AI agent decision pre-flight: GO / CAUTION / STOP with cryptographic receipt.

AI agent decision pre-flight + action audit + autonomous safety — should agent X execute action Y? Combined check: evidence freshness, provenance trace, policy gate, risk scoring, hallucination signals. Returns GO / CAUTION / STOP / INSUFFICIENT_EVIDENCE with cryptographic receipt. For autonomous agent workflows: tool calls, financial transactions, content publishing, regulatory submissions, multi-step task execution, agent compliance evidence, MCP gateway routing.