AgentGuard — 20-tool AI safety MCP: policy preflight, risk scoring, audit logging, rate limits.
Probing this server's capabilities…