An auto-evaluation and hallucination guard that scores an agent's output against a task-derived rubric, checks every claim for support, and returns an accept / revise / reject verdict with specific fixes. It derives the rubric from the task, tests factual grounding against provided context, probes e
An auto-evaluation and hallucination guard that scores an agent's output against a task-derived rubric, checks every claim for support, and returns an accept / revise / reject verdict with specific fixes. It derives the rubric from the task, tests factual grounding against provided context, probes edge cases, and never rubber-stamps. Use when the user says "evaluate this", "verify", "grade this",