The Prompting Company

Benchmark report

Task verified

5m 9s elapsed

The saved attempt met the documented success criterion.

Experience score

96/ 100

Agent-Ready
  • SInstall
  • SAuth
  • SExecute
  • ADocs
  • SCurrent
Experience measures friction in the tested task. The verdict says whether it succeeded.
Agent profile for this run: Install S, Auth S, Execute S, Docs A, Current SSInstallSAuthSExecuteADocsSCurrent
Agent profile · this run
Judge's note · what slowed the agent most
Guessing documentation URLs that 404'd before discovering the llms.txt index that lists the real page paths.

Where the time went

5:09 agent time
  • Research0:21
  • Attempt3:27
  • Verify1:21
5m 9sTotal elapsed
2:27Time to first success
0:00Waited on the owner

Judge reason

Saved code evidence/run.mjs uses the documented operations: kernel.browsers.create() (POST /browsers, bearerAuth per docs/create-session.md) returning session_id gywg9u5e5kr8b9fpqdgowq60 and cdp_ws_url, then chromium.connectOverCDP(created.cdp_ws_url) exactly as docs/control.md:183 shows. Transcript line 81-84 shows the live run with EXIT=0; run-output.txt and echoed-response.json carry the real httpbin echo (origin 108.62.48.73, a remote egress IP, Chrome/155 UA) with form.custname='Benchmark Test' and form.custtel='555-0100', and the printed JSON matches the saved file byte-for-byte. Termination is in the finally block via kernel.browsers.deleteByID(sessionId).asResponse() returning HTTP 204 No Content, which docs/delete-session.md:102 documents as 'Browser session deleted successfully'. The API key came from benchmark-request, was read from a file outside the home dir and sent only as the SDK's Authorization: Bearer header; all URL query params are redacted in saved artifacts and my own grep over evidence/ found no key or jwt. No mocking or hard-coding; Playwright drove the page, not Kernel's agent API.

Attempt record

  1. 2:13Answered from saved key: Kernel API key (KERNEL_API_KEY)no wait
  2. —First success verified

Attempt artifacts saved.