Checks a batch of prompts or requests for security issues before execution.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.