Checks whether an email is likely spam or risky before you send.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.