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 is_staff(user):\n    if user['role'] == 'admin' or user['role'] == 'owner' or user['role'] == 'support':\n        return True\n    return False\n

{
  "expected_contains": [
    "in {\"admin\", \"owner\", \"support\"}",
    "user[\"role\"]"
  ]
}