Scans submitted text for exposed secrets and credential-like patterns, then returns a risk score, summary, flags, and recommendations.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.