Synapse Health Interview Question

How would you properly handle LLM hallucinations in agentic workflows?