Your question is Guardrails to Reduce Unsafe Output. 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 design guardrails for a patient-facing AI agent to reduce hallucinations and unsafe recommendations?