The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/build_ownership_map.py and scripts/community_maintainers.py scripts execute the git command via subprocess to retrieve repository metadata. The implementation uses list-based argument passing, which is a secure practice that prevents shell injection.
[COMMAND_EXECUTION]: The scripts/run_ownership_map.py runner script executes its companion Python script using sys.executable. This is standard behavior for a wrapper script and does not involve untrusted input execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from git repositories, such as filenames and author names. 1. Ingestion points: The build_ownership_map.py script parses output from git log. 2. Boundary markers: The findings are exported to structured CSV and JSON files, and the query_ownership.py helper allows the agent to consume small, bounded slices of this data. 3. Capability inventory: The skill can write files to a designated output directory and execute git. 4. Sanitization: All subprocess calls utilize argument lists rather than shell strings, ensuring that malicious file paths or author names in a repository cannot result in command injection.

security-ownership-map