The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python script to execute local git commands for repository metadata and history analysis.
Evidence: Found in the run_git function within scripts/audit-docs.py which calls subprocess.run(["git", *arguments], ...).
Context: Command execution is limited to read-only git operations (rev-parse, show, ls-files) and uses list-based argument passing to prevent shell injection. All target paths are validated to ensure they remain within the repository root.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the target project repository, creating a potential surface for indirect prompt injection via analyzed documents.
Ingestion points: The scripts/audit-docs.py script reads Markdown files, repository mapping configurations (.governance/docs-map.json), and Git history logs.
Boundary markers: The audit script reads and processes file content directly for integrity checks (link verification, role mapping) without using explicit delimiters or instructions to ignore embedded agent directives.
Capability inventory: The skill possesses file system read access and the ability to execute git subprocesses.
Sanitization: The script implements path resolution checks using the resolved_within_root helper to prevent directory traversal attacks during file ingestion.

docs-governance