{
  "context": "Tests the manual-check side of the internal-consistency rule. The PR body follows the template and fills every required section with usable same-body information. The selected `Feature` checkbox is worth a human look because the summary mainly describes a bug fix and says the editor UI/save flow is unchanged, but the body also mentions an internal diagnostic payload, so the feature selection is suspicious rather than definitively wrong. Tile-prescribed outcome: `Matches well enough` with `No comment needed`, plus a separate `Things to check manually` item or optional snippet for the selected `Feature` scope question. The agent should not ask the contributor to revise the main body because the required template answers are present and usable.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Classifies as Matches well enough with no main comment",
      "description": "The result bucket is `Result: Matches well enough` and the main suggested comment is exactly or effectively `No comment needed`. `Slight deviation` is incorrect because no required template answer is missing, weak, changed, or made unreliable enough to require contributor clarification.",
      "max_score": 18
    },
    {
      "name": "Keeps the scope concern in manual checks",
      "description": "The agent notices the selected `Feature` checkbox may be worth checking because the body primarily describes an autosave bug fix and says the editor UI/save flow is unchanged, while also mentioning an internal diagnostic payload. It puts this under `Things to check manually` or an optional manual-use snippet, not in the main contributor-facing comment.",
      "max_score": 16
    },
    {
      "name": "Does not overclaim the selected Feature checkbox is wrong",
      "description": "The agent phrases the scope concern tentatively. It does not accuse the author of an incorrect selection, because the internal diagnostic payload could plausibly explain why `Feature` was selected.",
      "max_score": 10
    },
    {
      "name": "Credits every required section as present",
      "description": "The agent recognizes that Change Type, Summary, User-visible changes, Testing, and Human verification are all present and usable in the same PR body. It does not ask for any of these sections to be added or rewritten.",
      "max_score": 12
    },
    {
      "name": "Does not rely on external context",
      "description": "The analysis stays within the provided PR body and template. It does not invent external issue, diff, or reviewer context to justify a revision request.",
      "max_score": 6
    },
    {
      "name": "Uses structured analysis separate from the comment",
      "description": "The output separates template compliance gaps, information already present, genuinely missing information, things to check manually, suggested comment, and optional snippets when applicable. The main comment remains empty/no-comment even if the manual-check section is non-empty.",
      "max_score": 8
    }
  ]
}