The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/run_tests.py script utilizes subprocess.run to execute various testing frameworks (pytest, jest, vitest) based on the detected project environment, facilitating local code execution.
[COMMAND_EXECUTION]: Both scripts/skill_learn.py and scripts/loop_entry.py use dynamic loading techniques, including importlib.util.spec_from_file_location and sys.path modification via Path(__file__).parents[4], to load project-level modules from computed paths.
[PROMPT_INJECTION]: The skill processes external feature descriptions and codebase data, presenting an indirect prompt injection surface.
Ingestion points: Task descriptions, existing repository files, and test results via scripts/run_tests.py.
Boundary markers: Does not implement explicit delimiters or 'ignore' instructions when interpolating external task data into agent prompts.
Capability inventory: Includes command execution via subprocess, file system manipulation (write/delete), and behavior persistence.
Sanitization: Employs yaml.safe_load for internal agent metadata but lacks comprehensive input validation for user-provided task descriptions.

sc-implement