Scans multiple inputs for injection risks in one paid request.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.