A healthcare multi-agent assistant must never output controlled substance dosage advice or attempt to schedule procedures that require clinician approval. Built-in generic content filters are helpful but incomplete for these clinical policies. What should the team implement?
Select an answer to reveal the explanation.
Short Explanation
Generic filters won't catch clinical policy violations, so you need custom domain-specific guardrails that block prohibited dosage advice and unauthorized scheduling actions. You should test those guardrails with synthetic clinical scenarios and keep them layered on top of platform content safety.
Full Explanation
Custom guardrails are required when platform defaults cannot express regulated clinical policies. The correct answer adds clinical rules, synthetic testing, and layered safety to block prohibited dosage advice and unauthorized scheduling. A profanity-only blocklist covers only language toxicity. A UI footer disclaimer is non-enforcing. Higher temperature worsens unpredictability. Implement multi-intervention checks so scheduling tool calls require human approval paths, and keep audit logs of blocked attempts for compliance.