The Agent Skills Directory

[SAFE]: The skill implements security-conscious constraints, specifically instructing the agent to never commit files that likely contain secrets, such as environment files (.env), keys, or tokens, and to provide a warning to the user instead.
[SAFE]: It maintains a strict human-in-the-loop requirement, explicitly forbidding the agent from performing git commit or push operations unless the user provides a direct request in that specific turn.
[SAFE]: The skill uses defensive shell scripting practices, such as suggesting HEREDOC format for git commit messages, which mitigates the risk of command injection if the generated commit message contains shell-sensitive characters.
[SAFE]: No external network requests, unauthorized file system access outside of the repository, or third-party dependencies were detected.

prepare-commits