The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/usage_census.py script uses subprocess.run to call the git binary. The call is implemented securely using an argument list instead of a shell string, preventing command injection. This execution is scoped to listing tracked files to ensure the analysis respects the project's .gitignore configuration.- [SAFE]: The skill operates entirely on local resources. The included Python script only utilizes standard library modules (argparse, json, re, subprocess, sys, pathlib) and does not attempt to connect to any external servers or download remote content.- [SAFE]: The instructions in SKILL.md promote safe refactoring workflows, including the use of verification gates, architecture constraint checks, and the preservation of public APIs via shims. The documentation encourages a 'no action' verdict when refactoring would increase complexity, aligning with secure and maintainable coding practices.

less-code-same-behavior