Agent Guardrails

Define and check agent permissions, human approval and agent escalation for a real workflow. Use when you ask which actions an agent can take, when a person must approve or take over, or whether your agent guardrails match the runtime settings. Works from pasted tools, configs, traces, logs, transcripts and authorized files. Produces an action matrix, a control review and boundary findings tied to available evidence. Covers human in the loop and MCP security where they affect action authority.