The Agent Skills Directory

[COMMAND_EXECUTION]: The skill involves the execution of local Python scripts, specifically init_skill.py for scaffolding new skill folders and quick_validate.py for automated structure and content validation. These scripts are bundled with the skill, operate exclusively on local files within the .github/skills directory, and do not perform network operations or access sensitive system files.
[SAFE]: The skill provides instructions for the agent to mine conversation history to generate new skills, which represents an indirect prompt injection surface. However, the process includes mandatory validation gates and human-in-the-loop confirmation steps (Phase 5: Confirm before writing) to ensure generated content is accurate and safe before being written to the filesystem.

skill-creator