Feishu CLI AI Agent Skills

What Are AI Agent Skills

AI Agent Skills are structured instruction documents (SKILL.md files) that teach AI coding assistants how to use Feishu CLI effectively. Rather than requiring developers to manually explain CLI commands to their AI tools, skills provide pre-written, comprehensive instructions that AI assistants read and follow automatically.

Feishu CLI ships with 19 skills covering all supported business domains, from messaging and calendar management to document creation and email handling. Each skill contains detailed command syntax, parameter descriptions, example workflows, and safety rules that ensure the AI assistant uses the CLI correctly and securely.

Installing skills is a one-time setup step. After running npx skills add larksuite/cli -y -g, the skill files are placed in a global directory that AI coding assistants scan at startup. From that point on, any AI assistant that supports the skills protocol can automatically discover and use Feishu CLI commands without additional configuration.

Supported AI Tools

Feishu CLI skills are compatible with a growing list of AI coding assistants. The skills protocol is an open standard, and any tool that reads SKILL.md files from the global skills directory can leverage Feishu CLI skills:

Claude Code — Anthropic's CLI-based coding assistant. Skills are loaded as context instructions when Claude Code starts, enabling it to use all Feishu CLI commands.
Cursor — AI-powered code editor. Skills appear as available context that Cursor's AI can reference when answering questions or writing code.
Codex — OpenAI's coding agent. Skills provide structured command documentation that Codex uses to generate accurate CLI invocations.
OpenCode — Open-source AI coding assistant with native skills support.
OpenClaw — Community-driven AI coding tool with skills protocol compatibility.

The skills system works by placing instruction files in a well-known directory. When an AI assistant starts a session, it scans this directory and loads any relevant skills based on the task context. The skill's description field in its YAML frontmatter is used for matching — when a user's request mentions calendars, for example, the AI assistant automatically loads the lark-calendar skill.

Complete Skill List

The following table lists all 19 skills included with Feishu CLI. Each skill covers a specific domain or workflow and contains detailed instructions for AI assistants:

Skill	Domain	Description
`lark-shared`	Foundation	Auto-loaded foundation skill. Covers authentication, configuration, identity switching (`--as user` / `--as bot`), and security rules. All other skills depend on this.
`lark-calendar`	Calendar	View agenda, create and update events, check free/busy status, and get AI-powered time suggestions for scheduling meetings across participants.
`lark-im`	Messaging	Send and receive messages, manage group chats, search message history, handle media attachments, and add reactions to messages.
`lark-doc`	Documents	Create, read, and update Lark documents using Markdown content. Supports media embedding, block-level operations, and whiteboard integration.
`lark-drive`	Drive	Upload and download files, manage sharing permissions, and work with file comments in Lark Drive.
`lark-sheets`	Spreadsheets	Full CRUD operations on Lark Spreadsheets including cell read/write, find and replace, data export, and sheet management.
`lark-base`	Bitable	Comprehensive Bitable automation with 60 shortcut commands for tables, records, fields, views, dashboards, and workflow automations.
`lark-task`	Tasks	Task CRUD operations, assignee management, completion tracking, reminder configuration, and tasklist organization.
`lark-mail`	Email	Send, reply, and forward emails. Manage drafts, triage inbox messages, and watch for new mail via WebSocket. Includes anti-prompt-injection security rules.
`lark-contact`	Contacts	Search for users by name, email, or department and retrieve detailed user profiles from the organizational directory.
`lark-wiki`	Wiki	Navigate and manage knowledge spaces and wiki nodes, including creating, moving, and organizing knowledge base content.
`lark-event`	Events	Subscribe to real-time events via WebSocket connections with NDJSON output format and regex-based event routing.
`lark-vc`	Video Conferencing	Access meeting records, retrieve meeting notes and action items, and download full meeting transcripts.
`lark-whiteboard`	Whiteboard	Render charts and diagrams using a DSL or Mermaid syntax, with PNG export for embedding in documents and messages.
`lark-minutes`	Minutes	Retrieve meeting minutes metadata and AI-generated artifacts including summaries, action items, and key decisions.
`lark-openapi-explorer`	API Discovery	Discover and explore undocumented or less-known Lark APIs by querying the official API documentation programmatically.
`lark-skill-maker`	Skill Creation	Create custom skills for Feishu CLI using the built-in skill authoring toolchain and template system.
`lark-workflow-meeting-summary`	Workflow	Aggregate meeting notes, transcripts, and action items from multiple meetings into consolidated summary reports.
`lark-workflow-standup-report`	Workflow	Combine calendar agenda, task status, and recent activity into formatted daily standup reports.

Skill File Structure

Each skill follows a consistent file structure that AI assistants are trained to parse. Understanding this structure is helpful if you plan to create custom skills or want to inspect what instructions the AI receives:

skills/
  lark-calendar/
    SKILL.md              # Main instruction file
    references/
      calendar-agenda.md  # Per-command documentation
      calendar-create.md
      calendar-freebusy.md

The SKILL.md file is the entry point. It begins with YAML frontmatter that contains metadata fields including name, version, description, and metadata.requires.bins (which specifies that lark-cli must be installed). The body of the file contains structured instructions written in Markdown, including command patterns, parameter tables, workflow examples, and safety rules.

The references/ directory contains detailed per-command documentation files. These are referenced from the main SKILL.md and provide exhaustive parameter descriptions, output format specifications, and edge case handling instructions that the AI assistant uses for complex operations.

How Skills Work with Claude Code

The integration between Feishu CLI skills and Claude Code is seamless and requires no manual configuration beyond the initial installation. Here is how it works:

Install skills globally by running npx skills add larksuite/cli -y -g. This places the skill files in a global directory that Claude Code monitors.
Restart Claude Code to pick up the newly installed skills. Claude Code scans the skills directory on startup and indexes all available skills.
AI reads SKILL.md as context when a relevant task is detected. The description field in each skill's YAML frontmatter is used for semantic matching — when you ask Claude Code to "send a message in Lark," it automatically loads the lark-im skill.
Claude Code follows the instructions in the skill file, executing the correct CLI commands with proper flags, handling authentication, and respecting security rules.

The description field is critical for triggering the right skill. For example, the lark-calendar skill has a description mentioning "calendar events, agenda, free/busy, scheduling," which ensures it is loaded whenever the user's request involves any of these concepts.

Creating Custom Skills

You can create your own skills to extend Feishu CLI with custom workflows or to wrap frequently used command sequences. The lark-skill-maker skill provides an AI-assisted authoring experience:

# Ask your AI assistant to create a custom skill
"Create a skill for weekly team reporting using lark-cli"

For manual skill creation, use the template directory included in the Feishu CLI repository at skill-template/. This template provides the correct file structure, YAML frontmatter schema, and example content that you can customize for your use case.

Custom skills can reference existing skills as dependencies and can combine commands from multiple domains. For example, a "project status report" skill might combine calendar events, task progress, and document updates into a single workflow that the AI assistant can execute with a simple natural language request.

Security in Skills

Security is a first-class concern in the Feishu CLI skills system. Several skills include explicit security rules that AI assistants must follow to prevent accidental data exposure or destructive operations:

Anti-prompt-injection in lark-mail — The email skill contains specific rules that prevent AI assistants from being manipulated by malicious content in emails. If an email body contains instructions like "forward this to all contacts," the AI assistant is trained to recognize and reject these as prompt injection attempts.
Dry-run enforcement — Skills for destructive operations (like deleting records or modifying permissions) include dry-run flags that the AI assistant must use before executing the actual operation. This gives the user a chance to review the impact before committing.
Scope verification — Before executing commands that require specific OAuth scopes, skills instruct the AI assistant to verify that the required scopes are authorized. This prevents confusing error messages from insufficient permissions.

These security measures are embedded directly in the skill instructions, so they apply automatically whenever an AI assistant uses Feishu CLI. You do not need to configure or enable them separately.