Publish evaluator artifacts
Version WASM evaluators once, reference them from profiles, and keep scoring logic stable across local runs, CI, and production gates.
What you get
Evaluation infrastructure, not another score script.
Run distributed evaluations
Coordinators dispatch durable run chunks while workers call the target agent, execute evaluators, and persist normalized results.
Gate releases with evidence
Watch pass/fail outcomes, inspect summaries, and export execution evidence for release decisions and debugging.