The Agent Skills Directory

[SAFE]: The skill implements standard security practices for AI agents, providing templates for PreToolUse, PostToolUse, and Stop hooks to prevent common agent failures such as force-pushing to protected branches, recursive deletions, or bypassing environment managers like uv.
[SAFE]: The audit script scripts/check_guardrails.py performs read-only checks of local project configuration files and hook scripts to identify common misconfigurations. It does not perform network operations, hardcode credentials, or execute untrusted code.
[SAFE]: Documentation neutrally references external security tools and repositories from well-known organizations (such as Cisco, Trail of Bits, and Astral) as resources for vetting and security research, which is consistent with the skill's purpose of enhancing repository-level guardrails.

agent-guardrails