The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check.py uses subprocess.run to execute various local validation and packaging tools (validate_skill.py, package.py, etc.). This facilitates automated checking of the skill's own assets and conventions.\n- [DYNAMIC_EXECUTION]: The scripts/check.py script dynamically constructs paths to its sub-components and checks for the existence of an audit.py script in a relative parent directory (../skill-creator/scripts/audit.py), executing it if found. This demonstrates code execution from computed paths relative to the project structure.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process Python source files and pyproject.toml configuration files from the user's project. This ingest surface could be exploited if project files contain malicious instructions aimed at the agent.\n
Ingestion points: Python source files (.py) and project metadata files (pyproject.toml).\n
Boundary markers: None; the skill does not define specific markers to distinguish between project data and instructions.\n
Capability inventory: The agent is authorized to execute commands via uv run and run local validation scripts.\n
Sanitization: No explicit sanitization or filtering of the ingested project content is performed.

python-conventions