{
  "unit": "bug",
  "aggregation": "median",
  "headline": {
    "metric": "count",
    "higher_is_better": true,
    "unit": "seeded bugs caught"
  },
  "secondary": [
    {
      "id": "clean_pass",
      "metric": "mean",
      "higher_is_better": true,
      "unit": "fraction",
      "aggregation": "median"
    },
    {
      "id": "line_coverage",
      "metric": "mean",
      "higher_is_better": true,
      "unit": "fraction",
      "aggregation": "median"
    },
    {
      "id": "branch_coverage",
      "metric": "mean",
      "higher_is_better": true,
      "unit": "fraction",
      "aggregation": "median"
    },
    {
      "id": "mutation_score",
      "metric": "mean",
      "higher_is_better": true,
      "unit": "fraction",
      "aggregation": "median"
    }
  ],
  "seed": 17,
  "verdict_rule": {
    "bootstrap": "paired-items",
    "resamples": 10000,
    "confidence": 0.95,
    "prng": "mulberry32",
    "quantile": "type-7",
    "range": "absolute-difference-exceeds-wider-run-range",
    "label_tie": "json-lexicographic",
    "zero_division": 0,
    "degenerate_kappa": 1,
    "rounding": "half-away-from-zero",
    "decimals": 4
  },
  "record_shape": [
    "score"
  ],
  "max_tokens": 4096,
  "reasoning_effort": "medium",
  "tool_choice": {
    "type": "function",
    "function": {
      "name": "write_tests"
    }
  },
  "fixture_sha256": "ed78fc258cb79077a4ec3cd017b6506061dedf7f0d0bc018360cd565a723899d",
  "recipe_sha256": "3e832df4367347f054134428cfaa9ebad9709dce8cf513c0dd857ea6343de265",
  "checker_sha256": "31204ecc127bea13cc834b137542bd4e546b4e1609a1edadfb41a255e8a416f8",
  "runner_sha256": "493f1a269124d24910b79935c3f14353be3edf24e98a4dc825e3b6f891e03ca6",
  "sandbox": "macos",
  "fixture_source": {
    "source": "Original MIT pure-Python modules written for this page",
    "license": "MIT",
    "generators": ["Pareto coding agent"]
  },
  "planning": {
    "score_bounds": [0, 1]
  },
  "tuning_repeats": 1,
  "repeats": 3,
  "prompt": "Write a compact Python unittest file for the clean subject.py module. Import unittest and subject only. Use unittest.TestCase methods and explicit assertions for documented normal, empty, boundary and invalid inputs. Check expected values carefully against the contract before writing them. Do not test unspecified behavior. No file access, introspection, eval, exec, subprocesses, mocks, skipped tests or expected failures. Do not call unittest.main or use a __name__ guard; our loader runs the suite. Return the entire test file in exactly one write_tests tool call. No existing tests may be deleted or repaired."
}