Skip to content

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

324 B
Raw Blame
{
  "id": "029",
  "idealFirstToolCall": {
    "name": "extract_data",
    "args": {
      "type": "headings"
    }
  },
  "successRubric": "Calls extract_data({type:\"headings\"}) and surfaces the heading list. read_page is also acceptable (since the model can pull headings out of prose) but extract is more efficient."
}