PlugWiki

Lark Shared Skill for Claude Code: Auth and Safety

The base rules every lark-* skill uses: sign-in, user or bot identity, permissions and confirmation before risky writes.

Installs
480.8K
on skills.sh
Stars
17.9K
on GitHub
Files
6
besides SKILL.md
Updated
Oct 2026
last repo commit

lark-shared is not a task skill. It is the shared base for the whole lark-* set that drives Lark (Feishu) from the lark-cli program: how to sign in, whether to act as you or as the app bot, how to fix missing permissions, and how to treat risky actions.

Other lark skills tell Claude to read it first. It also applies when you ask for login, status, logout, permission scopes, revoking access or a _notice message. Its text is written in Chinese. It bundles six reference notes and no scripts.

Good for

  • Anyone setting up lark-cli with Claude for the first time
  • Teams deciding whether the agent acts as a user or as a bot
  • Admins who want risky Lark actions to need a confirmation

Install lark-shared

Claude Code

  1. 1

    Copy this prompt.

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

npx @larksuite/cli@latest install
npx skills add larksuite/cli -y -g

Then check that it loaded and tell me how to use it.
  2. 2

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

Requires Node.js.

The README gives two routes. Option 1 (recommended) is npx @larksuite/cli@latest install. Option 2 builds from source and then runs npx skills add larksuite/cli -y -g, which the README marks as required. It does not say whether Option 1 also installs the skills, so this box runs both. Every lark-* skill needs the lark-cli program and sends Claude to lark-shared for sign-in. auth login --recommend picks commonly used scopes; use --domain to limit them.

From the project README. Source.

What's in the skill

When it loads

Use for lark-cli setup/auth tasks: auth login/status/logout, user vs bot identity, business-domain permissions (--domain, including all/docs/drive), missing scopes, revoking authorization, or handling _notice JSON.

Sections in SKILL.md

  1. 通用准则
  2. 安全规则
  3. Reference 强触发索引

Key facts

Author
larksuite (larksuite/cli)
Installs on skills.sh
480,780 (Nov 2, 2026)
GitHub stars
17,911 on larksuite/cli (Nov 2, 2026)
License
MIT
Last repo commit
Oct 30, 2026
What installs
SKILL.md plus 6 bundled files, mostly reference notes under references/
skills.sh audits
Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass
Needs
The lark-cli program installed and signed in, plus the lark-shared skill
Rules
Never print secrets; confirm before writes and deletes; use --dry-run when a command has it
Relies on
lark-cli, per its metadata

What it includes

  • --as user acts as you; --as bot acts as the app

    A bot reading your personal resources returns an empty success, not an error.

    Read · Choose identity

  • auth login, status and logout; domain permissions and missing scopes

    Covers --domain choices such as all, docs and drive, and revoking authorization.

    Network · Auth and scopes

  • Shows a QR code beside any login or console URL

    Uses lark-cli auth qrcode and forwards the URL unchanged.

    QR codes for auth URLs

  • Success is ok == true or exit code 0

    Not code == 0, which only appears in errors.

    JSON output contract

  • Exit code 10 means stop and ask you

    Retries with the confirm flag only after your explicit yes.

    High-risk gate

  • No secrets in the terminal; confirm writes and deletes; relative file paths only

    Prefers --dry-run for dangerous requests.

    Safety rules

  • Five reference notes read on demand

    Identity and permissions, output contract, high-risk approval, config init, update notice.

    Reference triggers

From the official docs

Permissions

Medium risk

It controls sign-in and which identity the agent uses in Lark, so a wide login gives the agent wide reach. Limit permissions with --domain and keep the confirmation step on.

What it can do

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

Watch out for

  • It sets who the agent acts as in Lark

    With --as user the agent works with your permissions. Sign in with only the domains you need.

  • It adds a confirmation step for risky actions

    Exit code 10 stops the agent until you approve. Do not let it add the confirm flag on its own.

Technical details
Talks toLark/Feishu Open Platform
Scopes

  • Lark/Feishu sign-in and permission scopes

  • App configuration and credentials via lark-cli

Questions

What does lark-shared do?

It holds the shared rules for the lark-* skills: sign-in, user versus bot identity, scopes, the output format, and a confirmation step before risky writes.

How do I install it?

Follow the install box at the top of the page. Install the whole lark-* set from larksuite/cli, since each skill needs this one.

Do I use it directly?

Mostly no. Other lark skills tell Claude to read it. You meet it when you ask about login, status, logout, permissions or a _notice message.

What is the difference between --as user and --as bot?

--as user acts as you and reaches your calendar, drive and other personal resources. --as bot acts as the app and reaches only the app's own resources.

What happens at exit code 10?

It is a high-risk confirmation gate, not an error. The agent must stop, show you the action and risk, and retry with the confirm flag only after you say yes.

Is it safe?

README security note: after you authorize, the agent acts under your identity within the granted scope, and the README warns of prompt injection and data leaks. Sign in with limited domains and keep the confirmation gate on.

Information

updated Nov 2, 2026

Repository
larksuite/cli
Folder
skills/lark-shared

larksuite/cliskills.sh

Sources

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

Show all 3
  1. larksuite/cli README · github.com
  2. lark-shared SKILL.md · github.com
  3. skills.sh listing · skills.sh

Updated Nov 2, 2026.