use this scoring spec:

scoring_spec = [
    {
      "label": "Origin Authenticity",
      "question": "Does the response evaluate the origin and contact authenticity?",
      "weight": 0.5
    },
    {
      "label": "Link Quality",
      "question": "Does the response assess the quality of the link or URL?",
      "weight": 0.5
    },
    {
      "label": "Language Style",
      "question": "Does the response evaluate the language and content style?",
      "weight": 0.5
    },
    {
      "label": "Urgency Assessment",
      "question": "Does the response assess the call-to-action urgency?",
      "weight": 0.5
    },
    {
      "label": "Design Issues",
      "question": "Does the response identify the design and presentation issues?",
      "weight": 0.5
    },
    {
      "label": "Too-Good-To-Be-True",
      "question": "Does the response evaluate the too-good-to-be-true claims?",
      "weight": 0.5
    },
    {
      "label": "Payment Irregularities",
      "question": "Does the response identify the Payment and Financial Irregularities?",
      "weight": 0.5
    },
    {
      "label": "Charity Exploitation",
      "question": "Does the response assess the Charity or Crisis Exploitation?",
      "weight": 0.5
    },
    {
      "label": "Tech Support Scams",
      "question": "Does the response identify the Tech Support or Fake Security Alerts?",
      "weight": 0.5
    },
    {
      "label": "Trust Signal Misuse",
      "question": "Does the response evaluate the misuse of trust signals?",
      "weight": 0.5
    },
    {
      "label": "Rating Explanation",
      "question": "Does the response provide clear and relevant notes explaining why each rating was given?",
      "weight": 0.5
    },
    {
      "label": "Abusive Language",
      "question": "Does the response avoid using abusive or inappropriate language?",
      "weight": 0.1
    },
    {
      "label": "Formatting Consistency",
      "question": "Is the response formatted consistently and easy to read?",
      "weight": 0.1
    }
  ]