The Agent Skills Directory

[SAFE]: The skill implements proactive security measures by explicitly instructing the agent to scan for and exclude sensitive files such as .env, *.pem, and any files containing keywords like 'secret', 'token', or 'credentials' from being staged or committed.
[SAFE]: The workflow enforces a principle of least privilege by requiring specific file names for staging instead of broad commands like git add ., which reduces the risk of accidental data exposure.
[SAFE]: The skill includes a manual authorization step before any network operation (pushing to a remote repository), ensuring the user maintains control over external data transmission.
[SAFE]: Commands are executed within a restricted environment where model invocation is disabled, and standard shell practices (e.g., quoted heredocs for commit messages) are used to prevent unintended command injection or expansion.

commit