Deterministic content security scan for prompt injection and data exfiltration patterns. Returns a verdict, risk score, and matched findings for the supplied content.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.