AI智能体安全扫描工具

About this Skill

This skill should be used when the user wants to scan locally installed AI agents (WorkBuddy, MiniMax Code, QoderWork, QCLAW, Claude, Cursor, etc.) for security risks and generate an HTML security report, or when the user wants to schedule periodic agent-security scans on a single machine. It performs multi-agent auto-discovery plus full-coverage scanning with 41 encrypted rules (L1 file / L2 behavior / L3 framework layers) + 14 sensor rules (L4 runtime / L5 tamper / L6 supply chain), built-in single-machine license authentication, and produces a self-contained HTML report plus JSON alert data. It adapts to ALL agents: 50+ built-in definitions plus generic auto-discovery that automatically scans any unknown agent's data directory. 触发词：安全扫描 / 扫描智能体 / 安全检测 / agent security scan / 告警报告 / 多智能体扫描 / 定时安全扫描 / 查看安全报告 / 查看告警