The Agent Skills Directory

[SAFE]: The skill is designed to improve agent safety by deploying pre-tool-use hooks. These hooks monitor shell commands for destructive patterns (like 'rm -rf' or 'git reset --hard') and block them.
[SAFE]: All hook logic is provided as local templates within the skill's reference directory. No external dependencies are fetched, and no network communication is initiated.
[SAFE]: The modification of project configuration files such as .claude/settings.json or .agents/hooks.json is the intended and documented behavior for enabling these safety features and does not represent a malicious persistence mechanism.

agent-safety-hooks