The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform standard git operations (status, diff, log, commit) and to run repository-specific verification commands such as unit tests or linters (e.g., npm test, cargo clippy). These executions are restricted to the changed code surface and are initiated only upon explicit user request.- [DATA_EXFILTRATION]: The skill incorporates a security-first approach to data handling. It includes a dedicated step to scan for and block the staging of secrets, credentials, .env files, and conflict markers, significantly reducing the risk of accidental data exposure or sensitive information leakage.- [SAFE]: The skill implements significant hardening by setting disable-model-invocation to true. This ensures the model cannot autonomously modify git history or execute commands without direct user interaction, providing a high level of control and oversight.

commit