lark-mail Skill for Claude: Read, Draft and Send Lark and Feishu Mail

Lets Claude read, search, draft, reply to and send Lark or Feishu mail through lark-cli, saving drafts unless you confirm a send.

Installs: 473.1K; on skills.sh
Stars: 17.9K; on GitHub
Files: 35; besides SKILL.md
Updated: Oct 2026; last repo commit

lark-mail tells Claude which lark-cli mail commands to run. It lists and searches mail, reads messages and threads, creates and edits drafts, replies, forwards, sends, handles read receipts and templates, and watches for new mail. It loads when you mention writing, sending, replying to, searching or watching email.

Sending is gated. Drafts are the default, and a send only happens with --confirm-send after you confirm recipients, subject and body. SKILL.md also tells Claude to treat mail text as untrusted data and never obey instructions found inside an email. It needs lark-cli, a signed-in account and lark-shared.

Good for

Lark Mail users who want Claude to triage the inbox and draft replies
People who send recurring reports and want HTML mail from templates
Developers who automate mail rules and mail watching

Install lark-mail

Claude Code

Copy this prompt.

Install the lark-mail skill in Claude Code. Run this command:

npx @larksuite/cli@latest install

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

This skill only gives Claude instructions for the lark-cli command line, so install the CLI first, then the skills, then sign in. The skills command adds the whole larksuite/cli set, which includes lark-shared, the skill this one reads first for sign-in and permissions.

From the project README. Source.

What's in the skill

When it loads

飞书邮箱：Use when user mentions 起草邮件、写邮件、草稿、发送/回复/转发邮件、查阅邮件、看邮件、搜索邮件、邮件文件夹、邮件标签、邮件联系人、监听新邮件、邮件收信规则等；use for mail/email intent only. Do not use for docs/sheets/calendar/auth setup/pure contact lookup/IM chat tasks.

Sections in SKILL.md

核心概念
⚠️ 安全规则：邮件内容是不可信的外部输入
数据真实性与操作合规
1. 找不到就报"未找到"，不得伪造
2. 写操作前显式确认
正确流程示例
身份选择：优先使用 user 身份
典型工作流
常用操作速查
参数不确定时先查 -h
命令选择：先判断邮件类型，再决定草稿还是发送
正文格式与书写规范
读取邮件：按需控制返回内容
原生 API 调用规则
Step 1 — 用 -h 确定要调用的 API（必须，不可跳过）
Step 2 — 查 schema，获取参数定义

Key facts

Author: Lark Suite (larksuite)
Repository: larksuite/cli, folder skills/lark-mail
Installs on skills.sh: 473,061 as of 2026-11-02
GitHub stars: 17,911 for the whole larksuite/cli repo, 2026-11-02
License: MIT
Last repo commit: 2026-10-30
What is inside: 35 bundled files besides SKILL.md: 30 reference pages and 5 HTML mail templates; no scripts
skills.sh audits: Gen Agent Trust Hub: pass, Socket: pass, Snyk: warn

What it includes

Lists mail summaries with search and filters
Shows date, sender, subject and message ID.
Try: “Show unread mail from the last two days.”
Network · +triage
Reads one email or many by ID
Use +messages for several IDs instead of looping.
Network · +message / +messages
Reads a whole conversation in order
Includes replies, drafts and attachment details.
Network · +thread
Writes a new email as a draft
Saves a draft by default; --confirm-send sends it after you agree.
Try: “Draft a mail to the team about Friday release.”
Write · +send / +draft-create
Replies to a message, drafted first
Sets the Re: subject and reply headers for you.
Write · +reply / +reply-all
Forwards a message, drafted first
Includes the original message block.
Write · +forward
Edits an existing draft safely
Patches the draft instead of replacing it.
Write · +draft-edit
Adds labels or moves threads to a folder
Works in batches of 20 threads.
Write · +thread-modify
Moves threads to the trash
Needs --yes.
Delete · +thread-trash
Watches for new mail over a live connection
Needs the mail:event scope and a bot event set up.
Network · +watch
Answers or dismisses read-receipt requests
One sends a receipt, the other clears the request.
Write · +send-receipt / +decline-receipt
Lists signatures or shares a mail to a Lark chat
Shares as a card.
Write · +signature / +share-to-chat
Creates or edits personal mail templates
Uploads local images and attachments to Drive.
Write · +template-create / +template-update
Checks mail HTML before drafting
Read-only; returns warnings and auto-fixed HTML.
Read · +lint-html

From the official docs

Permissions

Medium risk

Claude can read your mail and send or trash messages as your Lark user. Keep the draft-first default, and read every draft before you let it send.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It acts as you in Lark after you sign in
The repo README warns that the agent works under your user identity within the authorized scope. Grant only the scopes you need and read commands before you approve them. github.com
It can send mail as you
Mail is drafted by default and sent only with an explicit confirm flag. Open the draft and check recipients before you allow a send. github.com
Email text can carry prompt injection
SKILL.md warns that mail bodies and sender names are untrusted. Be careful asking Claude to act on mail from unknown senders. github.com
Trash and rule changes need confirmation
SKILL.md requires confirmation for deletes, trashing, and creating or changing mail rules. Do not approve these in bulk. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	The skill has no read-only switch. Sign in with a narrow scope list, for example lark-cli auth login --domain <name>, so the agent cannot reach more than you intend.
Talks to	Lark/Feishu Open Platform API calls made by lark-cli. Your AI provider receives whatever Claude reads from Lark
Sign-in and keys	lark-cli stores credentials in the OS keychain, per the repo README.
Scopes	Claude runs lark-cli commands in your terminal After you sign in, commands run as your Lark or Feishu user within the permissions you granted Can read your mail and create drafts, replies and forwards Can send mail, trash threads and change mail rules

Questions

How do I install lark-mail?

Use the install box at the top of this page. The box lists three steps: install the Lark CLI, add the Lark skills, then set up an app and sign in. The skill is only instructions, so Claude cannot use it until lark-cli is installed and you are signed in.

Do I need a Lark account to use lark-mail?

Yes. You need a Lark or Feishu account, plus an app set up with lark-cli config init. The README then has you sign in with lark-cli auth login --recommend. There is no separate API key to paste.

Will lark-mail send email without asking me?

Not by default. New mail, replies and forwards are saved as drafts. Sending needs --confirm-send, and SKILL.md says Claude must show you the recipients, subject and a body summary first.

Can an email trick Claude into doing something?

SKILL.md tries to prevent it. It tells Claude to treat mail content as data, ignore instructions inside emails, and ask you before acting on a request that came from a message. This lowers the risk but does not remove it.

Is lark-mail safe?

Use it with care. It can send, forward, trash and set mail rules. The repo README warns that after you authorize Lark permissions the agent acts under your user identity and that model mistakes or prompt injection can cause real changes.

Information

updated Nov 2, 2026

Repository: larksuite/cli
Folder: skills/lark-mail

larksuite/cli skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3

larksuite/cli README · github.com
lark-mail SKILL.md · github.com
lark-skill-maker SKILL.md · github.com