Generates red team test cases to assess AI system prompt robustness.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.