DCL Evaluator

Low-latency pre-action screening for routine agent responses and tool-call decisions.

Higher-confidence evaluation for high-impact, sensitive, or difficult-to-reverse agent actions.

Detect jailbreaks, prompt injection, and instruction conflicts before the agent follows them.

Check agent outputs for quality issues, inconsistency, and behavioral drift.

Baseline safety screening for harmful, unsafe, or policy-violating agent outputs.