feishu-binding

Q: How do I install feishu-binding?

Run npx skills add https://github.com/starchild-ai-agent/official-skills --skill feishu-binding in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/feishu-binding/ (personal) or .claude/skills/feishu-binding/ (project).

by starchild-ai-agentProductivity & Automation

Feishu/Lark binding: device flow authorization, connect, disconnect, status check. Use when setting up or managing Feishu/Lark connection (e.g. connect Feishu, connect Lark, bind 飞书, check Feishu status, disconnect Lark).

Installs: 1.4K
Last 24h: —
Trending rank: —
Security: Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/starchild-ai-agent/official-skills --skill feishu-binding

01
Install
Run the command in your project. The CLI asks which agents to install for.
02
Describe the task
Ask your agent for something this skill covers. It loads feishu-binding automatically when relevant.
03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

📱 Feishu/Lark Binding

Connect / disconnect the user's Feishu (飞书) or Lark account so the agent can chat via the Feishu/Lark app.

The feishu tool stays built-in. This SKILL.md is the reference doc.

Brand selection

Brand	Region	Domain
`feishu`	China mainland (中国大陆)	feishu.cn / accounts.feishu.cn
`lark`	International	larksuite.com / accounts.larksuite.com

Ask the user which brand they use. Default to feishu if unclear. Chinese-speaking users almost always use feishu.

Typical binding flow

connect(brand) → user scans QR / opens link → poll(device_code) → done

Start device flow: feishu(action="connect", brand="feishu") — returns verification_uri, device_code, and a QR code image saved to workspace.
Show BOTH options to the user:
- QR code scan: Display the QR code image inline using the markdown_image from the result (e.g. ![Feishu QR Code](feishu_qrcode_xxx.png)). The user scans it with their Feishu/Lark app.
- Direct link: Show the verification_url as a clickable link. The user opens it in their browser, logs in to Feishu/Lark, and confirms.
Wait for the user to confirm. Don't auto-poll — let them say "done" / "confirmed" / "已确认" / "scanned" first.
Poll for completion: feishu(action="poll", device_code=<from step 1>, brand=<same brand>).
- status: "done" → binding complete, congratulate the user.
- status: "pending" → ask the user if they've scanned the QR code or opened the link and confirmed.
- status: "expired" → the device flow expired (typically 5 minutes). Start over with connect.
Confirm to user: "Feishu/Lark connected! You can now chat with your agent on Feishu/Lark."

Actions

action	required params	purpose
`status`	—	Current Feishu/Lark app state. Use to check if already connected.
`connect`	`brand` (optional, default "feishu")	Start device flow. Returns verification URL + device_code.
`poll`	`device_code`, `brand`	Check if user has confirmed authorization. Returns status.
`disconnect`	—	Unbind Feishu/Lark app (destructive — confirm with user first).

Channel-aware QR / link display

User channel	How to show
Web	Include `file_path` (QR image) in your reply — frontend renders it inline. Also show the link.
Telegram	`send_to_telegram(file_path=<qr_path>, message_type="photo")` + include the link in the caption
WeChat	Show the link only (user can't scan a QR inside WeChat for Feishu)
Feishu	(User is already on Feishu — they don't need to bind. Tell them it's already connected or check `status`.)

Key differences from WeChat / Telegram

Aspect	Feishu/Lark	WeChat	Telegram
Auth method	Device flow (URL)	QR code scan	Bot token
User action	Open URL + confirm in app	Scan QR + confirm	Create bot via @BotFather
Credential	None (device flow handles it)	bot_token (from QR)	bot_token (from BotFather)
Brand choice	feishu / lark	N/A	N/A

Critical rules

Don't auto-poll after connect. Wait for user confirmation that they've opened the link and confirmed in Feishu/Lark. Auto-polling wastes API calls.
Device flow expires in ~5 minutes. If poll returns expired, tell the user and start a new connect.
disconnect is destructive — confirm with the user before calling it. It will stop the Feishu/Lark gateway instance.
Brand matters — feishu and lark use different API domains. Using the wrong brand will fail silently or redirect to the wrong login page.
One app per user — a user can only have one Feishu/Lark app at a time. If they want to switch brands, they must disconnect first, then connect with the new brand.

Troubleshooting

Symptom	Likely cause	Fix
`connect` returns error	User already has an active app	Call `status` first; if active, ask if they want to `disconnect` and rebind
`poll` returns `expired`	User took too long to confirm	Start a new `connect`
`poll` returns `pending` repeatedly	User hasn't opened the URL yet	Remind them to open the verification URL
User says "I can't find the link"	URL was in a previous message	Re-run `connect` to get a fresh URL

FAQ

What does the feishu-binding skill do?

How do I install feishu-binding?

Run npx skills add https://github.com/starchild-ai-agent/official-skills --skill feishu-binding in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/feishu-binding/ (personal) or .claude/skills/feishu-binding/ (project).

Which AI agents can use feishu-binding?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is feishu-binding safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains feishu-binding?

It is published by starchild-ai-agent in starchild-ai-agent/official-skills. It has 1,391 recorded installs.