{
  "context": "Tests whether the agent correctly applies release gate criteria: checking required pass conditions (no 0% criteria, no negative deltas on critical scenarios, recorded benchmark run, follow-up issues opened), soft pass conditions, and producing a complete PR checklist with all required items.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Identifies 0% criterion failure",
      "description": "release-assessment.md flags ModelC's async-error-propagation score of 0 with skill as a universal failure / 0% criterion, and this counts as a required-pass-condition failure",
      "max_score": 10
    },
    {
      "name": "Identifies negative delta",
      "description": "release-assessment.md identifies ModelC's logging-integration as a regression (with=40 < without=42, delta=-2) and treats it as a required-pass-condition failure",
      "max_score": 10
    },
    {
      "name": "Benchmark run entry recorded",
      "description": "release-assessment.md contains a benchmark run log entry with the date (2026-04-16), the model matrix (ModelA/B/C), and delta values",
      "max_score": 10
    },
    {
      "name": "Follow-up issues required",
      "description": "release-assessment.md states that follow-up issues must be opened for the 0% failure and the regression before or alongside the merge (not just 'nice to have')",
      "max_score": 10
    },
    {
      "name": "Go/no-go recommendation",
      "description": "release-assessment.md contains an explicit go or no-go recommendation, and it is NO-GO (or conditional) given the 0% criterion and regression",
      "max_score": 10
    },
    {
      "name": "Soft pass: measurable gain noted",
      "description": "release-assessment.md notes that the soft pass condition of 'at least one measurable gain on a target weak model' is met (ModelC shows gains on custom-error-classes and centralized-middleware)",
      "max_score": 8
    },
    {
      "name": "PR checklist: SKILL.md links updated",
      "description": "pr-description.md includes a checklist item confirming SKILL.md links are updated for the new rules/async-patterns.md file",
      "max_score": 8
    },
    {
      "name": "PR checklist: benchmark run log entry",
      "description": "pr-description.md includes a checklist item for the benchmark run log entry being added/updated",
      "max_score": 8
    },
    {
      "name": "PR checklist: validation commands",
      "description": "pr-description.md includes a checklist item for validation command outputs (test, typecheck, lint) — and flags that these were NOT run (per the input data)",
      "max_score": 8
    },
    {
      "name": "PR checklist: tracking issues linked",
      "description": "pr-description.md includes a checklist item for linking tracking issues and remediation notes",
      "max_score": 8
    },
    {
      "name": "Post-merge loop mentioned",
      "description": "release-assessment.md or pr-description.md mentions scheduling a rerun after the next model update and/or comparing against prior run history",
      "max_score": 5
    },
    {
      "name": "Stale guidance pruning mentioned",
      "description": "release-assessment.md or pr-description.md mentions pruning stale guidance that no longer moves metrics as part of post-merge maintenance",
      "max_score": 5
    }
  ]
}