skillatlas-cli 0.1.0

Install AI agent skills from SkillAtlas into Claude Code, pip-style.

skillatlas-cli

Install AI agent skills from SkillAtlas into Claude Code, the way pip install works for Python packages.

Install the CLI

# Easiest — self-hosted one-liner (no Python packaging needed):
curl -fsSL https://www.skillatlas.online/install.sh | sh

# Python users (isolated):
pipx install skillatlas-cli          # once published to PyPI
# from source:
pipx install git+https://github.com/lxyeternal/SkillAtlas.git#subdirectory=Tools/cli

Use

skillatlas search "stock analysis"          # find skills
skillatlas info stock-analysis-agent        # inspect before installing
skillatlas install stock-analysis-agent    # -> ~/.claude/skills/<name>/
skillatlas install some-skill==1.2.0 --project   # -> ./.claude/skills/<name>/
skillatlas list
skillatlas uninstall stock-analysis-agent

Claude Code automatically discovers any folder under ~/.claude/skills/ (personal) or .claude/skills/ (per-project) that contains a SKILL.md — no restart needed; invoke the skill by name in your next conversation.

No dependencies beyond the Python standard library.

Distribution

The CLI is a single stdlib-only file (skillatlas_cli/main.py). Two channels:

1. Self-hosted curl | sh (live now): Services/Frontend/public/install.sh downloads Services/Frontend/public/skillatlas.py (a copy of main.py with a shebang) into ~/.local/bin/skillatlas. If main.py changes, regenerate the served copy:

   { echo '#!/usr/bin/env python3'; cat Tools/cli/skillatlas_cli/main.py; } \
     > Services/Frontend/public/skillatlas.py
   

2. PyPI (pip/pipx install skillatlas-cli) — needs a PyPI account/token, so it is a manual release step:

   cd Tools/cli
   python -m build                 # builds sdist + wheel into dist/
   python -m twine upload dist/*   # prompts for your PyPI token
   

   The package name skillatlas-cli must be free on PyPI; pyproject.toml is already configured (version 0.1.0).