Agent evaluation workspace

Evaluate agent behavior
before deployment.

Run saved scenarios against a Gemini simulation of an agent prompt. Inspect deterministic tool-policy checks, model-assisted findings, and ordered traces; actions stay in a mock environment.

Open Dashboard →Sign inGitHub
Scenario-based evaluationTool allowlist checksMock-tool executionEvidence-linked tracesModel-assisted judgeOffline deterministic testsEvaluation regression gate