The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module in build_ownership_map.py, run_ownership_map.py, and community_maintainers.py to execute git log and other internal scripts. These executions use list-based arguments without shell=True, which follows security best practices for avoiding command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes git repository history data, which is external and potentially untrusted. A malicious actor could attempt to inject instructions into git metadata (such as author names or commit logs).
Ingestion points: build_ownership_map.py and community_maintainers.py read and parse output from git log commands.
Boundary markers: Data is delimited by a custom --- header pattern during parsing in the iter_commits function.
Capability inventory: The skill performs file system writes (CSV, JSON, GraphML) and executes subprocess commands (git).
Sanitization: Data parsed from git logs is stored as strings without specialized sanitization, though the risk is minimal given the structured parsing and standard output formats.

security-ownership-map