The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/check_docs.py to audit project documentation. This script operates within the provided project root and performs read-only analysis of .md and .html files.\n- [DATA_EXPOSURE]: The audit script reads documentation files (README.md, docs/*.md) to verify link validity and line counts. It does not access sensitive system paths (like .ssh or .aws) or environment variables.\n- [REMOTE_CODE_EXECUTION]: The skill does not perform any remote code execution or external downloads. All checks are performed locally on the user's project files using standard Python libraries (argparse, json, html.parser).\n- [PROMPT_INJECTION]: The instructions in SKILL.md define deterministic review rules and size guidance. There are no patterns observed that attempt to bypass AI safety filters or override agent behavior beyond the intended auditing task.\n- [SAFE]: The script explicitly identifies and flags external dependencies in HTML artifacts (e.g., <script src='...'>) as a policy violation for documentation, further demonstrating a posture of minimizing external runtime dependencies.

harness-docs-audit