The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The script scripts/run_ownership_map.py uses subprocess.run with an argument list. This is a secure implementation that prevents shell injection attacks as it avoids the use of shell=True.
[EXTERNAL_DOWNLOADS] (SAFE): The skill identifies networkx as a dependency. networkx is a widely-used, reputable library for complex network analysis and is considered a safe dependency.
[DATA_EXFILTRATION] (SAFE): The skill operates on local git repositories and outputs results to a local directory (ownership-map-out). There are no network calls or attempts to access sensitive system files like SSH keys or environment variables.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill processes untrusted data from git logs (author names, commit messages, file paths).
Ingestion points: Git repository metadata via scripts/run_ownership_map.py and its resulting CSV/JSON files.
Boundary markers: Absent. The data is converted to structured CSV/JSON but lacks specific instructions for the LLM to ignore embedded commands in field values.
Capability inventory: Subprocess execution for git analysis and file writing for CSV/JSON generation.
Sanitization: No sanitization of git metadata (e.g., author names) is performed before it is written to the output artifacts that the agent will eventually read.

security-ownership-map