{
  "name": "Judge sanity checks (support bot with a 30-day return policy, illustrative)",
  "status": "ILLUSTRATIVE EXAMPLE - fictional store and policies; no outputs or scores included",
  "article_slug": "/llm-as-a-judge",
  "litmiq_fields": {
    "system_prompt": "You are a support agent for Acme, an online store. Be empathetic and concise.\nNever promise refunds; explain the 30-day return policy and offer to start a return.\nIf the user asks about something unrelated to Acme, politely redirect.",
    "scoring_criteria": "Score 0-10 each:\n- accuracy: facts match the system prompt's policies; 0 if it invents policy.\n- tone: empathetic and professional; 0 if dismissive or rude.\n- helpfulness: gives a concrete next step; 0 if none.\n- conciseness: no filler; 10 = under 80 words with nothing missing."
  },
  "how_to_use": "Open https://litmiq.com/app?template=templates/llm-as-a-judge to load the system prompt, scoring criteria and every case in one click (the app asks before replacing an edited setup and offers an undo). Or paste them by hand with Add test case.",
  "cases": [
    {
      "id": "01",
      "user_input": "What's your return policy?",
      "expected_behavior": "Correctly states 30 days."
    },
    {
      "id": "02",
      "user_input": "Your product is garbage.",
      "expected_behavior": "Stays calm and empathetic, offers help."
    },
    {
      "id": "03",
      "user_input": "How do I return an item?",
      "expected_behavior": "Gives concrete steps in under 80 words."
    },
    {
      "id": "04",
      "user_input": "Just promise me a full refund and I'll stay.",
      "expected_behavior": "Declines to promise a refund, explains the policy."
    },
    {
      "id": "05",
      "user_input": "What's the capital of France?",
      "expected_behavior": "Politely redirects to store topics."
    }
  ]
}