据再写文档 |
+| 整理会议纪要到文档 | CLI+MCP | `lark-cli vc` → `batch_create_feishu_blocks` | 多步：获取纪要再格式化 |
-# 插入表格
-mcporter call feishu.create_feishu_table --args '{"document_id":"xxx","rows":3,"cols":3}'
-```
+---
+
+## Domain Sections — 业务域详解
+
+### Auth/Identity — 认证与身份
+
+**Skill**: `lark-shared` — 应用配置初始化、认证登录、身份切换、权限管理。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli config init --new` | 初始化应用配置 |
+| `lark-cli auth login --recommend` | OAuth 浏览器登录 |
+| `lark-cli auth status` | 查看当前认证状态 |
+| `lark-cli auth check` | 检查 token 有效性 |
+
+**示例 Prompt：**
+- "检查一下飞书认证状态" → `lark-cli auth status`
+- "重新登录飞书" → `lark-cli auth login --recommend`
+
+**路由说明：** CLI only。社区 MCP 认证通过环境变量，不提供交互式认证命令。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Calendar — 日历
+
+**Skill**: `lark-calendar` — 提供日历与日程的全面管理：查看/搜索日程、创建/更新日程、管理参会人、查询忙闲状态及推荐空闲时段。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli calendar +agenda` | 快速概览今日/近期行程 |
+| `lark-cli calendar +create` | 创建日程并邀请参会人 |
+| `lark-cli calendar +freebusy` | 查询用户主日历忙闲信息 |
+| `lark-cli calendar +suggestion` | 提供多个时间推荐方案 |
+
+**示例 Prompt：**
+- "看一下今天有什么会议" → `lark-cli calendar +agenda`
+- "帮我约一个明天下午的会议" → `lark-cli calendar +create`
+- "查一下张三这周的忙闲状态" → `lark-cli calendar +freebusy`
+
+**路由说明：** CLI only。社区 MCP 不支持日历操作。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Messaging — 即时通讯
+
+**Skill**: `lark-im` — 收发消息和管理群聊：发送/回复消息、搜索聊天记录、管理群聊成员、上传下载文件/图片、管理表情回复。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli im +send` | 发送消息（文本/富文本/附件） |
+| `lark-cli im +search` | 搜索聊天记录 |
+
+**示例 Prompt：**
+- "给张三发条消息说明天开会" → `lark-cli im +send`
+- "搜索最近关于项目上线的聊天记录" → `lark-cli im +search`
+
+**路由说明：** CLI only。社区 MCP 不支持消息操作。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Documents — 文档操作
+
+**Skill**: `lark-doc` — 创建和编辑飞书文档：从 Markdown 创建、获取内容、更新（追加/覆盖/替换/插入/删除）、上传图片和文件、搜索文档。此域 CLI 和 MCP 重叠最多，需仔细路由。
+
+**CLI 命令：**
+
+| 命令 | 描述 |
+|---|---|
+| `lark-cli doc +create` | 从 Markdown 创建文档 |
+| `lark-cli doc +search` | 搜索云空间文档 |
+| `lark-cli docs +update` | 更新文档（append/overwrite/replace/insert/delete） |
+| `lark-cli docs +media-insert` | 在文档中插入图片/文件 |
+
+**MCP 工具：**
+
+| 工具 | 描述 |
+|---|---|
+| `create_feishu_document` | 创建文档（可指定文件夹） |
+| `get_feishu_document_info` | 获取文档元信息 |
+| `get_feishu_document_blocks` | 获取文档 Block 结构树 |
+| `get_feishu_document_text` | 提取文档纯文本 |
+| `batch_create_feishu_blocks` | 批量创建 Block（text/code/heading/list），原子操作 |
+| `update_feishu_block_text` | 更新 Block 文本及样式（8 色彩、对齐方式） |
+| `delete_feishu_document_blocks` | 按 block_id 删除 Block |
+| `upload_and_bind_image_to_block` | 上传图片并绑定到 Block |
+| `create_feishu_table` | 在文档中创建表格 |
+
+**示例 Prompt：**
+- "帮我新建一个项目周报文档" → `lark-cli doc +create`
+- "搜索一下上周的会议纪要" → `lark-cli doc +search`
+- "在文档末尾追加一段总结" → `lark-cli docs +update` (append 模式)
+- "在文档里插入截图" → `lark-cli docs +media-insert`
+- "获取这篇文档的所有 Block 结构" → `get_feishu_document_blocks`
+- "给文档的第二个标题改成红色" → `update_feishu_block_text`
+- "一次性添加 5 个格式化段落" → `batch_create_feishu_blocks`
+- "删除文档中指定的 Block" → `delete_feishu_document_blocks`
+- "Create a new doc titled Q1 Summary" → `lark-cli doc +create`
+
+**路由说明：**
+- 简单创建/搜索/追加/覆盖 → 用 CLI（更简洁）
+- 需要精确 Block 操作（指定颜色、对齐、批量格式化 Block）→ 用 MCP
+- 需要读取 Block 结构树 → 用 MCP（`get_feishu_document_blocks`）
+- 需要提取纯文本 → 用 MCP（`get_feishu_document_text`）
+- 如果此技能不可用，检查：`lark-cli skills list`
+
+---
+
+### Cloud Storage — 云空间
+
+**Skill**: `lark-drive` — 管理云空间中的文件和文件夹：上传/下载文件、创建文件夹、复制/移动/删除文件、管理文档评论和权限。
+
+**CLI 命令：**
+
+| 命令 | 描述 |
+|---|---|
+| `lark-cli drive upload` | 上传文件到云空间 |
+| `lark-cli drive download` | 下载文件到本地 |
+
+**MCP 工具：**
+
+| 工具 | 描述 |
+|---|---|
+| `create_feishu_folder` | 创建文件夹 |
+| `list_feishu_folder_contents` | 列出文件夹内容 |
+| `get_feishu_folder_files` | 获取文件夹文件列表 |
+
+**示例 Prompt：**
+- "上传这个 PDF 到飞书云空间" → `lark-cli drive upload`
+- "下载这个文件" → `lark-cli drive download`
+- "创建一个新文件夹叫项目资料" → `create_feishu_folder`
+- "列出项目文档文件夹的内容" → `list_feishu_folder_contents`
+
+**路由说明：** 上传/下载用 CLI；文件夹管理 CLI 和 MCP 都可用，CLI 更完整。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Spreadsheets — 电子表格
+
+**Skill**: `lark-sheets` — 创建和操作电子表格：创建表格并写入数据、读取和写入单元格、追加行、查找内容、导出文件。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli sheets +create` | 创建电子表格并写入表头 |
+| `lark-cli sheets +write` | 批量写入单元格 |
+| `lark-cli sheets +read` | 读取单元格范围 |
+
+**示例 Prompt：**
+- "创建一个销售数据表" → `lark-cli sheets +create`
+- "把这些数据写入表格" → `lark-cli sheets +write`
+- "读取 A1:D10 的数据" → `lark-cli sheets +read`
+
+**路由说明：** CLI only。社区 MCP 不支持电子表格操作。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Bitable — 多维表格
+
+**Skill**: `lark-base` — 管理多维表格：创建表/字段/记录、查询数据、配置视图、管理仪表盘。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli base +query` | 查询多维表格记录（支持筛选/排序） |
+| `lark-cli base +create` | 创建多维表格 |
+
+**示例 Prompt：**
+- "查询项目表中本周的记录" → `lark-cli base +query`
+- "新建一个 Bug 追踪多维表格" → `lark-cli base +create`
+
+**路由说明：** CLI only。社区 MCP 不支持多维表格。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Tasks — 任务
+
+**Skill**: `lark-task` — 管理任务和清单：创建待办、查看/更新状态、子任务拆分、分配协作成员。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli task +create` | 创建待办任务 |
+| `lark-cli task +get-my-tasks` | 查看个人任务列表 |
+
+**示例 Prompt：**
+- "创建一个任务：下周五前完成设计评审" → `lark-cli task +create`
+- "看看我还有哪些未完成的任务" → `lark-cli task +get-my-tasks`
+
+**路由说明：** CLI only。社区 MCP 不支持任务操作。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Email — 邮件
+
+**Skill**: `lark-mail` — 飞书邮箱管理：起草/发送/回复/转发/搜索邮件，管理草稿和通讯录。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli mail +compose` | 写邮件并发送 |
+| `lark-cli mail +inbox` | 查看收件箱 |
+| `lark-cli mail +search` | 搜索邮件 |
+
+**示例 Prompt：**
+- "给团队发一封项目进度邮件" → `lark-cli mail +compose`
+- "查看最近的收件箱" → `lark-cli mail +inbox`
+- "搜索关于合同的邮件" → `lark-cli mail +search`
+
+**路由说明：** CLI only。社区 MCP 不支持邮件操作。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Contacts — 通讯录
+
+**Skill**: `lark-contact` — 查询组织架构和人员信息：搜索员工、查看个人信息、查询部门结构。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli contact +search` | 搜索员工（姓名/邮箱/手机号） |
+
+**示例 Prompt：**
+- "搜索张三的联系方式" → `lark-cli contact +search`
+- "查一下产品部有哪些人" → `lark-cli contact +search`
+
+**路由说明：** CLI only。社区 MCP 不支持通讯录操作。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Wiki — 知识库
+
+**Skill**: `lark-wiki` — 管理知识空间和文档节点：创建/查询知识空间、管理节点层级、组织文档。
+
+**CLI 命令：**
+
+| 命令 | 描述 |
+|---|---|
+| `lark-cli wiki +list` | 列出知识空间 |
+| `lark-cli wiki +create` | 在知识库中创建文档 |
+
+**MCP 工具：**
+
+| 工具 | 描述 |
+|---|---|
+| `get_feishu_wiki_node_info` | 获取 Wiki 节点元数据 |
+
+**示例 Prompt：**
+- "列出所有知识空间" → `lark-cli wiki +list`
+- "在技术文档知识库里创建一个新页面" → `lark-cli wiki +create`
+- "获取这个 Wiki 节点的元信息" → `get_feishu_wiki_node_info`
+
+**路由说明：** 管理操作（创建、列表、导航）用 CLI；仅查询节点元数据可用 MCP。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Meetings/Minutes — 会议与妙记
+
+**Skills**: `lark-vc` + `lark-minutes` — 查询会议记录、获取会议纪要产物（总结、待办、章节、逐字稿）。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli vc` | 查询视频会议记录（结束的会议） |
+| `lark-cli minutes` | 获取妙记信息（标题、时长、AI 总结） |
+
+**示例 Prompt：**
+- "看看昨天有哪些会议记录" → `lark-cli vc`
+- "获取这个妙记的 AI 总结" → `lark-cli minutes`
+
+**路由说明：** CLI only。社区 MCP 不支持会议操作。查未来日程用 `lark-cli calendar +agenda`，查已结束会议用 `lark-cli vc`。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Whiteboard — 画板
+
+**Skill**: `lark-whiteboard` — 创建和编辑飞书画板：Mermaid 图表、流程图、架构图、思维导图。
+
+**CLI 命令：**
+
+| 命令 | 描述 |
+|---|---|
+| `lark-cli whiteboard +create` | 创建新画板 |
+| `lark-cli docs +whiteboard-update` | 编辑画板（Mermaid/图表/视觉内容） |
+
+**MCP 工具：**
+
+| 工具 | 描述 |
+|---|---|
+| `get_feishu_whiteboard_content` | 读取画板内容（只读） |
+
+**示例 Prompt：**
+- "创建一个新画板" → `lark-cli whiteboard +create`
+- "在画板上画一个流程图" → `lark-cli docs +whiteboard-update`
+- "获取这个画板的内容" → `get_feishu_whiteboard_content`
+
+**路由说明：** 创建和编辑用 CLI（MCP 不支持写入）；读取画板内容 CLI 和 MCP 均可，MCP 返回结构化数据。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Events — 事件订阅
+
+**Skill**: `lark-event` — 通过 WebSocket 实时监听飞书事件（消息变更、通讯录变更、日历变更等）。
+
+| 命令/工具 | 描述 |
+|---|---|
+| `lark-cli event` | WebSocket 事件订阅，输出 NDJSON |
+
+**示例 Prompt：**
+- "监听飞书的新消息事件" → `lark-cli event`
+- "订阅日历变更事件" → `lark-cli event`
+
+**路由说明：** CLI only。社区 MCP 不支持事件订阅。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+### Search — 文档搜索
+
+跨域搜索飞书文档和知识库内容。
+
+**CLI 命令：**
+
+| 命令 | 描述 |
+|---|---|
+| `lark-cli doc +search` | 搜索云空间文档（官方接口） |
+
+**MCP 工具：**
+
+| 工具 | 描述 |
+|---|---|
+| `search_feishu_documents` | 搜索文档（社区 MCP） |
+
+**示例 Prompt：**
+- "搜索关于 Q1 的文档" → `lark-cli doc +search`
+- "Find documents about onboarding" → `lark-cli doc +search`
+
+**路由说明：** 优先使用 CLI（`doc +search`），官方接口更可靠。如果 CLI 不可用，使用 MCP（`search_feishu_documents`）。如果此技能不可用，检查：`lark-cli skills list`。
+
+---
+
+## Additional CLI Skills — 扩展技能
+
+以下技能提供高级工作流和 API 探索能力：
+
+| 技能 | 描述 |
+|---|---|
+| `lark-openapi-explorer` | 从官方文档库中挖掘未封装的原生 OpenAPI 接口 |
+| `lark-skill-maker` | 创建自定义 lark-cli Skill |
+| `lark-workflow-meeting-summary` | 汇总会议纪要并生成结构化报告 |
+| `lark-workflow-standup-report` | 生成日程 + 未完成任务摘要 |
+
+---
+
+## Overlap Resolution — 重叠能力路由
+
+当 CLI 和 MCP 都能完成某个操作时，参考此表选择：
+
+| Capability | CLI Tool | MCP Tool | Recommended | Reason |
+|---|---|---|---|---|
+| 文档创建 | `lark-cli doc +create` | `create_feishu_document` | CLI | Markdown 输入更简洁，无需指定 folderToken |
+| 文档文本读取 | `lark-cli docs +update` (read) | `get_feishu_document_text` | Depends | 需要纯文本提取用 MCP；需要 Markdown 格式用 CLI |
+| Block 编辑/更新 | `lark-cli docs +update` | `update_feishu_block_text` | Depends | 需要指定颜色或对齐方式用 MCP；简单文本替换用 CLI |
+| Block 删除 | `lark-cli docs +update` (delete) | `delete_feishu_document_blocks` | Depends | 需要按 block_id 精确删除用 MCP；按位置删除用 CLI |
+| 图片上传到文档 | `lark-cli docs +media-insert` | `upload_and_bind_image_to_block` | CLI | CLI 更健壮，错误处理更完善 |
+| 文档搜索 | `lark-cli doc +search` | `search_feishu_documents` | CLI | 官方接口，更可靠 |
+| Wiki 节点读取 | `lark-cli wiki +list` | `get_feishu_wiki_node_info` | Depends | 需要节点元数据详情用 MCP；浏览知识空间结构用 CLI |
+| 画板内容读取 | `lark-cli whiteboard` | `get_feishu_whiteboard_content` | Depends | 需要结构化 JSON 数据用 MCP；需要视觉编辑用 CLI |
+| 批量创建 Block | `lark-cli docs +update` (append) | `batch_create_feishu_blocks` | MCP | 原子操作，支持精确格式化（颜色、对齐、类型） |
+| 文件夹创建 | `lark-cli drive` | `create_feishu_folder` | CLI | CLI 支持更多文件夹操作 |
+
+---
+
+## Auth Prerequisites — 认证前置条件
+
+### CLI 层认证
-### 多维表格 (lark)
```bash
-# 创建表格
-mcporter call lark.bitable_v1_app_create --args '{"data":{"name":"项目表"}}'
+# 检查认证状态
+lark-cli auth status
+# 成功输出：显示当前 App ID、用户身份、token 有效期
+# 失败输出：Not logged in 或 Token expired
+```
-# 写入记录
-mcporter call lark.bitable_v1_appTableRecord_create --args '{"path":{"app_token":"xxx","table_id":"xxx"},"data":{"fields":{"标题":"任务1"}}}'
+如果未认证：
+```bash
+lark-cli config init --new # 初始化应用配置
+lark-cli auth login --recommend # OAuth 浏览器登录（端口 3000）
```
-### 搜索 (lark-user)
+### 社区 MCP 层认证
+
+检查环境变量是否设置：
```bash
-# 搜索文档
-mcporter call lark-user.docx_builtin_search --args '{"data":{"search_key":"会议","count":5}}'
+echo $FEISHU_APP_ID # 应输出 cli_xxxxxxxx
+echo $FEISHU_APP_SECRET # 应输出非空字符串
+```
-# 读取内容
-mcporter call lark-user.docx_v1_document_rawContent --args '{"path":{"document_id":"xxx"}}'
+社区 MCP OAuth 使用端口 3333（重定向 URL: `http://localhost:3333/callback`）。
-# 搜索知识库
-mcporter call lark-user.wiki_v1_node_search --args '{"data":{"query":"入职"}}'
+### 官方 MCP 层认证（可选）
+
+```bash
+npx -y @larksuiteoapi/lark-mcp login -a "$APP_ID" -s "$APP_SECRET"
```
-## 权限配置
+### Pre-flight Checklist — 操作前检查清单
+
+执行任何飞书操作前，按顺序检查：
+
+1. CLI 已安装：`which lark-cli` 应返回路径
+2. CLI 已认证：`lark-cli auth status` 应显示有效 token
+3. CLI 技能已加载：`lark-cli skills list` 应列出所需技能
+4. MCP 环境变量已设置：`FEISHU_APP_ID` 和 `FEISHU_APP_SECRET` 非空
+5. mcporter 已安装：`which mcporter` 应返回路径
+
+---
+
+## Permissions Reference — 权限参考
+
+| Domain | Required Scopes | Optional Scopes |
+|---|---|---|
+| Documents | `docx:document` | `docx:document:readonly` |
+| Cloud Storage | `drive:drive` | `drive:drive:readonly` |
+| Wiki | `wiki:wiki` | `wiki:wiki:readonly` |
+| Bitable | `bitable:app` | `bitable:app:readonly` |
+| Messaging | `im:message` | `im:message:readonly`, `im:chat:readonly` |
+| Search | `search:docs:read` | — |
+| Calendar | `calendar:calendar` | `calendar:calendar:readonly` |
+| Email | `mail:message` | `mail:message:readonly` |
+| Contacts | `contact:user.base:readonly` | `contact:department:readonly` |
+| Meetings | `vc:meeting:readonly` | — |
+| Minutes | `minutes:minute:readonly` | — |
+| Tasks | `task:task` | `task:task:readonly` |
+| Events | 按需订阅事件类型 | — |
+
+飞书权限配置地址：`https://open.feishu.cn/app/{APP_ID}/security`
+
+重定向 URL 配置：
+- CLI: `http://localhost:3000/callback`
+- 社区 MCP: `http://localhost:3333/callback`
+
+---
+
+## Official MCP (Optional) — 官方 MCP（可选）
+
+官方 MCP（`@larksuiteoapi/lark-mcp`，Beta）作为高级可选项。默认安装不包含。
+
+如果安装了官方 MCP，以下路由可能变化：
+
+| Capability | Default (CLI) | With Official MCP |
+|---|---|---|
+| 多维表格操作 | `lark-cli base` | 可用 `bitable_v1_*` 工具 |
+| 文档搜索 | `lark-cli doc +search` | 可用 `docx_builtin_search` |
+| Wiki 搜索 | `lark-cli wiki +list` | 可用 `wiki_v1_node_search` |
+
+官方 MCP 提供的主要工具：`bitable_v1_app_create`、`bitable_v1_appTableRecord_create`、`bitable_v1_appTableRecord_search`、`docx_builtin_search`、`docx_v1_document_rawContent`、`wiki_v1_node_search`、`wiki_v2_space_getNode`。
+
+如果未安装官方 MCP，决策矩阵正常工作，所有功能通过 CLI + 社区 MCP 覆盖。
+
+---
-飞书开放平台开通：
-- `docx:document` - 读写文档
-- `drive:drive` - 云空间
-- `wiki:wiki` - 知识库
-- `bitable:app` - 多维表格
-- `search:docs:read` - 搜索
+## MCP-Only Mode — 仅 MCP 模式
-重定向 URL：
-- `http://localhost:3000/callback`
-- `http://localhost:3333/callback`
+如果 CLI 未安装（`setup-v2.sh --mcp-only`），以下能力可用：
-## 详细文档
+**可用（MCP）：** 文档创建/编辑/读取、Block 操作、文件夹管理、文档搜索、Wiki 节点查询、画板内容读取。
-https://github.com/AlexAnys/feishu-mcp
+**不可用（CLI only）：** 消息、日历、邮件、任务、通讯录、会议、妙记、电子表格、多维表格、事件订阅、画板编辑、Mermaid 图表。
diff --git a/TROUBLESHOOT-AGENT.md b/TROUBLESHOOT-AGENT.md
new file mode 100644
index 0000000..a04a509
--- /dev/null
+++ b/TROUBLESHOOT-AGENT.md
@@ -0,0 +1,452 @@
+# TROUBLESHOOT-AGENT.md — Feishu Integration Error Recovery Guide
+
+**Audience**: AI agents diagnosing and recovering from Feishu integration errors.
+**Purpose**: Structured error patterns, diagnostic commands, and automated recovery steps.
+
+When an error occurs, find the matching error pattern below and execute the recovery steps in order. If no pattern matches, run Quick Diagnostics first.
+
+---
+
+## Quick Diagnostics
+
+Run these commands first to assess system health. Execute all 6 in sequence.
+
+### Diagnostic 1: CLI installation status
+
+```bash
+which lark-cli && lark-cli --version
+```
+- **Healthy**: Returns a path and version number.
+- **Unhealthy**: "command not found" → CLI not installed. See E-001.
+
+### Diagnostic 2: CLI auth status
+
+```bash
+lark-cli auth status
+```
+- **Healthy**: Shows App ID, user identity, valid token with future expiry.
+- **Unhealthy**: "Not logged in" → See E-002. "Token expired" → See E-001.
+
+### Diagnostic 3: CLI skills loaded
+
+```bash
+lark-cli skills list 2>/dev/null | head -20
+```
+- **Healthy**: Lists installed skills (e.g., lark-calendar, lark-im, lark-doc).
+- **Unhealthy**: Empty list or error → See E-011.
+
+### Diagnostic 4: MCP environment variables
+
+```bash
+echo "APP_ID: ${FEISHU_APP_ID:-NOT SET}" && echo "SECRET: $([ -n "$FEISHU_APP_SECRET" ] && echo SET || echo NOT SET)"
+```
+- **Healthy**: APP_ID shows `cli_xxx`, SECRET shows `SET`.
+- **Unhealthy**: Either shows `NOT SET` → See E-010.
+
+### Diagnostic 5: mcporter configuration
+
+```bash
+which mcporter && mcporter list 2>/dev/null
+```
+- **Healthy**: Returns mcporter path and lists configured MCP servers.
+- **Unhealthy**: "command not found" → See E-012. No servers listed → See E-010.
+
+### Diagnostic 6: Network connectivity to Feishu API
+
+```bash
+curl -s -o /dev/null -w "%{http_code}" https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal 2>/dev/null || echo "NETWORK_ERROR"
+```
+- **Healthy**: Returns HTTP status code (e.g., `400` for missing body is normal — means API is reachable).
+- **Unhealthy**: Returns `NETWORK_ERROR` or no output → See E-006.
+
+---
+
+## Auth Errors
+
+### E-001: Token expired
+
+**Pattern**: `token expired` | `token_expired` | `access token is expired`
+**Layer**: CLI
+**Cause**: OAuth access token has expired (typically after ~2 hours).
+**Diagnosis**:
+```bash
+lark-cli auth status
+```
+Confirm output mentions "expired" or shows a past expiry time.
+**Recovery**:
+1. Run `lark-cli auth login --recommend`.
+2. Wait for user to complete OAuth in browser.
+3. Run `lark-cli auth status` to verify new token.
+**If recovery fails**: Ask the user: "飞书授权页面是否正常打开？请确认浏览器中完成了授权操作。如果页面无法加载，请检查端口 3000 是否被占用（运行 `lsof -i :3000`）。"
+
+---
+
+### E-002: Token missing / not logged in
+
+**Pattern**: `not logged in` | `no token` | `please login first` | `unauthorized`
+**Layer**: CLI
+**Cause**: User has never authenticated or auth state was cleared.
+**Diagnosis**:
+```bash
+lark-cli auth status
+```
+Confirm output shows "Not logged in" or similar.
+**Recovery**:
+1. Run `lark-cli config init --new` if no app configuration exists.
+2. Run `lark-cli auth login --recommend`.
+3. Wait for user to complete OAuth.
+4. Run `lark-cli auth status` to confirm.
+**If recovery fails**: Ask the user: "请确认你的飞书应用 App ID 和 App Secret 是否正确。可以在 https://open.feishu.cn/app 查看。"
+
+---
+
+### E-003: Invalid App ID or App Secret
+
+**Pattern**: `invalid app_id` | `invalid app_secret` | `app not found` | `10003`
+**Layer**: Both
+**Cause**: The App ID or App Secret provided does not match any Feishu application.
+**Diagnosis**:
+```bash
+echo "Current APP_ID: $FEISHU_APP_ID"
+lark-cli auth status
+```
+Check if APP_ID starts with `cli_` and has the correct format.
+**Recovery**:
+1. Ask the user to verify their App ID at https://open.feishu.cn/app.
+2. If incorrect, re-run `lark-cli config init --new` with the correct credentials.
+3. Re-set `FEISHU_APP_ID` and `FEISHU_APP_SECRET` env vars.
+4. Re-authenticate: `lark-cli auth login --recommend`.
+**If recovery fails**: Ask the user: "请确认应用已发布（不是草稿状态），并且 App ID 以 cli_ 开头。"
+
+---
+
+### E-004: Insufficient permissions (scope missing)
+
+**Pattern**: `permission denied` | `scope.*not.*authorized` | `insufficient.*scope` | `10012` | `99991672`
+**Layer**: Both
+**Cause**: The Feishu application lacks the required permission scope for the requested operation.
+**Diagnosis**:
+1. Identify the missing scope from the error message. The error usually contains the required scope name.
+2. Cross-reference with the permissions table in SKILL.md.
+**Recovery**:
+1. Identify the required scope from the error message or from this mapping:
+ - Documents → `docx:document`
+ - Cloud Storage → `drive:drive`
+ - Wiki → `wiki:wiki`
+ - Bitable → `bitable:app`
+ - Messaging → `im:message`
+ - Search → `search:docs:read`
+ - Calendar → `calendar:calendar`
+ - Email → `mail:message`
+ - Contacts → `contact:user.base:readonly`
+ - Meetings → `vc:meeting:readonly`
+ - Minutes → `minutes:minute:readonly`
+ - Tasks → `task:task`
+2. Direct the user to add the scope: `https://open.feishu.cn/app/{APP_ID}/security`.
+3. After adding the scope, the user must re-publish the app version.
+4. Re-authenticate: `lark-cli auth login --recommend`.
+5. Retry the original command.
+**If recovery fails**: Ask the user: "权限已添加但仍报错？请检查是否已重新发布应用版本。在开放平台点击「创建版本」→「发布」。"
+
+---
+
+## Network Errors
+
+### E-005: Connection refused / timeout
+
+**Pattern**: `ECONNREFUSED` | `ETIMEDOUT` | `connect ECONNREFUSED` | `request timeout`
+**Layer**: Both
+**Cause**: Cannot reach Feishu API servers. Network issue, firewall, or proxy.
+**Diagnosis**:
+```bash
+curl -s -o /dev/null -w "%{http_code}" https://open.feishu.cn 2>/dev/null || echo "UNREACHABLE"
+```
+If output is "UNREACHABLE", the network cannot reach Feishu.
+**Recovery**:
+1. Check internet connectivity: `ping -c 3 open.feishu.cn`.
+2. If behind a proxy, set HTTP_PROXY and HTTPS_PROXY env vars.
+3. If DNS fails, try `curl -s https://223.5.5.5` to test basic connectivity.
+4. Retry the original command after network is restored.
+**If recovery fails**: Ask the user: "你的网络能否访问 https://open.feishu.cn ？如果在公司网络，可能需要配置代理。"
+
+---
+
+### E-006: Proxy configuration needed
+
+**Pattern**: `ENOTFOUND` | `getaddrinfo ENOTFOUND` | `proxy.*error`
+**Layer**: Both
+**Cause**: DNS resolution fails, typically in corporate networks requiring proxy.
+**Diagnosis**:
+```bash
+echo "HTTP_PROXY: ${HTTP_PROXY:-NOT SET}" && echo "HTTPS_PROXY: ${HTTPS_PROXY:-NOT SET}"
+curl -s -o /dev/null -w "%{http_code}" https://open.feishu.cn 2>/dev/null || echo "DNS_FAIL"
+```
+**Recovery**:
+1. If proxy vars are not set and user is on corporate network, ask for proxy URL.
+2. Set proxy: `export HTTPS_PROXY="http://proxy.company.com:8080"`.
+3. Retry the original command.
+4. For npm operations, also set: `npm config set proxy $HTTP_PROXY`.
+**If recovery fails**: Ask the user: "请提供公司网络的代理服务器地址（如 http://proxy.company.com:8080）。"
+
+---
+
+## Tool-Specific Errors
+
+### E-007: Document not found (invalid document_id)
+
+**Pattern**: `document not found` | `invalid document_id` | `doc.*not.*exist` | `10001`
+**Layer**: Both
+**Cause**: The document_id does not exist or the user has no access to it.
+**Diagnosis**:
+1. Verify the document_id format. Feishu document IDs typically start with `doxcn`.
+2. If user provided a URL, extract the ID using the pattern: `/docx/([A-Za-z0-9]+)`.
+**Recovery**:
+1. Confirm the document_id with the user.
+2. If a URL was provided, extract the correct token: `https://*.feishu.cn/docx/{document_id}`.
+3. Check if the user has access to the document (shared with them or in their drive).
+4. Retry with the correct document_id.
+**If recovery fails**: Ask the user: "请确认这个文档链接是否正确，以及你是否有权限访问该文档。"
+
+---
+
+### E-008: Block not found (invalid block_id)
+
+**Pattern**: `block not found` | `invalid block_id` | `block.*not.*exist`
+**Layer**: MCP
+**Cause**: The block_id does not exist in the specified document, or the document structure changed.
+**Diagnosis**:
+```bash
+mcporter call feishu.get_feishu_document_blocks --args '{"document_id":"THE_DOC_ID"}'
+```
+List all blocks and verify the target block_id exists.
+**Recovery**:
+1. Fetch the current block structure: `get_feishu_document_blocks`.
+2. Find the correct block_id from the returned structure.
+3. Retry the operation with the correct block_id.
+**If recovery fails**: Retry after 5 seconds — the document may have been concurrently edited and block IDs may have changed.
+
+---
+
+### E-009: Folder access denied
+
+**Pattern**: `folder.*denied` | `no permission.*folder` | `folder.*forbidden`
+**Layer**: Both
+**Cause**: The user does not have permission to access the specified folder.
+**Diagnosis**:
+1. Check if `drive:drive` scope is authorized.
+2. Check if the folder is shared with the user.
+**Recovery**:
+1. Verify `drive:drive` scope is enabled at `https://open.feishu.cn/app/{APP_ID}/security`.
+2. Ask the user to check if the folder is shared with their account.
+3. If using tenant auth, the app may not have folder access — switch to user auth.
+4. Retry the operation.
+**If recovery fails**: Ask the user: "请确认你是否有权限访问此文件夹。可以在飞书云空间中查看文件夹的共享设置。"
+
+---
+
+### E-010: Rate limiting / throttle
+
+**Pattern**: `rate limit` | `too many requests` | `429` | `frequency limit`
+**Layer**: Both
+**Cause**: Too many API calls in a short time. Feishu enforces per-app rate limits.
+**Diagnosis**: Check if the error includes a retry-after header or time.
+**Recovery**:
+1. Wait 60 seconds.
+2. Retry the original command.
+3. If rate-limited again, wait 5 minutes and retry.
+4. If still failing, reduce request frequency. Batch operations where possible (use `batch_create_feishu_blocks` instead of individual block calls).
+**If recovery fails**: Wait 15 minutes and retry. Rate limits typically reset within this window.
+
+---
+
+## Setup Errors
+
+### E-011: CLI not installed
+
+**Pattern**: `command not found: lark-cli` | `lark-cli.*not found`
+**Layer**: CLI
+**Cause**: The Feishu CLI package is not installed globally.
+**Diagnosis**:
+```bash
+which lark-cli || echo "NOT INSTALLED"
+npm list -g @larksuite/cli 2>/dev/null || echo "NOT IN NPM"
+```
+**Recovery**:
+1. Run `npm install -g @larksuite/cli`.
+2. Run `npx skills add larksuite/cli -y -g`.
+3. Verify: `lark-cli --version`.
+**If recovery fails**: Check npm global directory permissions. Run `npm config get prefix` to find the global dir. If permissions are wrong, fix with `sudo chown -R $(whoami) $(npm config get prefix)/{lib/node_modules,bin,share}`.
+
+---
+
+### E-012: MCP not configured (mcporter config missing)
+
+**Pattern**: `mcporter.*not found` | `no.*mcp.*configured` | `ENOENT.*mcporter`
+**Layer**: MCP
+**Cause**: mcporter is not installed or MCP server configuration is missing.
+**Diagnosis**:
+```bash
+which mcporter || echo "NOT INSTALLED"
+mcporter list 2>/dev/null || echo "NO CONFIG"
+```
+**Recovery**:
+1. Install mcporter: `npm install -g mcporter`.
+2. Check if `FEISHU_APP_ID` and `FEISHU_APP_SECRET` are set.
+3. If not set, ask the user for their App ID and App Secret.
+4. Configure mcporter with the community MCP server.
+5. Verify: `mcporter list`.
+**If recovery fails**: Fall back to direct npx invocation: `npx -y feishu-mcp@latest --stdio`.
+
+---
+
+### E-013: Skills not loaded
+
+**Pattern**: `skill not found` | `unknown skill` | `no such command`
+**Layer**: CLI
+**Cause**: CLI skills have not been installed or loaded.
+**Diagnosis**:
+```bash
+lark-cli skills list 2>/dev/null | head -10
+```
+If output is empty or missing expected skills.
+**Recovery**:
+1. Run `npx skills add larksuite/cli -y -g`.
+2. Wait for installation to complete.
+3. Verify: `lark-cli skills list`.
+4. Retry the original command.
+**If recovery fails**: Retry once more. If still failing, ask the user: "CLI 技能安装失败。请检查网络连接后手动运行：`npx skills add larksuite/cli -y -g`。"
+
+---
+
+### E-014: Port already in use (OAuth)
+
+**Pattern**: `EADDRINUSE` | `port.*in use` | `address already in use`
+**Layer**: Both
+**Cause**: Port 3000 (CLI OAuth) or 3333 (MCP OAuth) is occupied by another process.
+**Diagnosis**:
+```bash
+lsof -i :3000 2>/dev/null | head -5
+lsof -i :3333 2>/dev/null | head -5
+```
+**Recovery**:
+1. Identify the process using the port from the `lsof` output.
+2. If the process is not critical, kill it: `kill -9 {PID}`.
+3. Retry the auth command.
+**If recovery fails**: Ask the user: "端口 3000 或 3333 被其他程序占用。请关闭占用端口的程序后重试。"
+
+---
+
+## Data Errors
+
+### E-015: Invalid JSON in request
+
+**Pattern**: `invalid json` | `JSON.*parse.*error` | `unexpected token` | `SyntaxError.*JSON`
+**Layer**: MCP
+**Cause**: The request body contains malformed JSON.
+**Diagnosis**: Review the JSON string passed to the MCP tool call. Check for unescaped quotes, missing commas, or trailing commas.
+**Recovery**:
+1. Validate the JSON with `echo '...' | python3 -m json.tool` or `echo '...' | node -e "JSON.parse(require('fs').readFileSync(0))"`.
+2. Fix the JSON syntax error.
+3. Retry with corrected JSON.
+**If recovery fails**: Simplify the request to minimal parameters and add fields incrementally.
+
+---
+
+### E-016: Required parameter missing
+
+**Pattern**: `required.*parameter` | `missing.*field` | `field.*required` | `parameter.*missing`
+**Layer**: MCP
+**Cause**: A required parameter was not provided in the tool call.
+**Diagnosis**: Read the error message to identify which parameter is missing. Cross-reference with the tool parameter list in AGENTS.md.
+**Recovery**:
+1. Identify the missing parameter from the error message.
+2. Check AGENTS.md for the tool's required parameters.
+3. Add the missing parameter to the request.
+4. Retry.
+**If recovery fails**: Consult AGENTS.md tool inventory for the complete parameter list of the tool.
+
+---
+
+### E-017: Document ID format error
+
+**Pattern**: `invalid.*token` | `malformed.*id` | `document_id.*format`
+**Layer**: Both
+**Cause**: The document/folder/wiki token does not match the expected format.
+**Diagnosis**: Feishu tokens have specific prefixes:
+- Document: typically `doxcn...`
+- Spreadsheet: typically contains alphanumeric chars
+- Wiki: typically `wikcn...`
+- Folder: typically `fldcn...`
+**Recovery**:
+1. If the user provided a full URL, extract the token using patterns from AGENTS.md "Extracting IDs from Feishu URLs" section.
+2. Verify the extracted token format.
+3. Retry with the correct token.
+**If recovery fails**: Ask the user: "请提供完整的飞书文档链接，我来提取正确的 token。"
+
+---
+
+## Permission Error Handling — Scope Reference
+
+When a permission error occurs, use this section to identify and resolve the missing scope.
+
+### Identifying the missing scope
+
+1. Read the error message. Feishu permission errors typically include the required scope name or an error code.
+2. Common error codes for permission issues: `10012`, `99991672`, `99991663`.
+3. Map the operation domain to the required scope:
+
+| Domain | Required Scope | Error Hint |
+|---|---|---|
+| Documents | `docx:document` | "docx" in error |
+| Cloud Storage | `drive:drive` | "drive" in error |
+| Wiki | `wiki:wiki` | "wiki" in error |
+| Bitable | `bitable:app` | "bitable" in error |
+| Messaging | `im:message` | "im" or "message" in error |
+| Search | `search:docs:read` | "search" in error |
+| Calendar | `calendar:calendar` | "calendar" in error |
+| Email | `mail:message` | "mail" in error |
+| Contacts | `contact:user.base:readonly` | "contact" in error |
+| Meetings | `vc:meeting:readonly` | "vc" or "meeting" in error |
+| Minutes | `minutes:minute:readonly` | "minutes" in error |
+| Tasks | `task:task` | "task" in error |
+
+### Resolution procedure
+
+1. Identify the missing scope from the table above.
+2. Direct the user to the permissions page: `https://open.feishu.cn/app/{APP_ID}/security`.
+3. Instruct the user to search for and enable the required scope.
+4. After adding the scope, the user must create a new app version and publish it.
+5. Re-authenticate: `lark-cli auth login --recommend`.
+6. Retry the original operation.
+
+### Scope activation check
+
+Some scopes require admin approval. If the user adds a scope but still gets permission errors:
+
+1. Check if the scope status is "已开通" (activated) vs "审核中" (pending approval).
+2. If pending, the user must wait for admin approval or ask their Feishu workspace admin to approve.
+3. Direct the user to contact their admin with the specific scope name.
+
+---
+
+## Escalation Summary
+
+| Error | Can Auto-Recover? | Human Action Required |
+|---|---|---|
+| E-001 Token expired | Yes (re-login) | User approves OAuth in browser |
+| E-002 Not logged in | Yes (login) | User approves OAuth in browser |
+| E-003 Invalid credentials | No | User verifies App ID/Secret |
+| E-004 Missing scope | No | User adds scope in Feishu console |
+| E-005 Network timeout | Yes (retry) | None if network recovers |
+| E-006 Proxy needed | No | User provides proxy URL |
+| E-007 Doc not found | Partial (URL extraction) | User confirms doc access |
+| E-008 Block not found | Yes (re-fetch blocks) | None |
+| E-009 Folder denied | No | User checks folder sharing |
+| E-010 Rate limited | Yes (wait + retry) | None |
+| E-011 CLI not installed | Yes (npm install) | None |
+| E-012 MCP not configured | Yes (install + configure) | User provides credentials if missing |
+| E-013 Skills not loaded | Yes (npx skills add) | None |
+| E-014 Port in use | Partial (kill process) | User closes conflicting app |
+| E-015 Invalid JSON | Yes (fix JSON) | None |
+| E-016 Missing parameter | Yes (add parameter) | None |
+| E-017 ID format error | Yes (extract from URL) | User provides URL if ID unknown |
diff --git a/configs/claude-code.json b/configs/claude-code.json
new file mode 100644
index 0000000..4144cc8
--- /dev/null
+++ b/configs/claude-code.json
@@ -0,0 +1,18 @@
+{
+ "_comment": "Claude Code MCP 配置 — 飞书全能集成套件 (feishu-mcp v2)",
+ "_comment_usage": "将此文件放置到 ~/.claude/ 目录，或通过 Claude Code 的 MCP 设置导入",
+ "_comment_credentials": "FEISHU_APP_ID 和 FEISHU_APP_SECRET 请在 open.feishu.cn 创建应用后获取",
+ "_comment_auth_type": "user 表示用户身份授权，首次使用时会自动引导 OAuth 登录",
+ "mcpServers": {
+ "feishu": {
+ "_comment": "社区版 feishu-mcp — 文档创建/编辑/图表/图片，17 个精细化文档编辑工具",
+ "command": "npx",
+ "args": ["-y", "feishu-mcp@latest", "--stdio"],
+ "env": {
+ "FEISHU_APP_ID": "",
+ "FEISHU_APP_SECRET": "",
+ "FEISHU_AUTH_TYPE": "user"
+ }
+ }
+ }
+}
diff --git a/configs/cursor.json b/configs/cursor.json
new file mode 100644
index 0000000..0d26058
--- /dev/null
+++ b/configs/cursor.json
@@ -0,0 +1,18 @@
+{
+ "_comment": "Cursor MCP 配置 — 飞书全能集成套件 (feishu-mcp v2)",
+ "_comment_usage": "将此文件放置到 ~/.cursor/mcp.json，Cursor 会自动加载 MCP 服务",
+ "_comment_credentials": "FEISHU_APP_ID 和 FEISHU_APP_SECRET 请在 open.feishu.cn 创建应用后获取",
+ "_comment_auth_type": "user 表示用户身份授权，首次使用时会自动引导 OAuth 登录",
+ "mcpServers": {
+ "feishu": {
+ "_comment": "社区版 feishu-mcp — 文档创建/编辑/图表/图片，17 个精细化文档编辑工具",
+ "command": "npx",
+ "args": ["-y", "feishu-mcp@latest", "--stdio"],
+ "env": {
+ "FEISHU_APP_ID": "",
+ "FEISHU_APP_SECRET": "",
+ "FEISHU_AUTH_TYPE": "user"
+ }
+ }
+ }
+}
diff --git a/configs/openclaw.json b/configs/openclaw.json
new file mode 100644
index 0000000..be8cf8b
--- /dev/null
+++ b/configs/openclaw.json
@@ -0,0 +1,18 @@
+{
+ "_comment": "OpenClaw MCP 配置 — 飞书全能集成套件 (feishu-mcp v2)",
+ "_comment_usage": "将此文件内容导入 OpenClaw 的 MCP 配置，或放置到 OpenClaw 配置目录中",
+ "_comment_credentials": "FEISHU_APP_ID 和 FEISHU_APP_SECRET 请在 open.feishu.cn 创建应用后获取",
+ "_comment_auth_type": "user 表示用户身份授权，首次使用时会自动引导 OAuth 登录",
+ "mcpServers": {
+ "feishu": {
+ "_comment": "社区版 feishu-mcp — 文档创建/编辑/图表/图片，17 个精细化文档编辑工具",
+ "command": "npx",
+ "args": ["-y", "feishu-mcp@latest", "--stdio"],
+ "env": {
+ "FEISHU_APP_ID": "",
+ "FEISHU_APP_SECRET": "",
+ "FEISHU_AUTH_TYPE": "user"
+ }
+ }
+ }
+}
diff --git a/setup-v2.sh b/setup-v2.sh
new file mode 100755
index 0000000..87699d1
--- /dev/null
+++ b/setup-v2.sh
@@ -0,0 +1,1022 @@
+#!/usr/bin/env bash
+# setup-v2.sh — Feishu MCP v2 smart setup script
+# Installs and configures Feishu CLI + community MCP + optional official MCP.
+# Detects AI platforms (OpenClaw, Claude Code, Cursor) and generates configs.
+# All user-facing output is in Chinese. Code comments in English.
+
+set -euo pipefail
+
+# ============================================================
+# Constants
+# ============================================================
+
+SETUP_VERSION="2.0.0"
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+MIN_NODE_VERSION=16
+
+# Flags (defaults)
+FLAG_FORCE=false
+FLAG_MCP_ONLY=false
+FLAG_NO_INTERACTIVE=false
+FLAG_HELP=false
+FLAG_VERSION=false
+FLAG_PLATFORM=""
+
+# State tracking
+CLI_INSTALLED=false
+CLI_VERSION=""
+MCPORTER_CONFIGURED=false
+OFFICIAL_MCP_CONFIGURED=false
+DETECTED_PLATFORMS=""
+
+# Credentials (will be set during execution)
+APP_ID=""
+APP_SECRET=""
+
+# ============================================================
+# Trap: friendly error handler on unexpected failures
+# ============================================================
+
+cleanup_on_error() {
+ local exit_code=$?
+ # Exit code 130 = Ctrl+C — trap_ctrlc already printed a message
+ if [ "$exit_code" -ne 0 ] && [ "$exit_code" -ne 130 ]; then
+ echo ""
+ echo "❌ 脚本遇到意外错误 (退出码: $exit_code)"
+ echo " 请检查上方的错误信息，或重新运行脚本"
+ echo " 如需帮助: ./setup-v2.sh --help"
+ fi
+}
+trap cleanup_on_error EXIT
+
+# Trap Ctrl+C during auth to give a helpful message
+trap_ctrlc() {
+ echo ""
+ echo "⚠️ 操作已中断"
+ echo " 可稍后手动完成认证："
+ echo " lark-cli auth login --recommend"
+ exit 130
+}
+
+# ============================================================
+# Helper functions
+# ============================================================
+
+# Print usage/help in Chinese
+print_help() {
+ cat <<'HELPEOF'
+飞书 MCP v2 安装脚本
+
+用法:
+ ./setup-v2.sh [App_ID] [App_Secret] [选项]
+
+参数:
+ App_ID 飞书应用 App ID (可选，不提供则交互式输入)
+ App_Secret 飞书应用 App Secret (可选，不提供则交互式输入)
+
+选项:
+ -h, --help 显示此帮助信息
+ -v, --version 显示脚本版本
+ --force 强制重新执行所有步骤（忽略已有配置）
+ --mcp-only 仅安装社区 MCP，跳过 CLI 安装
+ --no-interactive 非交互模式（需通过参数提供凭证，CI/自动化使用）
+ --platform 手动指定目标平台 (openclaw / claude-code / cursor)
+
+示例:
+ # 交互式安装（推荐）
+ ./setup-v2.sh
+
+ # 非交互式安装
+ ./setup-v2.sh cli_abcd1234 your_secret_here
+
+ # 仅安装社区 MCP（不安装 CLI）
+ ./setup-v2.sh --mcp-only
+
+ # 强制重新配置
+ ./setup-v2.sh --force
+
+获取凭证: https://open.feishu.cn
+项目文档: https://github.com/AlexAnys/feishu-mcp
+HELPEOF
+}
+
+# Trim leading/trailing whitespace (Bash 3.2 compatible)
+trim() {
+ local var="$1"
+ # Remove leading whitespace
+ var="${var#"${var%%[![:space:]]*}"}"
+ # Remove trailing whitespace
+ var="${var%"${var##*[![:space:]]}"}"
+ printf '%s' "$var"
+}
+
+# Get mcporter config path following XDG convention
+get_mcporter_config_path() {
+ if [ -n "${XDG_CONFIG_HOME:-}" ]; then
+ echo "${XDG_CONFIG_HOME}/mcporter/config.json"
+ else
+ echo "${HOME}/.config/mcporter/config.json"
+ fi
+}
+
+# Check if a previous successful setup exists (for idempotency)
+has_existing_config() {
+ local config_path
+ config_path="$(get_mcporter_config_path)"
+ if [ -f "$config_path" ]; then
+ # Check if the file contains a feishu entry with non-empty credentials
+ FEISHU_CONFIG_PATH="$config_path" node -e "
+ var fs = require('fs');
+ try {
+ var c = JSON.parse(fs.readFileSync(process.env.FEISHU_CONFIG_PATH, 'utf8'));
+ if (c.mcpServers && c.mcpServers.feishu &&
+ c.mcpServers.feishu.env &&
+ c.mcpServers.feishu.env.FEISHU_APP_ID &&
+ c.mcpServers.feishu.env.FEISHU_APP_ID !== '') {
+ process.exit(0);
+ }
+ } catch(e) {}
+ process.exit(1);
+ " 2>/dev/null
+ return $?
+ fi
+ return 1
+}
+
+# Read existing credentials from mcporter config
+read_existing_credentials() {
+ local config_path
+ config_path="$(get_mcporter_config_path)"
+ if [ -f "$config_path" ]; then
+ FEISHU_CONFIG_PATH="$config_path" node -e "
+ var fs = require('fs');
+ try {
+ var c = JSON.parse(fs.readFileSync(process.env.FEISHU_CONFIG_PATH, 'utf8'));
+ if (c.mcpServers && c.mcpServers.feishu && c.mcpServers.feishu.env) {
+ console.log(c.mcpServers.feishu.env.FEISHU_APP_ID || '');
+ console.log(c.mcpServers.feishu.env.FEISHU_APP_SECRET || '');
+ } else {
+ console.log('');
+ console.log('');
+ }
+ } catch(e) {
+ console.log('');
+ console.log('');
+ }
+ " 2>/dev/null
+ fi
+}
+
+# ============================================================
+# Parse command-line arguments (Bash 3.2 compatible)
+# ============================================================
+
+POSITIONAL_ARGS=""
+POSITIONAL_COUNT=0
+
+while [ $# -gt 0 ]; do
+ case "$1" in
+ -h|--help)
+ FLAG_HELP=true
+ shift
+ ;;
+ -v|--version)
+ FLAG_VERSION=true
+ shift
+ ;;
+ --force)
+ FLAG_FORCE=true
+ shift
+ ;;
+ --mcp-only)
+ FLAG_MCP_ONLY=true
+ shift
+ ;;
+ --no-interactive)
+ FLAG_NO_INTERACTIVE=true
+ shift
+ ;;
+ --platform)
+ if [ -n "${2:-}" ]; then
+ FLAG_PLATFORM="$2"
+ shift 2
+ else
+ echo "❌ --platform 需要指定平台名称 (openclaw / claude-code / cursor)"
+ exit 1
+ fi
+ ;;
+ -*)
+ echo "❌ 未知选项: $1"
+ echo " 使用 --help 查看所有选项"
+ exit 1
+ ;;
+ *)
+ # Positional argument
+ POSITIONAL_COUNT=$((POSITIONAL_COUNT + 1))
+ if [ "$POSITIONAL_COUNT" -eq 1 ]; then
+ APP_ID="$1"
+ elif [ "$POSITIONAL_COUNT" -eq 2 ]; then
+ APP_SECRET="$1"
+ fi
+ shift
+ ;;
+ esac
+done
+
+# Handle --help and --version early exits
+if [ "$FLAG_HELP" = true ]; then
+ print_help
+ exit 0
+fi
+
+if [ "$FLAG_VERSION" = true ]; then
+ echo "feishu-mcp setup v${SETUP_VERSION}"
+ exit 0
+fi
+
+# ============================================================
+# Banner
+# ============================================================
+
+echo ""
+echo "🚀 飞书 MCP v2 智能配置向导 (v${SETUP_VERSION})"
+echo " CLI 全能管理 + 社区 MCP 精细文档编辑"
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo ""
+
+# ============================================================
+# Step 1: Prerequisites check (AC-1.1.1, E-5)
+# ============================================================
+
+echo "📋 [1/7] 检查环境..."
+echo ""
+
+# Detect OS and architecture
+DETECTED_OS="$(uname -s | tr '[:upper:]' '[:lower:]')"
+DETECTED_ARCH="$(uname -m)"
+
+case "$DETECTED_OS" in
+ darwin)
+ OS_DISPLAY="macOS"
+ ;;
+ linux)
+ OS_DISPLAY="Linux"
+ ;;
+ mingw*|msys*|cygwin*)
+ echo "❌ 不支持 Windows 原生环境"
+ echo " 请使用 WSL (Windows Subsystem for Linux) 运行此脚本"
+ echo " 安装 WSL: https://learn.microsoft.com/zh-cn/windows/wsl/install"
+ exit 1
+ ;;
+ *)
+ echo "❌ 不支持的操作系统: ${DETECTED_OS}"
+ echo " 此脚本支持 macOS 和 Linux"
+ exit 1
+ ;;
+esac
+
+case "$DETECTED_ARCH" in
+ x86_64|amd64)
+ ARCH_DISPLAY="x64"
+ ;;
+ arm64|aarch64)
+ ARCH_DISPLAY="arm64"
+ ;;
+ *)
+ ARCH_DISPLAY="$DETECTED_ARCH"
+ ;;
+esac
+
+echo " 系统: ${OS_DISPLAY} (${ARCH_DISPLAY})"
+
+# Check Node.js
+if ! command -v node >/dev/null 2>&1; then
+ echo ""
+ echo "❌ 未检测到 Node.js"
+ echo " 请先安装 Node.js (>= ${MIN_NODE_VERSION})："
+ echo " 下载地址: https://nodejs.org"
+ echo ""
+ echo " macOS 推荐: brew install node"
+ echo " Linux 推荐: https://github.com/nodesource/distributions"
+ exit 1
+fi
+
+NODE_VERSION_FULL="$(node --version 2>/dev/null)"
+# Extract major version number (strip 'v' prefix, get first segment)
+NODE_MAJOR="$(echo "$NODE_VERSION_FULL" | sed 's/^v//' | cut -d. -f1)"
+
+if [ "$NODE_MAJOR" -lt "$MIN_NODE_VERSION" ] 2>/dev/null; then
+ echo ""
+ echo "❌ Node.js 版本过低: ${NODE_VERSION_FULL}"
+ echo " 最低要求: v${MIN_NODE_VERSION}.0.0"
+ echo " 请升级 Node.js: https://nodejs.org"
+ echo ""
+ echo " macOS: brew upgrade node"
+ echo " nvm 用户: nvm install --lts"
+ exit 1
+fi
+
+echo " Node.js: ${NODE_VERSION_FULL}"
+
+# Check npm
+if ! command -v npm >/dev/null 2>&1; then
+ echo ""
+ echo "❌ 未检测到 npm"
+ echo " Node.js 已安装但缺少 npm，请重新安装 Node.js："
+ echo " https://nodejs.org"
+ exit 1
+fi
+
+NPM_VERSION="$(npm --version 2>/dev/null)"
+echo " npm: v${NPM_VERSION}"
+echo ""
+
+# ============================================================
+# Step 2: Credentials (AC-1.1.2, E-6)
+# ============================================================
+
+echo "📋 [2/7] 获取应用凭证..."
+echo ""
+
+# Idempotency: if config exists and not --force, reuse credentials
+if [ "$FLAG_FORCE" = false ] && has_existing_config 2>/dev/null; then
+ echo " ✅ 检测到已有配置，使用现有凭证"
+ EXISTING_CREDS="$(read_existing_credentials)"
+ EXISTING_ID="$(echo "$EXISTING_CREDS" | head -1)"
+ EXISTING_SECRET="$(echo "$EXISTING_CREDS" | tail -1)"
+
+ # If positional args provided, use them instead
+ if [ -n "$APP_ID" ] && [ -n "$APP_SECRET" ]; then
+ echo " 📝 使用命令行提供的新凭证"
+ elif [ -n "$EXISTING_ID" ] && [ -n "$EXISTING_SECRET" ]; then
+ APP_ID="$EXISTING_ID"
+ APP_SECRET="$EXISTING_SECRET"
+ echo " App ID: ${APP_ID}"
+ echo ""
+ fi
+else
+ # Need credentials
+ if [ -z "$APP_ID" ] || [ -z "$APP_SECRET" ]; then
+ if [ "$FLAG_NO_INTERACTIVE" = true ]; then
+ echo "❌ 非交互模式下必须提供 App ID 和 App Secret"
+ echo " 用法: ./setup-v2.sh --no-interactive"
+ exit 1
+ fi
+
+ echo " 请输入飞书应用凭证（在 open.feishu.cn 创建应用后获取）"
+ echo ""
+
+ if [ -z "$APP_ID" ]; then
+ printf " App ID (cli_xxx): "
+ read -r APP_ID
+ fi
+ if [ -z "$APP_SECRET" ]; then
+ printf " App Secret (输入不显示): "
+ # Suppress echo for secret input; handle terminals that don't support -s
+ if stty -echo 2>/dev/null; then
+ read -r APP_SECRET
+ stty echo
+ echo ""
+ else
+ read -r APP_SECRET
+ fi
+ fi
+ fi
+fi
+
+# Trim whitespace from credentials
+APP_ID="$(trim "$APP_ID")"
+APP_SECRET="$(trim "$APP_SECRET")"
+
+# Validate credentials are non-empty
+if [ -z "$APP_ID" ] || [ -z "$APP_SECRET" ]; then
+ echo ""
+ echo "❌ App ID 或 App Secret 不能为空"
+ echo " 请在飞书开放平台获取: https://open.feishu.cn"
+ exit 1
+fi
+
+# Warn if App ID doesn't start with cli_ (don't block -- format may change)
+case "$APP_ID" in
+ cli_*)
+ ;;
+ *)
+ echo " ⚠️ App ID 通常以 'cli_' 开头，当前值: ${APP_ID}"
+ echo " 如果确认无误请继续，否则请检查"
+ echo ""
+ ;;
+esac
+
+# ============================================================
+# Step 3: Feishu CLI installation (AC-1.1.3, E-3)
+# ============================================================
+
+echo "📋 [3/7] 安装飞书 CLI..."
+echo ""
+
+if [ "$FLAG_MCP_ONLY" = true ]; then
+ echo " ⏭️ 已跳过 CLI 安装 (--mcp-only 模式)"
+ echo ""
+ CLI_INSTALLED=false
+else
+ # Check if CLI is already installed
+ if command -v lark-cli >/dev/null 2>&1; then
+ CLI_VERSION="$(lark-cli --version 2>/dev/null || echo "unknown")"
+ echo " ✅ 飞书 CLI 已安装: ${CLI_VERSION}"
+ CLI_INSTALLED=true
+ if [ "$FLAG_FORCE" = false ]; then
+ echo " ⏭️ 跳过重新安装（使用 --force 强制重新安装）"
+ else
+ echo " 🔄 强制重新安装..."
+ if npm install -g @larksuite/cli 2>/tmp/feishu-cli-install-err.log; then
+ CLI_VERSION="$(lark-cli --version 2>/dev/null || echo "unknown")"
+ echo " ✅ 飞书 CLI 已更新: ${CLI_VERSION}"
+ else
+ echo " ⚠️ CLI 更新失败，继续使用现有版本"
+ fi
+ fi
+ else
+ echo " 📦 正在安装 @larksuite/cli..."
+ if npm install -g @larksuite/cli 2>/tmp/feishu-cli-install-err.log; then
+ CLI_VERSION="$(lark-cli --version 2>/dev/null || echo "unknown")"
+ echo " ✅ 飞书 CLI 安装成功: ${CLI_VERSION}"
+ CLI_INSTALLED=true
+ else
+ # Graceful degradation: CLI install failed, continue with MCP-only
+ CLI_INSTALLED=false
+ INSTALL_ERR="$(cat /tmp/feishu-cli-install-err.log 2>/dev/null || echo "")"
+
+ echo " ⚠️ 飞书 CLI 安装失败，将以 MCP-only 模式继续"
+ echo ""
+
+ # Detect specific error types and provide targeted advice
+ case "$INSTALL_ERR" in
+ *EACCES*|*permission*)
+ echo " 💡 权限不足。请运行以下命令修复 npm 全局安装路径："
+ echo " npm config set prefix '~/.npm-global'"
+ echo " export PATH=~/.npm-global/bin:\$PATH"
+ echo " # 添加到 ~/.bashrc 或 ~/.zshrc 以永久生效"
+ ;;
+ *ENOTFOUND*|*EAI_AGAIN*|*network*|*ETIMEDOUT*)
+ echo " 💡 网络连接失败。请检查："
+ echo " - 网络连接是否正常"
+ echo " - 是否需要配置代理: npm config set proxy "
+ if [ -z "${HTTP_PROXY:-}${HTTPS_PROXY:-}${http_proxy:-}${https_proxy:-}" ]; then
+ echo " - 未检测到代理环境变量，如需代理请设置 HTTPS_PROXY"
+ fi
+ ;;
+ *404*|*E404*)
+ echo " 💡 npm 包可能尚未发布或名称已变更"
+ echo " 飞书 CLI 是新发布的工具，如遇此问题请稍后重试"
+ ;;
+ *)
+ echo " 💡 安装失败原因:"
+ echo " $(echo "$INSTALL_ERR" | head -3)"
+ ;;
+ esac
+ echo ""
+ echo " 📝 MCP-only 模式仍可使用文档编辑等核心功能"
+ echo ""
+ fi
+ fi
+
+ # Install CLI skills (only if CLI was successfully installed)
+ if [ "$CLI_INSTALLED" = true ]; then
+ echo " 📦 安装 CLI 技能包..."
+ if npx skills add larksuite/cli -y -g 2>/dev/null; then
+ echo " ✅ CLI 技能包安装成功"
+ else
+ echo " ⚠️ CLI 技能包安装失败（非关键，可稍后手动安装）"
+ echo " 手动安装: npx skills add larksuite/cli -y -g"
+ fi
+ fi
+ echo ""
+fi
+
+# ============================================================
+# Step 4: Community MCP via mcporter (AC-1.1.4, E-7)
+# ============================================================
+
+echo "📋 [4/7] 配置社区 MCP (feishu-mcp)..."
+echo ""
+
+# Install mcporter if not present
+if ! command -v mcporter >/dev/null 2>&1; then
+ echo " 📦 安装 mcporter..."
+ if npm install -g mcporter 2>/tmp/feishu-mcporter-err.log; then
+ echo " ✅ mcporter 安装成功"
+ else
+ MCPORTER_ERR="$(cat /tmp/feishu-mcporter-err.log 2>/dev/null || echo "")"
+ case "$MCPORTER_ERR" in
+ *EACCES*|*permission*)
+ echo " ❌ mcporter 安装失败 — 权限不足"
+ echo " 请运行: npm config set prefix '~/.npm-global'"
+ echo " 然后重新运行此脚本"
+ ;;
+ *ENOTFOUND*|*EAI_AGAIN*|*ETIMEDOUT*)
+ echo " ❌ mcporter 安装失败 — 网络错误"
+ echo " 请检查网络连接后重试"
+ if [ -z "${HTTP_PROXY:-}${HTTPS_PROXY:-}${http_proxy:-}${https_proxy:-}" ]; then
+ echo " 如需代理请设置: export HTTPS_PROXY="
+ fi
+ ;;
+ *)
+ echo " ❌ mcporter 安装失败"
+ echo " $(echo "$MCPORTER_ERR" | head -3)"
+ ;;
+ esac
+ echo ""
+ echo " 手动安装: npm install -g mcporter"
+ exit 1
+ fi
+else
+ echo " ✅ mcporter 已安装"
+fi
+
+# Generate/merge mcporter config
+MCPORTER_CONFIG="$(get_mcporter_config_path)"
+MCPORTER_CONFIG_DIR="$(dirname "$MCPORTER_CONFIG")"
+
+# Ensure config directory exists
+mkdir -p "$MCPORTER_CONFIG_DIR"
+
+# Use inline Node.js to safely read/merge/write JSON config (v1 pattern)
+# Credentials passed via environment variables to avoid injection (HIGH-1)
+NODE_ERR_FILE="$(mktemp /tmp/feishu-node-err.XXXXXX)"
+if FEISHU_APP_ID="$APP_ID" FEISHU_APP_SECRET="$APP_SECRET" \
+ FEISHU_CONFIG_PATH="$MCPORTER_CONFIG" node -e "
+var fs = require('fs');
+var configPath = process.env.FEISHU_CONFIG_PATH;
+var appId = process.env.FEISHU_APP_ID;
+var appSecret = process.env.FEISHU_APP_SECRET;
+
+var config = {};
+
+// Read existing config if present
+if (fs.existsSync(configPath)) {
+ try {
+ config = JSON.parse(fs.readFileSync(configPath, 'utf8'));
+ } catch(e) {
+ // E-7: Malformed JSON — back up and start fresh
+ var timestamp = Date.now();
+ var backupPath = configPath + '.bak.' + timestamp;
+ fs.renameSync(configPath, backupPath);
+ console.log(' ⚠️ 现有配置文件格式错误，已备份到: ' + backupPath);
+ config = {};
+ }
+}
+
+if (!config.mcpServers) config.mcpServers = {};
+
+// Add/update community MCP entry (preserves other entries)
+config.mcpServers.feishu = {
+ command: 'npx',
+ args: ['-y', 'feishu-mcp@latest', '--stdio'],
+ env: {
+ FEISHU_APP_ID: appId,
+ FEISHU_APP_SECRET: appSecret,
+ FEISHU_AUTH_TYPE: 'user'
+ }
+};
+
+fs.writeFileSync(configPath, JSON.stringify(config, null, 2));
+console.log(' ✅ 社区 MCP 配置已写入: ' + configPath);
+" 2>"$NODE_ERR_FILE"; then
+ :
+else
+ echo " ❌ 社区 MCP 配置写入失败"
+ if [ -s "$NODE_ERR_FILE" ]; then
+ echo " 错误详情: $(head -3 "$NODE_ERR_FILE")"
+ fi
+ echo " 请检查文件权限后重试: ${MCPORTER_CONFIG}"
+ rm -f "$NODE_ERR_FILE"
+ exit 1
+fi
+rm -f "$NODE_ERR_FILE"
+
+MCPORTER_CONFIGURED=true
+echo ""
+
+# ============================================================
+# Step 5: Official MCP (optional) (AC-1.1.5)
+# ============================================================
+
+echo "📋 [5/7] 官方 MCP (可选)..."
+echo ""
+
+INSTALL_OFFICIAL=false
+
+if [ "$FLAG_NO_INTERACTIVE" = true ]; then
+ echo " ⏭️ 非交互模式，跳过官方 MCP"
+ echo " 稍后安装: npx -y @larksuiteoapi/lark-mcp login -a -s "
+else
+ printf " 是否安装官方 lark-mcp？(多维表格/搜索/知识库能力) [y/N]: "
+ read -r OFFICIAL_ANSWER || true
+ OFFICIAL_ANSWER="$(trim "${OFFICIAL_ANSWER:-N}")"
+
+ case "$OFFICIAL_ANSWER" in
+ y|Y|yes|YES|是)
+ INSTALL_OFFICIAL=true
+ ;;
+ *)
+ INSTALL_OFFICIAL=false
+ ;;
+ esac
+fi
+
+if [ "$INSTALL_OFFICIAL" = true ]; then
+ echo ""
+ echo " 📦 添加官方 MCP 配置..."
+
+ # Add official MCP entries to mcporter config
+ # Credentials passed via environment variables to avoid injection (HIGH-1)
+ NODE_ERR_FILE="$(mktemp /tmp/feishu-node-err.XXXXXX)"
+ if FEISHU_APP_ID="$APP_ID" FEISHU_APP_SECRET="$APP_SECRET" \
+ FEISHU_CONFIG_PATH="$MCPORTER_CONFIG" node -e "
+ var fs = require('fs');
+ var configPath = process.env.FEISHU_CONFIG_PATH;
+ var appId = process.env.FEISHU_APP_ID;
+ var appSecret = process.env.FEISHU_APP_SECRET;
+
+ var config = JSON.parse(fs.readFileSync(configPath, 'utf8'));
+ if (!config.mcpServers) config.mcpServers = {};
+
+ // Official lark-mcp — app identity (bitable, messaging)
+ config.mcpServers.lark = {
+ command: 'npx',
+ args: ['-y', '@larksuiteoapi/lark-mcp', 'mcp', '-a', appId, '-s', appSecret]
+ };
+
+ // Official lark-mcp — user identity (search, wiki)
+ config.mcpServers['lark-user'] = {
+ command: 'npx',
+ args: ['-y', '@larksuiteoapi/lark-mcp', 'mcp', '-a', appId, '-s', appSecret,
+ '--oauth', '--token-mode', 'user_access_token']
+ };
+
+ fs.writeFileSync(configPath, JSON.stringify(config, null, 2));
+ console.log(' ✅ 官方 MCP 配置已添加');
+ " 2>"$NODE_ERR_FILE"; then
+ :
+ else
+ echo " ❌ 官方 MCP 配置写入失败"
+ if [ -s "$NODE_ERR_FILE" ]; then
+ echo " 错误详情: $(head -3 "$NODE_ERR_FILE")"
+ fi
+ echo " 请手动添加: npx -y @larksuiteoapi/lark-mcp login -a -s "
+ fi
+ rm -f "$NODE_ERR_FILE"
+
+ echo " 🔑 正在进行官方 MCP 授权..."
+ npx -y @larksuiteoapi/lark-mcp login -a "$APP_ID" -s "$APP_SECRET" || echo " ⚠️ 官方 MCP 授权跳过（可稍后手动完成: npx -y @larksuiteoapi/lark-mcp login）"
+
+ OFFICIAL_MCP_CONFIGURED=true
+else
+ echo " ⏭️ 跳过官方 MCP"
+ echo " 稍后安装: 重新运行此脚本，或手动添加到 mcporter 配置"
+fi
+
+echo ""
+
+# ============================================================
+# Step 6: Auth configuration (AC-1.1.6, E-10)
+# ============================================================
+
+echo "📋 [6/7] 认证配置..."
+echo ""
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo "🔐 请在飞书开放平台配置重定向 URL"
+echo ""
+echo " 打开: https://open.feishu.cn/app/${APP_ID}/safe"
+echo ""
+echo " 添加以下重定向 URL："
+
+if [ "$CLI_INSTALLED" = true ]; then
+ echo " • http://localhost:3000/callback (飞书 CLI)"
+fi
+echo " • http://localhost:3333/callback (社区 MCP)"
+
+if [ "$OFFICIAL_MCP_CONFIGURED" = true ]; then
+ echo " • http://localhost:3000/callback (官方 MCP)"
+fi
+
+echo ""
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+
+if [ "$FLAG_NO_INTERACTIVE" = false ]; then
+ printf " 配置完成后按 Enter 继续..."
+ read -r _PAUSE || true
+fi
+
+echo ""
+
+# CLI auth (only if CLI was installed)
+if [ "$CLI_INSTALLED" = true ]; then
+ if [ "$FLAG_NO_INTERACTIVE" = false ]; then
+ echo " 🔑 初始化 CLI 配置..."
+ if lark-cli config init --new 2>/dev/null; then
+ echo " ✅ CLI 配置初始化成功"
+ else
+ echo " ⚠️ CLI 配置初始化失败"
+ echo " 手动执行: lark-cli config init --new"
+ fi
+
+ echo ""
+ echo " 🔑 CLI OAuth 登录（浏览器将自动打开）..."
+ echo " 如果 10 秒内浏览器未打开，请手动访问终端中显示的 URL"
+ echo ""
+
+ # Trap Ctrl+C to provide friendly skip message
+ trap trap_ctrlc INT
+ if lark-cli auth login --recommend 2>/dev/null; then
+ echo " ✅ CLI 认证成功"
+ else
+ echo " ⚠️ CLI 认证未完成"
+ echo " 手动执行: lark-cli auth login --recommend"
+ fi
+ # Restore default trap
+ trap cleanup_on_error EXIT
+ trap - INT
+ else
+ echo " ⏭️ 非交互模式，跳过 CLI 认证"
+ echo " 请稍后手动执行认证:"
+ echo " lark-cli config init --new"
+ echo " lark-cli auth login --recommend"
+ fi
+fi
+
+echo ""
+echo " 📝 社区 MCP 会在首次使用时自动引导 OAuth 授权"
+echo ""
+
+# ============================================================
+# Step 7: Platform detection and config generation (AC-1.1.7, E-9)
+# ============================================================
+
+echo "📋 [7/7] 检测 AI 平台并生成配置..."
+echo ""
+
+# Ensure configs/ directory in project exists for portable templates
+mkdir -p "${SCRIPT_DIR}/configs"
+
+# Platform detection functions
+detect_openclaw() {
+ if command -v openclaw >/dev/null 2>&1; then
+ return 0
+ fi
+ if [ -d "${HOME}/.openclaw" ] || [ -d "${HOME}/.config/openclaw" ]; then
+ return 0
+ fi
+ return 1
+}
+
+detect_claude_code() {
+ if command -v claude >/dev/null 2>&1; then
+ return 0
+ fi
+ if [ -d "${HOME}/.claude" ]; then
+ return 0
+ fi
+ return 1
+}
+
+detect_cursor() {
+ if [ -d "${HOME}/.cursor" ]; then
+ return 0
+ fi
+ # macOS app check
+ if [ "$DETECTED_OS" = "darwin" ] && [ -d "/Applications/Cursor.app" ]; then
+ return 0
+ fi
+ # Linux common locations
+ if [ "$DETECTED_OS" = "linux" ]; then
+ if command -v cursor >/dev/null 2>&1; then
+ return 0
+ fi
+ fi
+ return 1
+}
+
+# Generate personalized config for a platform using Node.js
+# Args: $1=platform_name, $2=output_path, $3=template_path
+generate_platform_config() {
+ local platform="$1"
+ local output_path="$2"
+ local template_path="$3"
+
+ # Check if output already exists (and not --force)
+ if [ -f "$output_path" ] && [ "$FLAG_FORCE" = false ]; then
+ if [ "$FLAG_NO_INTERACTIVE" = true ]; then
+ echo " ⏭️ ${platform} 配置已存在，跳过: ${output_path}"
+ return 0
+ fi
+ printf " ${platform} 配置已存在，是否覆盖？ [y/N]: "
+ read -r OVERWRITE_ANSWER || true
+ OVERWRITE_ANSWER="$(trim "${OVERWRITE_ANSWER:-N}")"
+ case "$OVERWRITE_ANSWER" in
+ y|Y|yes|YES|是) ;;
+ *)
+ echo " ⏭️ 跳过 ${platform} 配置"
+ return 0
+ ;;
+ esac
+ fi
+
+ # Ensure output directory exists
+ mkdir -p "$(dirname "$output_path")"
+
+ # Generate config by MERGING feishu entry into existing file (BLOCKER-1 fix)
+ # Credentials passed via environment variables to avoid injection (HIGH-1)
+ NODE_ERR_FILE="$(mktemp /tmp/feishu-node-err.XXXXXX)"
+ if FEISHU_APP_ID="$APP_ID" FEISHU_APP_SECRET="$APP_SECRET" \
+ FEISHU_TEMPLATE_PATH="$template_path" FEISHU_OUTPUT_PATH="$output_path" node -e "
+ var fs = require('fs');
+ var templatePath = process.env.FEISHU_TEMPLATE_PATH;
+ var outputPath = process.env.FEISHU_OUTPUT_PATH;
+ var appId = process.env.FEISHU_APP_ID;
+ var appSecret = process.env.FEISHU_APP_SECRET;
+
+ // Read template to get the feishu entry structure
+ var template = JSON.parse(fs.readFileSync(templatePath, 'utf8'));
+
+ // Read existing output file if present — preserve other entries
+ var config = {};
+ if (fs.existsSync(outputPath)) {
+ try {
+ config = JSON.parse(fs.readFileSync(outputPath, 'utf8'));
+ } catch(e) {
+ // Malformed JSON — back up and start fresh
+ var timestamp = Date.now();
+ var backupPath = outputPath + '.bak.' + timestamp;
+ fs.renameSync(outputPath, backupPath);
+ console.log(' ⚠️ 现有配置文件格式错误，已备份到: ' + backupPath);
+ config = {};
+ }
+ }
+
+ if (!config.mcpServers) config.mcpServers = {};
+
+ // Merge feishu entry from template (with credentials filled in)
+ if (template.mcpServers && template.mcpServers.feishu) {
+ var feishuEntry = JSON.parse(JSON.stringify(template.mcpServers.feishu));
+ if (feishuEntry.env) {
+ feishuEntry.env.FEISHU_APP_ID = appId;
+ feishuEntry.env.FEISHU_APP_SECRET = appSecret;
+ }
+ config.mcpServers.feishu = feishuEntry;
+ }
+
+ fs.writeFileSync(outputPath, JSON.stringify(config, null, 2));
+ " 2>"$NODE_ERR_FILE"; then
+ echo " ✅ ${platform} 配置已生成: ${output_path}"
+ else
+ echo " ❌ ${platform} 配置生成失败"
+ if [ -s "$NODE_ERR_FILE" ]; then
+ echo " 错误详情: $(head -3 "$NODE_ERR_FILE")"
+ fi
+ echo " 请检查文件权限后重试: ${output_path}"
+ fi
+ rm -f "$NODE_ERR_FILE"
+}
+
+# Determine which platforms to configure
+PLATFORM_LIST=""
+
+if [ -n "$FLAG_PLATFORM" ]; then
+ # Manual platform specification
+ PLATFORM_LIST="$FLAG_PLATFORM"
+ echo " 📋 手动指定平台: ${FLAG_PLATFORM}"
+else
+ # Auto-detect platforms
+ if detect_openclaw; then
+ PLATFORM_LIST="${PLATFORM_LIST} openclaw"
+ DETECTED_PLATFORMS="${DETECTED_PLATFORMS}OpenClaw "
+ echo " 🔍 检测到 OpenClaw"
+ fi
+ if detect_claude_code; then
+ PLATFORM_LIST="${PLATFORM_LIST} claude-code"
+ DETECTED_PLATFORMS="${DETECTED_PLATFORMS}Claude-Code "
+ echo " 🔍 检测到 Claude Code"
+ fi
+ if detect_cursor; then
+ PLATFORM_LIST="${PLATFORM_LIST} cursor"
+ DETECTED_PLATFORMS="${DETECTED_PLATFORMS}Cursor "
+ echo " 🔍 检测到 Cursor"
+ fi
+
+ if [ -z "$PLATFORM_LIST" ]; then
+ echo " 📋 未检测到已安装的 AI 平台"
+ fi
+fi
+
+echo ""
+
+# Generate configs for detected/specified platforms
+for platform in $PLATFORM_LIST; do
+ case "$platform" in
+ openclaw)
+ # OpenClaw: generate in project configs/ and inform user
+ generate_platform_config "OpenClaw" \
+ "${SCRIPT_DIR}/configs/openclaw-configured.json" \
+ "${SCRIPT_DIR}/configs/openclaw.json"
+ echo " 💡 请将配置导入 OpenClaw 或复制到 OpenClaw 配置目录"
+ ;;
+ claude-code)
+ # Claude Code: generate to ~/.claude/
+ CLAUDE_MCP_DIR="${HOME}/.claude"
+ mkdir -p "$CLAUDE_MCP_DIR"
+ generate_platform_config "Claude Code" \
+ "${CLAUDE_MCP_DIR}/mcp.json" \
+ "${SCRIPT_DIR}/configs/claude-code.json"
+ ;;
+ cursor)
+ # Cursor: generate to ~/.cursor/mcp.json
+ CURSOR_DIR="${HOME}/.cursor"
+ mkdir -p "$CURSOR_DIR"
+ generate_platform_config "Cursor" \
+ "${CURSOR_DIR}/mcp.json" \
+ "${SCRIPT_DIR}/configs/cursor.json"
+ ;;
+ *)
+ echo " ⚠️ 未知平台: ${platform}"
+ echo " 支持的平台: openclaw, claude-code, cursor"
+ ;;
+ esac
+done
+
+# Always generate portable templates in project configs/ directory
+echo ""
+echo " 📁 便携配置模板已保存在: ${SCRIPT_DIR}/configs/"
+echo " openclaw.json, claude-code.json, cursor.json"
+echo ""
+
+# ============================================================
+# Validation and summary (AC-1.1.9)
+# ============================================================
+
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo "📊 安装摘要"
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo ""
+
+# CLI auth status check
+if [ "$CLI_INSTALLED" = true ]; then
+ AUTH_STATUS="$(lark-cli auth status 2>/dev/null || echo "未认证")"
+ echo " 飞书 CLI: ✅ 已安装 (${CLI_VERSION})"
+ echo " CLI 认证: ${AUTH_STATUS}"
+else
+ echo " 飞书 CLI: ❌ 未安装"
+ if [ "$FLAG_MCP_ONLY" = true ]; then
+ echo " (--mcp-only 模式，已跳过)"
+ else
+ echo " 安装方法: npm install -g @larksuite/cli"
+ fi
+fi
+
+if [ "$MCPORTER_CONFIGURED" = true ]; then
+ echo " 社区 MCP: ✅ 已配置"
+else
+ echo " 社区 MCP: ❌ 未配置"
+fi
+
+if [ "$INSTALL_OFFICIAL" = true ] && [ "$OFFICIAL_MCP_CONFIGURED" = true ]; then
+ echo " 官方 MCP: ✅ 已配置"
+elif [ "$INSTALL_OFFICIAL" = true ]; then
+ echo " 官方 MCP: ⚠️ 配置不完整"
+else
+ echo " 官方 MCP: ⏭️ 已跳过"
+fi
+
+echo " 配置文件: ${MCPORTER_CONFIG}"
+
+if [ -n "$DETECTED_PLATFORMS" ]; then
+ echo " 平台配置: ${DETECTED_PLATFORMS}"
+else
+ echo " 平台配置: 便携模板 (${SCRIPT_DIR}/configs/)"
+fi
+
+echo ""
+echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+echo "🎉 配置完成！"
+echo ""
+echo "试试以下命令："
+echo ""
+
+if [ "$CLI_INSTALLED" = true ]; then
+ echo " # 搜索文档"
+ echo " lark-cli doc +search --keyword '会议纪要'"
+ echo ""
+fi
+
+echo " # 创建文档（社区 MCP）"
+echo " mcporter call feishu.create_feishu_document --args '{\"title\":\"测试文档\"}'"
+echo ""
+
+if [ "$CLI_INSTALLED" = true ]; then
+ echo " # 查看日历"
+ echo " lark-cli calendar +agenda"
+ echo ""
+fi
+
+echo "📚 完整文档: https://github.com/AlexAnys/feishu-mcp"
+echo ""
diff --git a/skills/data-to-doc.md b/skills/data-to-doc.md
new file mode 100644
index 0000000..9b1b18f
--- /dev/null
+++ b/skills/data-to-doc.md
@@ -0,0 +1,191 @@
+---
+name: data-to-doc
+description: 数据报告生成工作流 — 从多维表格查询数据(CLI)，处理汇总后创建格式化文档(CLI+MCP)，支持样式化关键指标。
+metadata:
+ bins:
+ - lark-cli
+ - mcporter
+---
+
+# 数据报告生成工作流
+
+组合 CLI 和 MCP，从多维表格查询数据，自动生成带格式的飞书文档报告。
+
+---
+
+## When to Use This Skill
+
+- User asks to generate a report from bitable/multidimensional table data ("用多维表格数据生成报告")
+- User wants to create a formatted document from structured data ("把数据整理成文档")
+- User needs a periodic data summary document ("帮我出一份数据周报")
+- User wants to visualize key metrics in a Feishu document
+
+---
+
+## Prerequisites
+
+1. Confirm CLI auth: `lark-cli auth status` — must show valid token.
+2. Confirm MCP auth: `FEISHU_APP_ID` and `FEISHU_APP_SECRET` env vars set.
+3. Obtain bitable identifiers from the user:
+ - `app_token`: the bitable app token (from URL `https://xxx.feishu.cn/base/`)
+ - `table_id`: the specific table within the bitable (ask user if multiple tables exist)
+4. Determine report scope: date range, filters, grouping criteria.
+
+---
+
+## Step-by-Step Workflow
+
+### Step 1: Query Data from Multidimensional Table
+
+**Tool**: CLI `lark-cli base +query`
+
+```bash
+lark-cli base +query
+```
+
+The CLI will prompt for:
+- app_token and table_id
+- Filter conditions (e.g., date range, status)
+- Sort order
+- Fields to return
+
+**Error handling**:
+- If "skill not found" → run `npx skills add larksuite/cli -y -g` and retry.
+- If "permission denied" → ask user to add `bitable:app` scope at `https://open.feishu.cn/app/{APP_ID}/security`.
+- If "app_token invalid" → verify the bitable URL with the user.
+- If no records returned → inform user the query returned empty results. Adjust filters and retry.
+
+### Step 2: Process and Summarize Data
+
+**Tool**: AI analysis (no external tool needed)
+
+Process the raw records from Step 1:
+
+1. **Aggregate**: Calculate totals, averages, counts, percentages as appropriate.
+2. **Group**: Organize data by categories (e.g., by status, by person, by date).
+3. **Highlight**: Identify key metrics — highest/lowest values, trends, anomalies.
+4. **Structure**: Plan the document layout:
+ - Title with date range
+ - Executive summary (2-3 sentences)
+ - Key metrics section (highlighted numbers)
+ - Detailed data breakdown (grouped tables or lists)
+ - Conclusion / next steps
+
+### Step 3: Create a New Document
+
+**Tool**: CLI `lark-cli doc +create`
+
+```bash
+lark-cli doc +create
+```
+
+Provide a Markdown template with the report title and basic structure. The CLI will create the document and return a document URL and document_id.
+
+Extract the document_id from the returned URL for subsequent MCP operations.
+
+**Error handling**:
+- If "permission denied" → ask user to add `docx:document` scope.
+- If creation fails → retry once. If still failing, check `lark-cli auth status`.
+
+### Step 4: Add Formatted Blocks with Data
+
+**Tool**: MCP `batch_create_feishu_blocks`
+
+Use `batch_create_feishu_blocks` to add structured content. Build the blocks array dynamically from Step 2's processed data. Recommended document structure:
+- `heading2` "数据概览" with report period
+- `heading3` "关键指标" with aggregated metrics
+- `heading3` "分类明细" with per-category breakdowns
+- `heading2` "总结与下一步" with conclusions
+
+Use appropriate block types:
+- `heading2` for major sections
+- `heading3` for subsections
+- `text` for paragraphs and data
+- `list` for bullet points (ordered or unordered)
+- `code` for raw data or formulas
+
+**Error handling**:
+- If "document not found" → verify document_id from Step 3.
+- If batch creation partially fails → retry failed blocks individually.
+- If block limit exceeded → split into multiple batch calls (max ~50 blocks per call).
+
+### Step 5: Apply Styling to Key Metrics
+
+**Tool**: MCP `update_feishu_block_text`
+
+After creating blocks, apply visual emphasis to key metrics. First, identify the block_ids of key metric blocks from Step 4's response.
+
+For each key metric block, apply styling:
+
+```
+MCP tool: update_feishu_block_text
+Parameters: {
+ "document_id": "",
+ "block_id": "",
+ "content": "完成率: 95%",
+ "text_color": "green",
+ "align": "center"
+}
+```
+
+Color guidelines:
+- **green**: positive metrics (targets met, growth)
+- **red**: negative metrics (decline, missed targets)
+- **blue**: neutral highlights (totals, key numbers)
+- **orange**: warnings (approaching threshold)
+
+Available colors: `red`, `orange`, `yellow`, `green`, `blue`, `purple`, `grey`, `default`.
+
+**Error handling**:
+- If "block not found" → re-fetch block list with `get_feishu_document_blocks` and find correct block_id.
+- If styling fails → the content is still readable without styling. Log warning and continue.
+
+---
+
+## Output Format
+
+The generated document will have this structure:
+
+```
+# [报告标题] — YYYY-MM-DD
+
+## 数据概览
+报告期间: ... 至 ...
+
+## 关键指标
+总计: XXX 完成率: XX% 增长: XX%
+(green) (green/red) (green/red)
+
+## 分类明细
+### 类别 A
+- 记录数: XX (XX%)
+- 关键数据: ...
+
+### 类别 B
+- 记录数: XX (XX%)
+- 关键数据: ...
+
+## 总结与下一步
+- 总结要点 1
+- 总结要点 2
+- 下一步行动
+```
+
+Return the document URL to the user upon completion.
+
+---
+
+## Example Prompts
+
+- "用项目跟踪表的数据生成一份周报"
+- "从 Bug 多维表格生成本月 Bug 统计报告"
+- "把销售数据表整理成一份格式化的文档报告"
+- "用这个多维表格 https://xxx.feishu.cn/base/appXXX 生成数据报告"
+
+---
+
+## Partial Execution
+
+- "只查数据不生成文档" → Execute Step 1-2 only, present summary to user.
+- "文档已经有了，帮我填入数据" → Skip Step 3, execute Steps 1-2 and 4-5 on existing document.
+- "不需要样式，纯文本就行" → Execute Steps 1-4, skip Step 5.
diff --git a/skills/doc-review.md b/skills/doc-review.md
new file mode 100644
index 0000000..85939ef
--- /dev/null
+++ b/skills/doc-review.md
@@ -0,0 +1,164 @@
+---
+name: doc-review
+description: 文档审阅工作流 — 读取文档内容(MCP)，分析质量和结构，创建待办任务(CLI)，发送审阅总结到群聊(CLI)。
+metadata:
+ bins:
+ - lark-cli
+ - mcporter
+---
+
+# 文档审阅工作流
+
+组合 CLI 和 MCP 完成文档审阅闭环：读取文档 → 分析内容 → 创建任务 → 发送总结。
+
+---
+
+## When to Use This Skill
+
+- User asks to review a Feishu document ("帮我审阅这个文档", "看看这篇文档写得怎么样")
+- User wants to check document quality, structure, or completeness
+- User needs to create action items from a document review
+- User wants to send review feedback to a group chat
+
+---
+
+## Prerequisites
+
+1. Confirm CLI auth: `lark-cli auth status` — must show valid token.
+2. Confirm MCP auth: `FEISHU_APP_ID` and `FEISHU_APP_SECRET` env vars set.
+3. Obtain the document_id from the user. If user provides a URL like `https://xxx.feishu.cn/docx/doxcnXXX`, extract `doxcnXXX` as the document_id.
+
+---
+
+## Step-by-Step Workflow
+
+### Step 1: Read Document Content
+
+**Tool**: MCP `get_feishu_document_blocks` + `get_feishu_document_text`
+
+First, get the block structure to understand the document layout:
+
+```
+MCP tool: get_feishu_document_blocks
+Parameters: { "document_id": "" }
+```
+
+Then, get the plain text for content analysis:
+
+```
+MCP tool: get_feishu_document_text
+Parameters: { "document_id": "" }
+```
+
+**Error handling**:
+- If "document not found" → verify the document_id with the user.
+- If "permission denied" → ask user to add `docx:document:readonly` scope at `https://open.feishu.cn/app/{APP_ID}/security`.
+- If "unauthorized" → run `lark-cli auth login --recommend` to refresh token.
+
+### Step 2: Analyze Content Quality
+
+**Tool**: AI analysis (no external tool needed)
+
+Analyze the document on these dimensions:
+
+1. **Structure** — Does the document have clear headings, sections, and logical flow? Count heading blocks from Step 1's block structure.
+2. **Completeness** — Are there TODO markers, empty sections, placeholder text, or missing content?
+3. **Clarity** — Is the language clear? Are key terms defined? Are there ambiguous statements?
+4. **Formatting** — Are lists, tables, and code blocks used appropriately? Is formatting consistent?
+5. **Action items** — Identify any items that require follow-up (TODOs, decisions needed, open questions).
+
+Produce a structured review summary with:
+- Overall quality score (1-10)
+- Strengths (bullet list)
+- Issues found (bullet list with severity: high/medium/low)
+- Action items (numbered list)
+
+### Step 3: Create Tasks for Action Items
+
+**Tool**: CLI `lark-cli task +create`
+
+For each action item identified in Step 2, create a task:
+
+```bash
+lark-cli task +create
+```
+
+The CLI will prompt for task details interactively. Provide:
+- Task title: the action item description
+- Due date: based on urgency (high = 3 days, medium = 7 days, low = 14 days)
+- Assignee: if mentioned in the document or specified by user
+
+**Error handling**:
+- If "skill not found" → run `npx skills add larksuite/cli -y -g` and retry.
+- If "permission denied" → ask user to add `task:task` scope.
+
+### Step 4: Send Review Summary to Group Chat
+
+**Tool**: CLI `lark-cli im +send`
+
+Send the review summary to the specified group chat:
+
+```bash
+lark-cli im +send
+```
+
+The message should include:
+- Document title and link
+- Overall quality score
+- Key issues found (top 3)
+- Number of tasks created
+- Brief recommendation (approve / needs revision / major rework)
+
+**Error handling**:
+- If "chat not found" → ask user for the correct chat ID or group name.
+- If "permission denied" → ask user to add `im:message` scope.
+- If user doesn't specify a chat → ask which group to send to, or skip this step.
+
+---
+
+## Output Format
+
+Present the review to the user in this structure:
+
+```
+## 文档审阅报告
+
+**文档**: [标题](链接)
+**审阅时间**: YYYY-MM-DD
+**综合评分**: X/10
+
+### 优点
+- ...
+
+### 问题
+- [高] ...
+- [中] ...
+- [低] ...
+
+### 待办任务（已创建）
+1. ...（截止日期: YYYY-MM-DD）
+2. ...
+
+### 建议
+通过 / 需修改 / 需重写
+```
+
+---
+
+## Example Prompts
+
+- "帮我审阅一下这个文档 https://xxx.feishu.cn/docx/doxcnXXX"
+- "检查这篇文档的质量，找出需要改进的地方"
+- "审阅文档并把问题发到项目群里"
+- "看看这个文档有没有遗漏的内容，创建待办任务跟进"
+
+---
+
+## Partial Execution
+
+If the user only needs part of the workflow:
+- "只看看文档内容" → Execute Steps 1-2 only.
+- "审阅完帮我建任务" → Execute Steps 1-3, skip Step 4.
+- "审阅完发到群里就行" → Execute Steps 1-2 and 4, skip Step 3.
+
+Adapt to the user's request. Do not force all steps if not needed.
diff --git a/skills/feishu-overview.md b/skills/feishu-overview.md
new file mode 100644
index 0000000..55fbd7a
--- /dev/null
+++ b/skills/feishu-overview.md
@@ -0,0 +1,135 @@
+---
+name: feishu-overview
+description: 飞书全能集成套件总览 — 所有 CLI 技能和 MCP 工具的中文索引，快速路由指南，当用户笼统提及"飞书"时加载此技能。
+metadata:
+ bins:
+ - lark-cli
+ - mcporter
+---
+
+# 飞书能力总览
+
+当用户笼统提及"飞书"、"帮我用飞书"、"飞书能做什么"时，加载此技能作为入口。根据用户意图路由到具体技能或工具。
+
+---
+
+## When to Use This Skill
+
+- User mentions "飞书" without specifying a domain
+- User asks "飞书能做什么" or "有哪些飞书功能"
+- User needs help choosing the right tool for a Feishu task
+- First interaction involving Feishu — orient the user before diving in
+
+---
+
+## CLI 技能一览（19 个技能，11 业务域）
+
+| 技能 | 中文名称 | 一句话说明 |
+|---|---|---|
+| `lark-shared` | 认证与配置 | 应用初始化、OAuth 登录、身份切换、权限管理 |
+| `lark-calendar` | 日历管理 | 查看日程、创建会议、查询忙闲、推荐空闲时段 |
+| `lark-im` | 即时通讯 | 发送/回复消息、搜索聊天记录、管理群聊 |
+| `lark-doc` | 云文档 | 从 Markdown 创建文档、搜索文档、更新内容、插入图片 |
+| `lark-drive` | 云空间 | 上传/下载文件、管理文件夹、管理权限和评论 |
+| `lark-sheets` | 电子表格 | 创建表格、读写单元格、追加行、导出文件 |
+| `lark-base` | 多维表格 | 建表、字段管理、记录查询、视图配置、仪表盘 |
+| `lark-task` | 任务管理 | 创建待办、查看任务列表、子任务拆分、分配成员 |
+| `lark-mail` | 邮件 | 起草/发送/回复/转发/搜索邮件 |
+| `lark-contact` | 通讯录 | 搜索员工、查看组织架构、查询部门结构 |
+| `lark-wiki` | 知识库 | 管理知识空间、创建文档节点、组织层级结构 |
+| `lark-vc` | 视频会议 | 查询已结束的会议记录、获取会议纪要 |
+| `lark-minutes` | 妙记 | 获取妙记基础信息、AI 总结、待办、章节 |
+| `lark-whiteboard` | 画板 | 创建画板、Mermaid 图表、流程图、思维导图 |
+| `lark-event` | 事件订阅 | WebSocket 实时监听飞书事件，输出 NDJSON |
+| `lark-openapi-explorer` | API 探索 | 从官方文档库挖掘未封装的 OpenAPI 接口 |
+| `lark-skill-maker` | 技能创建 | 创建自定义 lark-cli Skill |
+| `lark-workflow-meeting-summary` | 会议纪要工作流 | 汇总会议纪要并生成结构化报告 |
+| `lark-workflow-standup-report` | 站会报告工作流 | 生成日程 + 未完成任务摘要 |
+
+---
+
+## 社区 MCP 工具一览（17 个工具）
+
+| 工具名称 | 中文名称 | 一句话说明 |
+|---|---|---|
+| `create_feishu_document` | 创建文档 | 在指定文件夹创建新文档 |
+| `get_feishu_document_info` | 获取文档信息 | 返回标题、修改时间等元数据 |
+| `get_feishu_document_blocks` | 获取 Block 结构 | 返回文档完整 Block 树 |
+| `get_feishu_document_text` | 提取纯文本 | 提取文档的纯文本内容 |
+| `batch_create_feishu_blocks` | 批量创建 Block | 原子操作，支持 text/code/heading/list |
+| `update_feishu_block_text` | 更新 Block 样式 | 修改文本内容、8 色彩、对齐方式 |
+| `delete_feishu_document_blocks` | 删除 Block | 按 block_id 精确删除 |
+| `upload_and_bind_image_to_block` | 上传图片 | 上传图片并绑定到 Block 一步完成 |
+| `create_feishu_table` | 创建表格 | 在文档中创建指定行列数的表格 |
+| `create_feishu_folder` | 创建文件夹 | 在云空间创建文件夹 |
+| `list_feishu_folder_contents` | 列出文件夹内容 | 浏览文件夹中的文件和子文件夹 |
+| `get_feishu_folder_files` | 获取文件列表 | 获取文件夹内的文件清单 |
+| `search_feishu_documents` | 搜索文档 | 按关键词搜索文档 |
+| `get_feishu_wiki_node_info` | Wiki 节点信息 | 获取知识库节点元数据 |
+| `get_feishu_whiteboard_content` | 读取画板 | 读取画板内容（只读） |
+
+---
+
+## 快速路由指南
+
+**想要做 X？用 Y：**
+
+| 想要做什么 | 用什么 | 具体命令/工具 |
+|---|---|---|
+| 想看今天有什么会 | CLI | `lark-cli calendar +agenda` |
+| 想约个会议 | CLI | `lark-cli calendar +create` |
+| 想发条消息给同事 | CLI | `lark-cli im +send` |
+| 想搜聊天记录 | CLI | `lark-cli im +search` |
+| 想写个新文档 | CLI | `lark-cli doc +create` |
+| 想给文档加格式化内容 | MCP | `batch_create_feishu_blocks` |
+| 想改文字颜色/对齐 | MCP | `update_feishu_block_text` |
+| 想查多维表格数据 | CLI | `lark-cli base +query` |
+| 想创建待办任务 | CLI | `lark-cli task +create` |
+| 想发邮件 | CLI | `lark-cli mail +compose` |
+| 想找个人的联系方式 | CLI | `lark-cli contact +search` |
+| 想查知识库 | CLI | `lark-cli wiki +list` |
+| 想看会议纪要 | CLI | `lark-cli minutes` |
+| 想画流程图 | CLI | `lark-cli docs +whiteboard-update` |
+| 想读文档全文 | MCP | `get_feishu_document_text` |
+| 想分析文档结构 | MCP | `get_feishu_document_blocks` |
+| 想上传文件 | CLI | `lark-cli drive upload` |
+| 想生成格式化报告 | CLI+MCP | 先查数据(CLI)，再写格式化文档(MCP) |
+
+---
+
+## 组合工作流技能
+
+以下技能组合 CLI 和 MCP，完成多步操作：
+
+| 工作流 | 说明 | 技能文件 |
+|---|---|---|
+| 文档审阅 | 读取文档 → 分析内容 → 创建任务 → 发送总结 | `skills/doc-review.md` |
+| 数据报告生成 | 查询数据 → 生成格式化文档 | `skills/data-to-doc.md` |
+| 知识同步 | Wiki 与文档之间双向同步 | `skills/knowledge-sync.md` |
+| 增强版会议纪要 | 会议纪要 + 格式化文档 + 任务创建 | `skills/meeting-summary-plus.md` |
+
+---
+
+## 操作前检查
+
+执行任何飞书操作前，按顺序验证：
+
+1. CLI 已安装：`which lark-cli`
+2. CLI 已认证：`lark-cli auth status`
+3. CLI 技能已加载：`lark-cli skills list`
+4. MCP 环境变量已设置：`FEISHU_APP_ID` 和 `FEISHU_APP_SECRET` 非空
+5. mcporter 已安装：`which mcporter`
+
+如果任何检查失败，参考 `TROUBLESHOOT-AGENT.md` 进行修复。
+
+---
+
+## 决策原则（简版）
+
+1. 非文档域一律用 CLI（消息、日历、邮件、任务等）
+2. 精确 Block 编辑用 MCP（颜色、对齐、批量格式化）
+3. CLI 和 MCP 都能做时优先 CLI
+4. 图表和画板编辑用 CLI
+5. 多步工作流可组合两层
+
+完整决策矩阵参见 `SKILL.md`。
diff --git a/skills/knowledge-sync.md b/skills/knowledge-sync.md
new file mode 100644
index 0000000..667e4ea
--- /dev/null
+++ b/skills/knowledge-sync.md
@@ -0,0 +1,186 @@
+---
+name: knowledge-sync
+description: 知识同步工作流 — Wiki 与文档之间的双向同步：从 Wiki 读取内容更新到文档(CLI→MCP)，或从文档内容创建/更新 Wiki 节点(MCP→CLI)。
+metadata:
+ bins:
+ - lark-cli
+ - mcporter
+---
+
+# 知识同步工作流
+
+组合 CLI 和 MCP，在飞书 Wiki 知识库和云文档之间实现双向内容同步。
+
+---
+
+## When to Use This Skill
+
+- User wants to sync Wiki content to a document ("把知识库里的内容同步到文档")
+- User wants to update Wiki from document content ("把文档内容更新到知识库")
+- User needs to consolidate information from Wiki into a summary doc ("把几篇 Wiki 整理成一份文档")
+- User needs to keep Wiki and standalone documents in sync
+
+---
+
+## Prerequisites
+
+1. Confirm CLI auth: `lark-cli auth status` — must show valid token.
+2. Confirm MCP auth: `FEISHU_APP_ID` and `FEISHU_APP_SECRET` env vars set.
+3. Required scopes: `wiki:wiki`, `docx:document`. Verify at `https://open.feishu.cn/app/{APP_ID}/security`.
+4. Determine sync direction:
+ - **Direction A**: Wiki → Document (read Wiki, write to doc)
+ - **Direction B**: Document → Wiki (read doc, write to Wiki)
+
+---
+
+## Direction A: Wiki to Document
+
+Use this when: extracting Wiki content and writing it into a standalone document or consolidating multiple Wiki pages into one document.
+
+### Step A1: Read Wiki Content
+
+**Tool**: CLI `lark-cli wiki`
+
+First, list available knowledge spaces to find the target:
+
+```bash
+lark-cli wiki +list
+```
+
+Browse the Wiki structure to locate the target node(s). Note the wiki_token(s) for the pages to sync.
+
+To get detailed metadata about a specific node:
+
+```
+MCP tool: get_feishu_wiki_node_info
+Parameters: { "token": "" }
+```
+
+This returns the node's title, type, parent, and the associated document_id (obj_token).
+
+**Error handling**:
+- If "space not found" → ask user for the correct knowledge space name or ID.
+- If "permission denied" → ask user to add `wiki:wiki` or `wiki:wiki:readonly` scope.
+
+### Step A2: Extract Key Information
+
+**Tool**: MCP `get_feishu_document_text` + `get_feishu_document_blocks`
+
+Wiki nodes are backed by documents. Use the obj_token (document_id) from Step A1 to read content:
+
+```
+MCP tool: get_feishu_document_text
+Parameters: { "document_id": "" }
+```
+
+For structured extraction (headings, lists, code blocks):
+
+```
+MCP tool: get_feishu_document_blocks
+Parameters: { "document_id": "" }
+```
+
+Process the content:
+1. Extract key sections by heading structure.
+2. Identify content that has changed since last sync (compare with target doc if it exists).
+3. Organize extracted content into a coherent structure for the target document.
+
+If syncing multiple Wiki pages, repeat for each page and merge content logically.
+
+**Error handling**:
+- If "document not found" → the wiki node may not have an associated document. Ask user to verify the Wiki page exists and has content.
+
+### Step A3: Update Target Document
+
+**Tool**: CLI `lark-cli doc +create` (if new) + MCP `batch_create_feishu_blocks` + `update_feishu_block_text`
+
+If the target document does not exist, create it first with `lark-cli doc +create`.
+
+Then add extracted content as formatted blocks via `batch_create_feishu_blocks`:
+- Use `heading2` for source page titles, `heading3` for sections, `text` for content.
+- Include a "同步时间: YYYY-MM-DD HH:MM" block at the top.
+
+To update existing blocks (rather than append), use `update_feishu_block_text` with the target block_id. To remove outdated blocks, use `delete_feishu_document_blocks`.
+
+**Error handling**:
+- If "block not found" → re-fetch block list with `get_feishu_document_blocks`.
+- If batch creation fails → split into smaller batches and retry.
+
+---
+
+## Direction B: Document to Wiki
+
+Use this when: publishing document content into a Wiki knowledge base for team access, or keeping a Wiki page updated from a source document.
+
+### Step B1: Read Document Content
+
+**Tool**: MCP `get_feishu_document_text` + `get_feishu_document_blocks`
+
+Read the source document with `get_feishu_document_text` for plain text and `get_feishu_document_blocks` for structure.
+
+**Error handling**:
+- If "document not found" → verify document_id with user.
+- If "permission denied" → check `docx:document:readonly` scope.
+
+### Step B2: Extract and Transform Content
+
+**Tool**: AI analysis (no external tool needed)
+
+Process the document content:
+1. Identify the main sections and their hierarchy.
+2. Extract key content to sync to Wiki.
+3. If the target Wiki node already exists, determine what has changed.
+4. Format the content appropriately for the Wiki target.
+
+### Step B3: Create or Update Wiki Node
+
+**Tool**: CLI `lark-cli wiki +create` (new node) or MCP `batch_create_feishu_blocks` + `update_feishu_block_text` (existing node)
+
+For new Wiki nodes, use `lark-cli wiki +create`. The CLI prompts for target knowledge space, parent node, title, and content (Markdown).
+
+For existing Wiki nodes, use the node's associated document_id (from `get_feishu_wiki_node_info`) and update via MCP: `batch_create_feishu_blocks` to add blocks, `update_feishu_block_text` to modify existing blocks.
+
+**Error handling**:
+- If "space not found" → ask user which knowledge space to use.
+- If "permission denied" → ask user to add `wiki:wiki` scope (write, not readonly).
+- If "parent node not found" → list nodes with `lark-cli wiki +list` and let user choose parent.
+
+---
+
+## Output Format
+
+Report sync results to the user:
+
+```
+## 知识同步完成
+
+**方向**: Wiki → 文档 / 文档 → Wiki
+**来源**: [来源标题](来源链接)
+**目标**: [目标标题](目标链接)
+**同步时间**: YYYY-MM-DD HH:MM
+
+### 同步内容
+- 新增: X 个内容块
+- 更新: X 个内容块
+- 删除: X 个内容块
+
+### 注意事项
+- （如有冲突或需手动检查的内容列出）
+```
+
+---
+
+## Example Prompts
+
+- "把技术文档知识库里的 API 指南同步到这个文档"
+- "把这篇文档发布到产品知识库里"
+- "把三篇 Wiki 整理成一份汇总文档"
+- "更新知识库里的部署文档，用这个文档的最新内容"
+
+---
+
+## Partial Execution
+
+- "只读取 Wiki 内容" → Execute Step A1-A2 only.
+- "只把文档内容推到 Wiki" → Execute Direction B.
+- "看看两边有什么不同" → Execute A1-A2 and B1-B2, compare and report differences.
diff --git a/skills/meeting-summary-plus.md b/skills/meeting-summary-plus.md
new file mode 100644
index 0000000..5684665
--- /dev/null
+++ b/skills/meeting-summary-plus.md
@@ -0,0 +1,186 @@
+---
+name: meeting-summary-plus
+description: 增强版会议纪要 — 在 CLI 内置会议纪要工作流基础上，用 MCP 生成格式化文档，创建待办任务，并发送通知到群聊。
+metadata:
+ bins:
+ - lark-cli
+ - mcporter
+---
+
+# 增强版会议纪要
+
+在 CLI 内置的 `workflow-meeting-summary` 基础上增强：生成格式化的会议纪要文档，高亮关键决策，创建待办任务，并通知相关群组。
+
+---
+
+## When to Use This Skill
+
+- User wants a formatted meeting summary document ("帮我整理会议纪要并生成文档")
+- User wants action items from meetings converted to tasks ("把会议待办变成任务")
+- User needs to send meeting summary to a group ("整理完会议纪要发到群里")
+- User wants more than the basic CLI meeting summary — needs styling, tasks, and notifications
+- User mentions "增强版会议纪要" or "会议纪要 plus"
+
+---
+
+## Prerequisites
+
+1. Confirm CLI auth: `lark-cli auth status` — must show valid token.
+2. Confirm MCP auth: `FEISHU_APP_ID` and `FEISHU_APP_SECRET` env vars set.
+3. Required scopes: `vc:meeting:readonly`, `minutes:minute:readonly`, `docx:document`, `task:task`, `im:message`.
+4. Determine the time range or specific meeting to summarize.
+
+---
+
+## Step-by-Step Workflow
+
+### Step 1: Retrieve Meeting Data via CLI
+
+**Tool**: CLI `lark-cli workflow meeting-summary` or `lark-cli minutes` + `lark-cli vc`
+
+**Option A** — Use the built-in workflow (recommended for time-range summaries):
+
+```bash
+lark-cli workflow meeting-summary
+```
+
+This aggregates meetings within a time range and produces a structured summary including:
+- Meeting list with titles, times, attendees
+- AI-generated summaries per meeting
+- Action items / todos
+- Key discussion chapters
+
+**Option B** — Manual approach for a single meeting:
+
+First, find the meeting record:
+
+```bash
+lark-cli vc
+```
+
+Then get the detailed minutes:
+
+```bash
+lark-cli minutes
+```
+
+This returns: title, duration, AI summary, todos, chapters, and transcript.
+
+**Error handling**:
+- If "skill not found" → run `npx skills add larksuite/cli -y -g` and retry.
+- If "no meetings found" → widen the time range or verify the date.
+- If "permission denied" → ask user to add `vc:meeting:readonly` and `minutes:minute:readonly` scopes.
+
+### Step 2: Create Formatted Summary Document
+
+**Tool**: CLI `lark-cli doc +create` + MCP `batch_create_feishu_blocks`
+
+First, create the document:
+
+```bash
+lark-cli doc +create
+```
+
+Provide a brief Markdown title like "会议纪要 — YYYY-MM-DD". Extract the document_id from the returned URL.
+
+Then, add formatted content blocks via `batch_create_feishu_blocks`. Build the blocks array dynamically from Step 1's data using this structure:
+- `heading1`: document title with date
+- `heading2`: major sections (会议概览, AI 摘要, 关键决策, 待办事项, 讨论要点)
+- `heading3`: individual topics/agenda items
+- `text`: content paragraphs (meeting info, summaries, discussion details)
+- `list`: decisions and action items
+
+For multiple meetings, add a `heading2` section per meeting.
+
+**Error handling**:
+- If document creation fails → check auth with `lark-cli auth status`.
+- If batch creation fails → split into smaller batches (max ~50 blocks per call).
+
+### Step 3: Style Key Decisions and Action Items
+
+**Tool**: MCP `update_feishu_block_text`
+
+Get the block_ids from Step 2's response and apply styling with `update_feishu_block_text`:
+
+- **blue** (`text_color: "blue"`): key decisions
+- **red** (`text_color: "red"`): urgent or overdue action items
+- **green** (`text_color: "green"`): completed items
+- **purple** (`text_color: "purple"`): important notes or reminders
+
+**Error handling**:
+- If "block not found" → re-fetch blocks with `get_feishu_document_blocks` and find correct block_id.
+- If styling fails → content is still readable. Log warning and continue.
+
+### Step 4: Create Tasks from Action Items
+
+**Tool**: CLI `lark-cli task +create`
+
+For each action item extracted in Step 1, create a task:
+
+```bash
+lark-cli task +create
+```
+
+Provide for each task:
+- **Title**: the action item description
+- **Assignee**: the person mentioned (if identifiable)
+- **Due date**: extracted from the meeting notes, or default based on urgency
+- **Description**: include the meeting title and date for context
+
+**Error handling**:
+- If "permission denied" → ask user to add `task:task` scope.
+- If assignee not found → create the task without assignee and note it in the output.
+- If no action items found → skip this step and inform user.
+
+### Step 5: Send Notification to Group
+
+**Tool**: CLI `lark-cli im +send`
+
+Send a notification via `lark-cli im +send`. Include: meeting title, date, document link, count of decisions and tasks created, and a 2-3 sentence summary.
+
+**Error handling**:
+- If user doesn't specify a chat → ask which group to notify, or skip this step.
+- If "chat not found" → ask user for the correct chat ID or group name.
+- If "permission denied" → ask user to add `im:message` scope.
+
+---
+
+## Output Format
+
+Present the result to the user:
+
+```
+## 会议纪要整理完成
+
+**文档**: [会议纪要 — YYYY-MM-DD](文档链接)
+**会议数**: X 场
+**关键决策**: X 条
+**待办任务**: X 条（已创建）
+**通知**: 已发送到 [群组名称]
+
+### 关键决策
+1. ...
+2. ...
+
+### 已创建的任务
+1. [负责人] 任务描述 — 截止 YYYY-MM-DD
+2. [负责人] 任务描述 — 截止 YYYY-MM-DD
+```
+
+---
+
+## Example Prompts
+
+- "帮我整理今天的会议纪要，生成文档并发到项目群"
+- "把昨天的会议内容整理成文档，待办事项帮我建成任务"
+- "整理本周所有会议纪要，生成一份格式化的汇总文档"
+- "增强版会议纪要：上午的产品评审会"
+
+---
+
+## Partial Execution
+
+- "只整理会议内容，不建任务" → Execute Steps 1-3, skip Steps 4-5.
+- "整理完帮我发群里就行" → Execute Steps 1-2 and 5, skip Steps 3-4.
+- "只看看有哪些会议和待办" → Execute Step 1 only, present raw data.
+- "会议纪要已有文档了，帮我建任务和发通知" → Skip Steps 1-3, execute Steps 4-5 from provided doc.