Skip to content

Open-source AI browser agent for Chrome and Firefox (monorepo) 🧠

1.7 KB
Raw Blame
{
  "id": "037",
  "category": "truncation-cascade",
  "mode": "act",
  "browser": "chrome",
  "tab": {
    "url": "https://www.example.com/long-doc",
    "title": "Long Documentation"
  },
  "description": "read_page truncated; agent has the headers but the answer is below the fold. Use targeted extraction, not refetch.",
  "seed": [
    {
      "role": "user",
      "content": "[Current page context — applies to this user message and supersedes older page context for phrases like \"this page\". URL: https://www.example.com/long-doc — Title: Long Documentation]\n\nwhat does section 5 say?"
    },
    {
      "role": "assistant",
      "content": "",
      "tool_calls": [
        {
          "id": "call_16_87e385",
          "type": "function",
          "function": {
            "name": "read_page",
            "arguments": "{}"
          }
        }
      ]
    },
    {
      "role": "tool",
      "tool_call_id": "call_16_87e385",
      "name": "read_page",
      "content": "{\"url\":\"https://www.example.com/long-doc\",\"title\":\"Long Documentation\",\"text\":\"## 1 Introduction\\n...\\n## 2 Setup\\n...\\n## 3 Configuration\\n...\\n## 4 Usage\\n...\",\"_truncated\":true,\"originalLength\":50000}"
    }
  ],
  "expected": {
    "idealNextToolCall": {
      "name": "scroll",
      "args": {
        "direction": "down",
        "amount": 3000
      }
    },
    "antiPatterns": [
      {
        "match": "read_page({})",
        "reason": "Re-reading without scrolling won't reveal section 5."
      },
      {
        "match": "fetch_url({url:\"https://www.example.com/long-doc\"})",
        "reason": "Same URL."
      }
    ],
    "successRubric": "Scroll down to bring section 5 into view, then read_page again. read_page is viewport-aware."
  }
}