Implements security hooks and permission controls to govern AI agent behavior and protect sensitive data.
The Agent Governance skill provides a robust framework for managing the security and operational boundaries of custom Claude agents. By utilizing PreToolUse and PostToolUse hooks, developers can intercept tool calls to block unauthorized file access, prevent dangerous CLI commands, and maintain detailed audit trails. This skill is essential for production-grade agent deployments where security compliance, rate limiting, and behavioral oversight are critical requirements for maintaining system integrity.
主な機能
01Hook-based permission management via PreToolUse and PostToolUse
02Comprehensive audit logging for tool execution and session history
03Dangerous command validation for Bash and terminal-based tools
041 GitHub stars
05Sensitive file access blocking for credentials and configuration files
06Customizable rate limiting and output content filtering
ユースケース
01Securing enterprise agents by preventing access to secrets, .env files, and private keys.
02Implementing regulatory audit trails for compliance in automated coding workflows.
03Creating safe execution sandboxes for agents performing CLI-based system operations.
