Classifies text as spam, likely spam, suspicious, or clean using signal detection.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.