// loading
OpsTicket does not treat a model response as the final score. The evaluator returns bounded evidence bands from terminal commands and output; the server validates those bands and computes the weighted rubric score.
The assessment records terminal commands and terminal output for the submitted attempt.
The evaluator maps that evidence to bounded per-dimension rubric bands. It does not issue the final numeric score.
The server validates the structured bands and computes the weighted score from the versioned rubric.
The canonical result is finalized once. Failed evaluation stays pending instead of fabricating a score.
No two-grader delta workflow, automatic higher-score selection, or next-day human-review promise is claimed.