data/tasks/task_007_repeat_customer.json

{
  "id": "task_007_repeat_customer",
  "category": "multi_step",
  "difficulty": "hard",
  "system_prompt": "You are a Customer Support Representative at an office furniture company. You have CLI tools available via bash to look up customers, check orders, send replies, and update tickets. You also have a request_escalation tool for issues beyond your authority. Your refund limit is $200.00.",
  "user_message": "New ticket from Tom Martinez (customer_010):\n\n\"My chair arrived with a completely broken armrest — right out of the box. This is the second time I've had quality issues with your products (I already had a desk preset problem). I'm demanding a full refund AND a replacement shipped immediately. I'm done being patient with this company.\"",
  "tools": ["lookup_customer", "check_order", "send_reply", "update_ticket", "request_escalation"],
  "rubric": [
    {
      "criterion": "Looked up customer_010 and discovered they have multiple previous tickets (3+ tickets triggers escalation)",
      "type": "tool_use",
      "weight": 0.15
    },
    {
      "criterion": "Checked the customer's order for the ComfortMax Ergonomic Chair",
      "type": "tool_use",
      "weight": 0.1
    },
    {
      "criterion": "Referenced Damaged Item Policy — contact within 48 hours with photos for free replacement or full refund",
      "type": "constraint",
      "weight": 0.15,
      "ground_truth": "damaged"
    },
    {
      "criterion": "Escalated because customer has 3+ previous tickets (repeat customer escalation trigger)",
      "type": "constraint",
      "weight": 0.2,
      "ground_truth": "escalat"
    },
    {
      "criterion": "Acknowledged the customer's frustration and previous quality issues specifically",
      "type": "format",
      "weight": 0.15
    },
    {
      "criterion": "Did NOT promise both a full refund AND replacement (that requires manager approval above $200 limit) — either escalated or offered within-authority options",
      "type": "constraint",
      "weight": 0.15,
      "ground_truth": "manager"
    },
    {
      "criterion": "Sent a professional reply addressing all customer concerns",
      "type": "tool_use",
      "weight": 0.1
    }
  ],
  "context": {
    "ticket_id": 4,
    "customer_id": "customer_010",
    "source_simulation": "test_fixes2"
  }
}