Benchmark report

Onboarding verified

2m 53s elapsed

The saved attempt met the documented success criterion.

Experience score

100/ 100

Agent-Ready
  • SInstall
  • SAuth
  • SExecute
  • SDocs
  • 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 S, Current SSInstallSAuthSExecuteSDocsSCurrent
Agent profile · this run
Judge's note · what slowed the agent most
Only real delay was obtaining the API key; after that the whole run took ~30s, with a brief detour reconciling one API result that had an empty title.

Where the time went

0:04 on the owner · 2:49 agent time
Waited on the owner 0:04
  • Starting0:02
  • Research0:16
  • Attempt1:22
  • Verify1:09
2m 53sTotal elapsed
0:30Time to first success
0:04Waited on the owner

Judge reason

Saved evidence proves a live keyed call: run_search.sh/command_redacted.txt show POST https://api.parallel.ai/v1/search with x-api-key and an objective+search_queries body; http_status.txt=200 and transcript tool output 'HTTP_STATUS=200 bytes=49361'. search_response.json (identical to evidence_search_response.json) contains search_id 'search_c378edebd4a6b775d97073f26415b857', session_id, usage [{sku_search:1}] and 10 results; 9/10 have non-empty url, title and a non-empty excerpt (lengths 2000-22586 chars). Result index 9 (linkedin.com/company/parallel-web) came back with an empty title but valid url and excerpt — upstream data, not a call failure, and the agent disclosed it in result.json and NOTE.md rather than hiding it. No install needed; key obtained via benchmark-request and never exposed.

Attempt record

  1. 0:34Asked the owner for Parallel API keywaited 4s · provided
  2. —First success verified

Attempt artifacts saved.