Analyzes text for prompt injection and secret-exfiltration risk. Returns a risk score, severity, flags, evidence, and mitigation recommendations.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.