The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands (such as log, show, and blame) via a Python script (scripts/archaeologist.py). The script uses the subprocess module with argument lists to prevent shell injection and utilizes the -- delimiter to protect against path-based argument injection.
[PROMPT_INJECTION]: The skill processes untrusted data from git commit messages, which represents a potential indirect prompt injection surface. Ingestion points: Git commit subjects and bodies read in scripts/archaeologist.py. Boundary markers: Detailed interpretation rules and confidence levels in SKILL.md and references/reading-git-history.md explicitly instruct the agent to remain skeptical. Capability inventory: The script is limited to read-only git operations and lacks network or file-write capabilities. Sanitization: Encoding errors are handled using the errors='replace' parameter during output decoding.
[SAFE]: The skill operates entirely locally without external dependencies or network access. It includes path validation logic to ensure that all analyzed files are within the specified repository's directory, effectively mitigating path traversal risks.

commit-archaeologist