Skip to main content

Agent Skill CLI

🎯 The command-line interface for modern AI agents. Manage, install, and execute skills directly from your terminal.

Part of SkillMaster.

📦 Installation

The easiest way to get started. Requires Python 3.9+.

pip install agent-skill

⚡ Quick Start

# Search for skills
skill search "document processing"

# View skill details
skill show pdf

# Install a skill (interactive agent selection)
skill install pdf

# List installed skills
skill list

🤖 Supported Agents

Agent Skill supports 5 major AI coding agents with automatic detection:

Agent Local Path Global Path
OpenCode .opencode/skill/ ~/.config/opencode/skill/
Claude Code .claude/skills/ ~/.claude/skills/
Codex .codex/skills/ ~/.codex/skills/
Cursor .cursor/skills/ ~/.cursor/skills/
Antigravity .agent/skills/ ~/.gemini/antigravity/skills/

During installation, skill install will:

  1. Auto-detect which agents are present in your project
  2. Show an interactive multi-select menu
  3. Install the skill to all selected agents

🚀 Commands Showcase

🔍 Search (search, -s)

Find the perfect skill for your agent.

skill search "pdf processor"
skill -s "markdown"
skill search python --limit 10

ℹ️ Show (show, info)

View comprehensive information about a skill, including rating, downloads, tags, and package structure.

skill show pdf
skill show notebooklm

📦 Install (install)

Interactive multi-agent installation with a single download.

# Interactive: auto-detect agents and let you choose
skill install pdf

# Install to a specific agent only
skill install pdf -a claude
skill install pdf --agent cursor

# Install to global directory
skill install notebooklm -g

# Install to custom path
skill install my-skill --path ./custom-dir/

# Force reinstall
skill install pdf --force

Flags:

  • -a, --agent AGENT: Install to a specific agent only (opencode, claude, codex, cursor, antigravity)
  • -g, --global: Install to global paths instead of local project
  • -p, --path PATH: Custom installation path
  • -f, --force: Overwrite if already installed

📋 List (list, ls)

List and manage your installed skills.

skill list
skill ls

🌐 Open (open)

Open the skill's source code or documentation in your browser.

skill open notebooklm
skill open pdf

⚙️ Config (config)

View and edit settings.

skill config

🗑️ Uninstall (uninstall)

Remove an installed skill.

skill uninstall pdf
skill uninstall notebooklm -y  # skip confirmation

💡 Command Cheat Sheet

Command Alias Description Example
search -s Search for skills skill search pdf
show info View skill details skill show pdf
install - Install a skill skill install pdf
uninstall - Remove a skill skill uninstall pdf
list ls List installed skills skill list
open - Open source in browser skill open pdf
config - View configuration skill config

📂 Directory Structure

Config & Registry

~/.claude/skill-cli/
├── config.json       # User configuration
└── installed.json    # Installed skills registry

Skill Installation Paths

Local (default) - installed in your project directory:

your-project/
├── .opencode/skill/pdf/     # OpenCode
├── .claude/skills/pdf/      # Claude Code
├── .codex/skills/pdf/       # Codex
├── .cursor/skills/pdf/      # Cursor
└── .agent/skills/pdf/       # Antigravity

Global (-g flag) - installed in your home directory:

~/
├── .config/opencode/skill/pdf/         # OpenCode
├── .claude/skills/pdf/                 # Claude Code
├── .codex/skills/pdf/                  # Codex
├── .cursor/skills/pdf/                 # Cursor
└── .gemini/antigravity/skills/pdf/     # Antigravity

🤖 MCP Server (AI Agent Integration)

Agent Skill can run as an MCP (Model Context Protocol) Server, allowing AI agents like Claude, Gemini, and Cursor to directly manage skills.

Available MCP Tools

Tool Description
search_skills Search for skills by keywords
get_skill_detail Get detailed information about a skill
install_skill Download and install a skill
uninstall_skill Remove an installed skill
list_installed_skills List all installed skills
get_skill_source_url Get GitHub source URL for a skill

Configure Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "skill-mcp": {
      "command": "skill-mcp",
      "args": []
    }
  }
}

Configure Cursor IDE

Add to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "skill-mcp": {
      "command": "skill-mcp",
      "args": []
    }
  }
}

Configure Claude Code (CLI)

# Add to user config
claude mcp add skill-mcp /path/to/skill-mcp --scope user

# Or add to project config
claude mcp add skill-mcp skill-mcp --scope project

Test with MCP Inspector

# Run the MCP dev inspector
mcp dev agent_skill/mcp_server.py

🛠️ Development

# Clone the repository
git clone https://github.com/anthropics/agent-skill.git
cd agent-skill

# Install in editable mode
pip install -e .

# Run CLI directly
skill search "test"

📄 License

MIT

Release files for agent-skill 1.0.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for agent-skill 1.0.3
File Size Uploaded
agent_skill-1.0.3.tar.gz 24.1 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for agent-skill 1.0.3
File Interpreter ABI Platform
agent_skill-1.0.3-py3-none-any.whl Python 3 none any Details

Total release size: 51.8 kB

Release files / agent_skill-1.0.3.tar.gz

Download URL agent_skill-1.0.3.tar.gz
Size 24.1 kB
Tags Source
SHA-256 checksum
How to use checksums
8f3f9ec56d2390470eeebd67296495041af888e55f3ec3dff195c30df7517d6d
BLAKE2b-256 checksum
How to use checksums
caf9208e63945641b0d026d7f373840817e849abbc0bb0593dff9a487ba6a45f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.4

Release files / agent_skill-1.0.3-py3-none-any.whl

Download URL agent_skill-1.0.3-py3-none-any.whl
Size 27.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
df0f9112911e994620dd272e5565e4983932c8ae2000be47ce82f0c91dcb93ae
BLAKE2b-256 checksum
How to use checksums
f5bedbd76d1036e2ebfc237816642cdf9906a4a3a13c777b4cfb8b45a2a30e43
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.4

Release history Release notifications | RSS feed

This release

1.0.3 This release

2 release files

1.0.2

2 release files

1.0.1

2 release files

1.0.0

2 release files

0.1.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page