Refactor-Report-Export-Workflow

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. Report export code mixes data loading, permission checks, file rendering, and upload cleanup. Refactor into an orchestrated workflow with clear failure handling.

{
  "metrics": {
    "intent": {
      "checks": [
        {
          "contains": [
            "orchestrated workflow"
          ]
        },
        {
          "contains": [
            "clear failure handling"
          ]
        },
        {
          "contains": [
            "report export"
          ]
        }
      ]
    },
    "visible": {
      "checks": [
        {
          "contains": [
            "data loading"
          ]
        },
        {
          "contains": [
            "permission checks"
          ]
        },
        {
          "contains": [
            "upload cleanup"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "same files"
          ]
        },
        {
          "contains": [
            "cleanup on failure"
          ]
        },
        {
          "contains": [
            "tests"
          ]
        }
      ]
    }
  }
}