Your question is Robust Error Handling for Tools. 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).
Given a specific agentic task, how would you design a robust error-handling mechanism for failed tool calls?
Focus on how the agent should recover, retry, escalate, or stop safely when a tool fails. Keep your answer grounded in practical agent behavior, not generic backend retries.