Productivity TOOL
Agent Output Evaluator
Paste an AI agent task, final answer, claimed changes, evidence, acceptance tests, and stop rules to score whether the work is actually done.
Reviewed and refreshed on 2026-06-17.
Browser-side analysis
No signup
Copyable report
Productivity
WHY THIS EXISTS
Built for AI-era search and real users
Paste an AI agent task, final answer, claimed changes, evidence, acceptance tests, and stop rules to score whether the work is actually done. The output stays inspectable: users see the input, classification or score, proof checks, and the limits before they trust the result.
- Closes the loop after AI-agent work instead of trusting a polished final answer.
- Separates claimed work, visible evidence, acceptance tests, and remaining risk.
- Produces a JSON verdict and remediation prompt that can be handed back to an agent.
Boundary: Not for legal audit certification, regulated compliance sign-off, security approval, medical or financial decisions, or verifying files and accounts it cannot inspect.