The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git, curl, and gh api to search for and clone repositories. It validates the target skill's documentation by executing its included commands ('live checking') and instructs the agent to run scripts for backtesting and scoring.
[EXTERNAL_DOWNLOADS]: Retrieves material from external sources, including GitHub and skill registries, to perform peer analysis. This process includes cloning user-specified repositories and referencing well-known community skills for benchmarking.
[SAFE]: The instructions establish rigorous 'Safety Boundaries' and 'Forced Stop Points' that prevent the agent from performing destructive or public actions without explicit, imperative user authorization. It also prohibits the inclusion of secrets or private data in any polished output.

luban