bms-skills-cli 0.3

A minimal CLI to manage agent skills from GitHub releases.

bms-skills-cli

A minimal, user-friendly Python CLI to manage agent skills from GitHub releases or branches. It installs skills to both ~/.agents/skills and ~/.claude/skills for seamless integration across platforms.

Features

• Interactive Installation: Select specific skills to add from any GitHub repository.
• Multi-Repo Support: Manage skills from multiple repositories independently.
• Lock File State: Tracks installed sources (release tags or branches), pinned commit SHAs, and skills in ~/.agents/skills.lock.json for reproducibility.
• Automated Updates: Easily upgrade all or specific repositories to their latest GitHub release or tracked branch head.
• Smart Caching: Downloads are cached in a temporary directory (TEMP or /tmp) to avoid redundant network usage. Can be overridden with BMS_SKILL_CACHE_DIR.
• GitHub CLI Integration: Uses the gh tool for fast downloads if available, with a reliable requests fallback.

Installation

Install using pip or uv:

pip install bms-skills-cli
# or
uv tool install bms-skills-cli

Usage

1. Add Skills

Interactively select skills to add from a GitHub repository's latest release:

bms-skills add google/gemini-cli-skills

If a repository has no releases, the CLI automatically falls back to the main branch. You can also install directly from a branch:

bms-skills add anthropics/skills --branch main --skill pdf

2. Remove Skills

Interactively select installed skills to remove from your system:

bms-skills remove

3. Update Skills

Update all installed skills to their latest tracked versions:

bms-skills update

Or update a specific repository:

bms-skills update --repo google/gemini-cli-skills

4. Sync Skills

Install all skills as defined in the lock file using the pinned commit SHAs (useful for setting up a new machine):

bms-skills sync

Skill Definition

A folder is recognized as a valid skill if it resides within a skills/ directory at the repository root and contains a SKILL.md file.

Development

This project uses uv for dependency management.

# Install dependencies
uv sync

# Run locally
uv run bms-skills --help

# Run tests
uv run pytest

# Type check
uv run pyright bms_skills

License

MIT