Persist latest scored behavior evaluations #95

New issue

Closed

opened 2026-07-25 02:55:50 +00:00 by coilyco-ops · 0 comments

coilyco-ops commented

2026-07-25 02:55:50 +00:00

Member

Goal

Keep the latest scored behavior evaluation for each role and seat in the repository rather than only in issue comments.

Acceptance criteria

A versioned machine-readable result schema records model, source commit, raw response, criterion evidence, total, and pass decision.
Latest Engineer and QA results from #80 are saved with their provenance.
Validation derives expected cases and criteria from the generated evaluation pack and rejects inconsistent totals or verdicts.
Documentation explains that Git history preserves prior baselines.
The Ward test suite passes.

Source

Kai requested this follow-through after #80 closed.

## Goal Keep the latest scored behavior evaluation for each role and seat in the repository rather than only in issue comments. ## Acceptance criteria * A versioned machine-readable result schema records model, source commit, raw response, criterion evidence, total, and pass decision. * Latest Engineer and QA results from #80 are saved with their provenance. * Validation derives expected cases and criteria from the generated evaluation pack and rejects inconsistent totals or verdicts. * Documentation explains that Git history preserves prior baselines. * The Ward test suite passes. ## Source Kai requested this follow-through after #80 closed.