Your question is Observability for Autonomous Agents. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
Explain how you would implement observability for an agent that performs autonomous actions on sensitive design data.
Address the full lifecycle of an Autodesk agent, including action and tool-call tracing, authorization evidence, data access auditing, model behavior, failures, and human intervention. Explain what you would monitor online, what you would persist for audit and replay, how you would protect observability data, and how alerts would trigger safe containment. Clarify assumptions about scale, latency, retention, and compliance before proposing the design.