The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute scripts, a core capability for its validation and testing workflows. In scripts/test_skill.py, it identifies and executes Python scripts from other skill directories to verify their functionality. These calls use list-based arguments and avoid shell=True, reducing common command injection risks.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes content from other skills' SKILL.md files, creating a potential vector for indirect prompt injection.\n
Ingestion points: Files are read in scripts/validate_skill.py, scripts/batch_fix_skills.py, and scripts/update_skill.py.\n
Boundary markers: The skill does not implement delimiters or instructions to ignore embedded agent commands in the files it processes.\n
Capability inventory: The skill can modify local files and execute Python scripts via subprocess.run.\n
Sanitization: The ingested content is parsed with regular expressions and rewritten to the file system without formal sanitization steps.\n- [DYNAMIC_EXECUTION]: scripts/test_skill.py implements a dynamic discovery mechanism to find executable scripts in target skill folders. It scans for Python files containing the if __name__ == \"__main__\" entry point and executes them. This allows the skill to run code discovered at runtime that may not be part of the skill's own codebase.

skill-creation-update