lark-cli

Lark CLI

Install and bootstrap the official Lark/Feishu CLI for later Lark operations.

This skill is for CLI setup and shared operating rules. It does not replace domain skills such as lark-im for messages, lark-contact for resolving people, or lark-markdown for Drive-native Markdown files.

Source

This workflow follows the official larksuite/cli README quick start for AI agents, with one repo rule: always pass an explicit --brand.

npx @larksuite/cli@latest install
lark-cli config init --new --brand lark
lark-cli auth login --recommend
lark-cli auth status

The upstream quick start omits --brand. This skill must not omit it: use --brand lark when the user explicitly says Lark, otherwise use --brand feishu.

What This Skill Does

• checks whether lark-cli is already available
• installs or upgrades the CLI through the official npm package
• explains where the command and package are installed
• initializes Lark/Feishu Open Platform app credentials with the correct product suite
• starts OAuth login with recommended scopes and the saved product suite
• prints setup/auth URLs from the CLI for the user to open manually
• verifies the active auth state
• gives shared rules for identity, output parsing, and side-effect safety

What This Skill Needs

• Node.js with npm and npx
• network access to npm and the binary download source
• a browser or manual access to the URLs printed by config init and auth login
• user approval for app setup and OAuth authorization

If Node.js/npm is missing, stop and ask the user to install Node.js first. Do not try to install system Node.js from this skill.

Install Timing

Use this skill before other Lark skills when:

• the user asks to set up Lark/Feishu CLI
• lark-cli is not found
• a Lark skill fails because CLI config or auth is missing
• the user asks where lark-cli is installed
• the user wants to switch or verify user versus bot execution

Do not reinstall on every Lark request. First run:

bash <skill-path>/scripts/setup_lark_cli.sh check

Install only if the check says lark-cli is missing, or if the user explicitly asks to upgrade/reinstall.

Install Method

Preferred installer:

npx @larksuite/cli@latest install

The installer performs a guided setup. It globally installs @larksuite/cli, installs the upstream Lark AI skills, then offers app configuration and authorization.

For non-interactive agent setup, use the wrapper script:

bash <skill-path>/scripts/setup_lark_cli.sh setup --suite lark

This runs:

1. install when lark-cli is missing
2. lark-cli config init --new --brand <lark|feishu>
3. lark-cli auth login --recommend
4. lark-cli auth status

The config and login steps may print browser URLs. Send those URLs to the user and wait for them to finish authorization when needed.

Product Suite

Choose the product suite before config init.

• If the user says Lark, international Lark, larksuite, global, overseas, or English Lark, pass --suite lark.
• If the user says Feishu, Feishu, domestic Feishu, or Chinese Feishu, pass --suite feishu.
• If the user does not specify Lark versus Feishu, default to --suite feishu.
• If the CLI is already configured, auth login uses the saved product suite. To change the authorization link family, rerun config-init or setup with the desired --suite.

The wrapper maps --suite to the official lark-cli config init --brand option. If --suite is omitted, the wrapper still passes --brand feishu explicitly:

bash <skill-path>/scripts/setup_lark_cli.sh config-init --suite lark
bash <skill-path>/scripts/setup_lark_cli.sh config-init --suite feishu
bash <skill-path>/scripts/setup_lark_cli.sh config-init

Do not confuse product suite with auth business domains:

• --suite lark|feishu chooses the Lark versus Feishu setup/authorization link family.
• --domain calendar|docs|drive|all on auth login chooses permission business domains.

Setup Command Rules

Avoid raw ambiguous setup commands during agent setup:

lark-cli config init
lark-cli config init --new

Raw lark-cli config init commands rely on upstream defaults and make it too easy to miss a user request for Lark.

Prefer the wrapper:

bash <skill-path>/scripts/setup_lark_cli.sh config-init --suite lark
bash <skill-path>/scripts/setup_lark_cli.sh config-init --suite feishu
bash <skill-path>/scripts/setup_lark_cli.sh config-init
bash <skill-path>/scripts/setup_lark_cli.sh setup --suite lark
bash <skill-path>/scripts/setup_lark_cli.sh setup --suite feishu
bash <skill-path>/scripts/setup_lark_cli.sh setup

The wrapper default is Feishu. If an existing run accidentally created a Feishu profile when the user requested Lark, rerun config-init --suite lark before auth login; plain auth login will keep using the saved Feishu brand.

Authorization Links

The wrapper does not open browser links automatically. Run setup/login and copy the printed setup or authorization URL to the user:

bash <skill-path>/scripts/setup_lark_cli.sh setup --suite lark

Keep URLs intact and send them to the user exactly as printed. Treat setup/auth URLs as opaque strings: do not edit, re-encode, or add punctuation inside the URL.

Install Location

The official npm package installs globally.

On macOS/Linux, the command normally resolves to:

$(npm prefix -g)/bin/lark-cli

The npm package normally lives at:

$(npm root -g)/@larksuite/cli

Inside that package, the downloaded native binary is:

$(npm root -g)/@larksuite/cli/bin/lark-cli

On Windows, the command is normally under:

%APPDATA%\npm\lark-cli.cmd

Use the wrapper to print the current machine's resolved paths without exposing credentials:

bash <skill-path>/scripts/setup_lark_cli.sh check

Commands

Check current install

bash <skill-path>/scripts/setup_lark_cli.sh check

Install the CLI

bash <skill-path>/scripts/setup_lark_cli.sh install

Initialize app credentials

bash <skill-path>/scripts/setup_lark_cli.sh config-init

This defaults to lark-cli config init --new --brand feishu. Use --suite lark for international Lark. The command may require the user to open a browser URL and complete app setup.

Log in

bash <skill-path>/scripts/setup_lark_cli.sh login

This runs lark-cli auth login --recommend. If it prints an authorization URL, send that URL to the user and wait for them to complete authorization.

Use exact scopes only when the user or downstream skill requires them:

bash <skill-path>/scripts/setup_lark_cli.sh login --scope "im:message:send_as_bot"

Verify auth

bash <skill-path>/scripts/setup_lark_cli.sh status

Identity Rules

Lark commands can run as a user or bot. Do not guess when it matters.

• Use --as user when the user wants actions under their personal Lark identity.
• Use --as bot when sending from or operating as the configured app bot.
• For user-facing write actions, confirm the target, message/content, and identity before executing without dry-run.

Example:

lark-cli im +messages-send --as bot --chat-id "oc_xxx" --text "Hello" --dry-run

Output Rules

Prefer JSON for parsing:

lark-cli auth status --format json

With --format json, success is indicated by:

{ "ok": true }

Errors are written to stderr with:

{ "ok": false, "error": { "type": "...", "message": "..." } }

Check exit code or ok == true. Do not expect a top-level code: 0 in shortcut command success envelopes.

Safety Rules

• For side-effect commands, run --dry-run first when the command supports it.
• Do not print tokens, cookies, app secrets, refresh tokens, or raw local credential files.
• Do not relax lark-cli default security settings unless the user explicitly asks and understands the risk.
• If a command reports missing scopes, use lark-cli auth check <scope> or rerun lark-cli auth login with the needed scope/domain.
• If a user asks to send a Lark message, use lark-im after this setup skill has verified install and auth.
• If a user provides only a person name or email for messaging, use a contact lookup skill before sending.
• If a user asks to manage Drive-native .md files, use lark-markdown; it is not the IM message sender.

Example Prompts

Chinese

• "Help me install lark-cli, and explain where it is installed"
• "Check whether this machine has lark-cli; if it is not installed, initialize it"
• "Configure Lark CLI app credentials, then log in with the recommended permissions"
• "Check whether the current lark-cli is a user or bot identity"
• "Before sending a Feishu message, first install and log in to lark-cli"

English

• "Install lark-cli and show me where it was installed."
• "Set up Lark CLI for AI agent usage."
• "Run lark-cli config init and auth login with recommended scopes."
• "Check whether lark-cli is authenticated."
• "Prepare lark-cli before using lark-im to send a message."