Benchmark report
Task verified
4m 15s elapsed
The saved attempt met the documented success criterion.
94/ 100
Agent-Ready- AInstall
- SAuth
- SExecute
- ADocs
- SCurrent
A PEP 668 pip block forcing a venv, plus the plan's browser-use.com/playwright page being unreachable and a guessed quickstart URL returning 404.
Where the time went
4:16 agent time- Research0:33
- Attempt2:22
- Verify1:21
Judge reason
Live end-to-end run verified from saved tool output, not prose. run_task.py creates a session via the documented POST https://api.browser-use.com/api/v4/browsers with the X-Browser-Use-API-Key header; session_create_response.json shows HTTP 201, id 07aea751-dfaa-48b9-bed0-f31ecc2a2837, status active, cdpUrl https://07aea751-...cdp.browser-use.com (no credential in URL). Playwright 1.65.2 (installed in a venv, import confirmed) connects via chromium.connect_over_cdp(cdpUrl, headers={'X-Browser-Use-API-Key': key}); run_log.txt from the live Bash result shows load of https://httpbin.org/forms/post, post-submit URL https://httpbin.org/post, and echoed_response.json with form.custname 'Benchmark Test' and form.custtel '555-0100' plus in-script asserts passing. The echo carries remote-browser markers (Chrome/156 macOS UA, origin 68.165.186.185) inconsistent with any local/mocked run; no AI agent/task API was used. Termination runs in a finally block: PATCH /api/v4/browsers/{id} {"action":"stop"} -> HTTP 200, status 'stopped', finishedAt set, saved in session_stop_response.json. Printed log matches saved echo byte-for-byte; agent's grep for the key across artifacts returned 0 files and my independent scan of /workspace/home found no credential or key-bearing query string (only liveUrl's wss= pointing at the plain cdp host).
Attempt record
- 0:47Answered from saved key: Browser Use Cloud API keyno wait
- —First success verified
Attempt artifacts saved.