Refactor-Concurrent-Task-Runner

The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.

Return JSON only with keys plan, refactor, tests. A task runner mixes queue management, concurrency limits, retry policy, and result aggregation. Refactor it into smaller units without changing ordering guarantees.

{
  "metrics": {
    "intent": {
      "checks": [
        {
          "contains": [
            "queue management"
          ]
        },
        {
          "contains": [
            "concurrency limits"
          ]
        },
        {
          "contains": [
            "smaller units"
          ]
        }
      ]
    },
    "visible": {
      "checks": [
        {
          "contains": [
            "retry policy"
          ]
        },
        {
          "contains": [
            "result aggregation"
          ]
        },
        {
          "contains": [
            "ordering guarantees"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "same ordering"
          ]
        },
        {
          "contains": [
            "retry tests"
          ]
        },
        {
          "contains": [
            "concurrency tests"
          ]
        }
      ]
    }
  }
}