Cleanup

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

Refactor without changing behavior. Return only Python code.\n\ndef normalize_email(email):\n    if email is None:\n        return None\n    email = email.strip()\n    if email == '':\n        return None\n    return email.lower()\n

{
  "expected_contains": [
    "email.strip()",
    "return email.lower()"
  ]
}