Checks multiple LLM outputs for numeric, sentiment, and length consistency.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.