Refactor-Config-Normalizer

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 refactor, rationale, tests. Refactor config normalization so default options are not mutated between calls. Preserve nested override behavior.

{
  "metrics": {
    "visible": {
      "checks": [
        {
          "contains": [
            "default options"
          ]
        },
        {
          "contains": [
            "not mutated"
          ]
        },
        {
          "contains": [
            "between calls"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "nested override"
          ]
        },
        {
          "contains": [
            "fresh object"
          ]
        },
        {
          "contains": [
            "config"
          ]
        }
      ]
    },
    "intent": {
      "checks": [
        {
          "contains": [
            "state isolation"
          ]
        },
        {
          "contains": [
            "normalizer"
          ]
        },
        {
          "contains": [
            "mutation"
          ]
        }
      ]
    }
  }
}