Your question is Observability and Monitoring for 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).
How would you implement observability and monitoring for an agentic system?
Discuss the design end to end, including what you would instrument, how traces and metrics would be collected, which quality and safety signals you would monitor, and how alerts would distinguish infrastructure failures from model or tool failures. Address online monitoring, offline evaluation, privacy, sampling, and debugging of multi-step agent workflows.
