skillx is a tool for AI agents that lets you try out skills directly from any GitHub repository without formal installation. It is specifically designed to keep your AI environment clean while you experiment with new tools.

• Clutter-Free Testing: Run any GitHub-based skill without adding files to your AI agent's main skill directory.
• Dynamic Usage: Instantly use tools from owner/repo or full URLs.
• Safe Cache: Uses an isolated, deterministic cache that is easy to manage and clean up.

• bash (4.0 or later)
• git
• An AI agent capable of running custom skills or scripts (e.g., Claude Code, Copilot CLI)

Run the following command within your AI agent:

/skillx <REPO> [OPTIONS...]

• <REPO>: The GitHub repository containing the skill. Supports several formats:
  • owner/repo
  • https://github.com/owner/repo
  • https://github.com/owner/repo/tree/branch/sub/path (for skills in a specific subdirectory or branch)
• [OPTIONS...]: Arguments to pass to the downloaded skill.

Examples:

/skillx subasaogawa/session-stocker-skill
/skillx https://github.com/anthropics/skills/tree/main/skills/skill-creator

1. Normalization: It converts the provided <REPO> into a standard owner/repo format.
2. Setup:
   • It identifies a cache location: ~/.local/share/skillx/skills/<owner>/<repo>.
   • If the skill isn't cached, it clones it into a temporary directory and moves it to the cache.
   • If already cached, it verifies the repository integrity and updates it via git pull --ff-only.
3. Execution: It reads SKILL.md from the cached repository and provides the instructions to the AI agent.
4. Cleanup: After the task is complete, the agent can invoke the cleanup command to remove the specific skill cache.