Evaluates text or model outputs against specified values or policies and returns a numeric alignment score with explanatory feedback.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.