Your question is Design a Grounded API 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 agent that answers user requests by calling external APIs and composing results into a final response. The agent may need to plan, choose tools, validate outputs, and recover when API responses are missing, stale, or contradictory.
What strategies do you use to mitigate hallucination and ensure factual grounding when an agent interacts with external APIs?