skill-deployer

Install AI coding CLI skills to Claude Code and Codex with one command.

Both platforms use the same skill format (SKILL.md + optional references/ and scripts/ directories) but store them in different locations and register MCP servers differently. This tool bridges that gap.

Install

pip install skill-deployer

Usage

Check detected platforms

skill-deployer status

Install skills

# Install all bundled skills to all detected platforms
skill-deployer install

# Install a specific bundled skill
skill-deployer install my-skill-name

# Install from a GitHub URL
skill-deployer install https://github.com/user/repo/tree/main/skills/my-skill

# Install from a local directory
skill-deployer install ./path/to/skill

# Target a specific platform
skill-deployer install --claude
skill-deployer install --codex

# Overwrite existing skills
skill-deployer install --force

List skills

# List skills installed on detected platforms
skill-deployer list

# List bundled skills available in this package
skill-deployer list --bundled

Uninstall skills

skill-deployer uninstall my-skill-name

Register MCP servers

Register an MCP server on both Claude Code and Codex in one command:

skill-deployer mcp picocalc picocalc-mcp
skill-deployer mcp picocalc python3 /path/to/mcp_server.py

This writes to .mcp.json for Claude Code and runs codex mcp add for Codex.

How it works

skill-deployer detects which platforms are installed, copies skill directories to the right location, and handles MCP registration for each platform's config format.

For skill authors

Bundle your skills in the package by adding them to src/skill_deployer/_skills/ and updating manifest.json:

{
  "skills": [
    {
      "name": "my-skill",
      "version": "1.0.0",
      "description": "What my skill does"
    }
  ]
}

Each skill directory needs a SKILL.md with YAML frontmatter:

---
name: my-skill
description: Trigger description for the AI assistant
---

# Skill content here

Requirements

• Python 3.9+
• Zero runtime dependencies (stdlib only)

License

MIT