skillz

Agent skills package manager — install, manage, and discover AI agent skills from GitHub.

Install

# Zero-install via uvx (like npx for Python):
uvx skillz add kayaman/skills@semver

# Or install globally:
uv tool install skillz
skillz add kayaman/skills@semver

Commands

skillz add <source>

Install a skill from GitHub or a local path.

# Install a specific skill from a repo
skillz add kayaman/skills@semver

# Install all skills from a repo
skillz add kayaman/skills

# Install globally (to ~/.agents/skills/)
skillz add kayaman/skills@semver -g

# Install from a local directory
skillz add ./my-local-skill

# Skip confirmation
skillz add kayaman/skills@semver -y

skillz list

Show installed skills (project + global).

skillz list            # Table output
skillz list --json     # JSON output
skillz list -g         # Global only
skillz ls              # Alias

skillz remove <name>

Uninstall a skill and clean the lockfile.

skillz remove semver
skillz rm semver -g    # Remove from global

skillz update [name]

Check for upstream changes and re-fetch updated skills.

skillz update          # Update all skills
skillz update semver   # Update a specific skill

skillz init [name]

Scaffold a new SKILL.md template.

skillz init my-skill
skillz init my-skill --dir ./skills

skillz find <query>

Search GitHub for repos containing agent skills.

skillz find "react testing"
skillz find deploy --limit 5

Skill format

A skill is a directory containing a SKILL.md file with YAML frontmatter:

---
name: my-skill
description: What this skill does and when to use it.
---

# My Skill

Instructions for the AI agent...

Skills are typically organized in a GitHub repo under a skills/ directory:

my-repo/
└── skills/
    ├── skill-a/
    │   └── SKILL.md
    └── skill-b/
        ├── SKILL.md
        └── references/
            └── extra-context.md

Lockfile

Installed skills are tracked in .skill-lock.json (v3 format, compatible with vercel-labs/skills). Skills installed by either tool are visible to the other.

Supported agents

skillz detects and works with: Cursor, Claude Code, Codex, GitHub Copilot, Amp, Cline, Continue, and Kiro.

Development

uv sync                  # Install deps
uv run pytest -v         # Run tests
uv run ruff check src/   # Lint
uv run ruff format src/  # Format
uv run skillz --help     # Run locally

License

Apache-2.0