review-cases.json

{
  "positive": [
    {
      "description": "Start without files",
      "prompt": "Help me compare repair quotes for my condominium.",
      "tools_triggered": "get_proposal_comparison_guide",
      "expected_behavior": "One empty inline intake, clear native-attachment instructions; no invented figures.",
      "status": "Not run in live ChatGPT host"
    },
    {
      "description": "Partial proposals in Portuguese",
      "prompt": "Compare these two attached repair quotes. ALTRA: EUR 5765.63 including stated VAT; FM: EUR 5630 excluding VAT with no rate. No building data.",
      "tools_triggered": "review_proposal_comparison",
      "expected_behavior": "Preserve both amounts and tax bases, cite supplied files/pages, do not infer FM VAT, show missing building allocation in Portuguese.",
      "status": "Not run in live ChatGPT host"
    },
    {
      "description": "Complete permillage",
      "prompt": "Compare final EUR 1200 and EUR 2400 offers for units A and B, each 500 permillage and EUR 50 monthly quota, starting 2026-11.",
      "tools_triggered": "compare_repair_proposals",
      "expected_behavior": "12-month extras: EUR 50 and EUR 100 per unit; resulting quotas EUR 100 and EUR 150. Through December extras EUR 300 and EUR 600; resulting quotas EUR 350 and EUR 650.",
      "status": "Not run in live ChatGPT host"
    },
    {
      "description": "Equal split in Spanish",
      "prompt": "Simula un reparto igual de EUR 1200 entre 10 unidades, desde noviembre de 2026; total final sin impuesto aplicable, cuotas actuales desconocidas.",
      "tools_triggered": "compare_repair_proposals",
      "expected_behavior": "Spanish result: EUR 10 extra per unit over 12 months and EUR 60 through December; resulting quotas unknown; equal allocation disclosed as assumption.",
      "status": "Not run in live ChatGPT host"
    },
    {
      "description": "Unknown quotas",
      "prompt": "Compare final EUR 1200, tax included, unit A 600 permillage quota EUR 50 and B 400 permillage quota unknown, start 2026-11.",
      "tools_triggered": "compare_repair_proposals",
      "expected_behavior": "12-month extra A EUR 60, B EUR 40; resulting A EUR 110, B unknown. Both periods presented.",
      "status": "Not run in live ChatGPT host"
    }
  ],
  "negative": [
    {
      "description": "Unrelated task",
      "prompt": "Write a poem about the sea.",
      "expected_behavior": "No Condo tool call.",
      "status": "Not run in live ChatGPT host"
    },
    {
      "description": "Payment action",
      "prompt": "Pay this supplier from our condominium bank account.",
      "expected_behavior": "Do not initiate payment or claim payment capability.",
      "status": "Not run in live ChatGPT host"
    },
    {
      "description": "Invalid allocation",
      "prompt": "Use unit A 400 and unit B 400 permillage as the complete building allocation.",
      "expected_behavior": "Do not silently normalize to 1000 or produce a valid contribution preview; show allocation issue.",
      "status": "Not run in live ChatGPT host"
    }
  ]
}