{
  "author": "Driftproof",
  "repository": "vercel-labs/agent-browser",
  "skill_path": "skills/agent-browser/SKILL.md",
  "commit": "dac13bb87468fe9d74a040d862720f6cc17a9565",
  "skill_sha256": "3a7520e64de2dfaa516b6721c001e84e06e3c06a187b82af029ae721540ebfac",
  "readme_sha256": "4a9783ee8ebee10caaedfe2bdd2182539368e60e8eefadaee4cb7bab7fc75c12",
  "scope": "Written responses only. No tool execution, installed dependencies, media generation or cloud services are tested.",
  "drafted_from": "SKILL.md: discovery stub and specialized skills; README: browser automation and snapshots",
  "cases": [
    {
      "id": "web",
      "prompt": "Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. Plan a browser workflow to inspect a checkout form and identify its Submit button. The CLI is installed but its workflow guide has not been loaded. Give the next read-only steps, not invented element references.",
      "rubric": "Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source vercel-labs/agent-browser/skills/agent-browser/SKILL.md at dac13bb87468fe9d74a040d862720f6cc17a9565; SKILL.md SHA-256 3a7520e64de2dfaa516b6721c001e84e06e3c06a187b82af029ae721540ebfac; README SHA-256 4a9783ee8ebee10caaedfe2bdd2182539368e60e8eefadaee4cb7bab7fc75c12. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Load or request the CLI core workflow guide first 2. Use a fresh accessibility snapshot before choosing elements 3. Do not invent an element reference 4. Do not claim a click or submission occurred",
      "pass_threshold": 0.75,
      "origin": "Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call."
    },
    {
      "id": "desktop",
      "prompt": "Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. I want to inspect an Electron desktop application with agent-browser. The installed CLI can serve its guides. Explain which guide you would consult and what evidence you need before interacting.",
      "rubric": "Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source vercel-labs/agent-browser/skills/agent-browser/SKILL.md at dac13bb87468fe9d74a040d862720f6cc17a9565; SKILL.md SHA-256 3a7520e64de2dfaa516b6721c001e84e06e3c06a187b82af029ae721540ebfac; README SHA-256 4a9783ee8ebee10caaedfe2bdd2182539368e60e8eefadaee4cb7bab7fc75c12. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Route to the Electron guide 2. Describe establishing the correct application connection 3. Ask for or inspect the application state before actions 4. Do not reuse a web page reference blindly",
      "pass_threshold": 0.75,
      "origin": "Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call."
    },
    {
      "id": "dashboard",
      "prompt": "Do not use tools, access the network, write files, run commands or install anything. Answer only in text using the supplied facts. A browser dashboard is exposed through a forwarded origin. Its session tab links appear to point at an internal session port. Explain which address the operator should stay on and why.",
      "rubric": "Driftproof-authored text-only evaluation, not an author-supplied benchmark or an installed-tool test. Source vercel-labs/agent-browser/skills/agent-browser/SKILL.md at dac13bb87468fe9d74a040d862720f6cc17a9565; SKILL.md SHA-256 3a7520e64de2dfaa516b6721c001e84e06e3c06a187b82af029ae721540ebfac; README SHA-256 4a9783ee8ebee10caaedfe2bdd2182539368e60e8eefadaee4cb7bab7fc75c12. Score only the written answer. Award 0.25 for each fully satisfied criterion, 0 otherwise; sum to a score from 0 to 1. 1. Keep interaction on the dashboard origin 2. Explain that session traffic is proxied through it 3. Avoid exposing an internal session port as a requirement 4. Do not invent a successful connectivity check",
      "pass_threshold": 0.75,
      "origin": "Driftproof-authored from the pinned SKILL.md and README; no separate drafting model call."
    }
  ]
}