Analyzes AI prompts and returns a structured risk score in JSON.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.