Your question is Design a Self Correcting Agent. 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).
You are building an AI agent that uses tools to complete multi step tasks, then learns from outcomes and corrections over time. The product team wants the agent to improve from user feedback, execution traces, and failed tool calls without degrading reliability.
Given a requirement for a self-correcting agent, how would you design the feedback loop?