Benchmark report

Task verified

5m 25s 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
The canonical docs hosts were unreachable from the sandbox, so the agent spent several calls locating the equivalent docs.gtm.ai mirror before coding.

Where the time went

5:25 agent time
  • Research0:32
  • Attempt3:26
  • Verify1:26
5m 25sTotal elapsed
2:17Time to first success
0:00Waited on the owner

Judge reason

Saved code code/enrich_stripe.py performs the documented ZoomInfo GTM Data API flow: OAuth client-credentials POST https://api.zoominfo.com/gtm/oauth/v1/token (HTTP Basic, scope api:data:company) then POST https://api.zoominfo.com/gtm/data/v1/companies/enrich with Content-Type/Accept application/vnd.api+json and body {"data":{"type":"CompanyEnrich","attributes":{"matchCompanyInput":[{"companyWebsite":"stripe.com"}],"outputFields":[...]}}}. This matches the saved OpenAPI snapshot docs/reference_enrichinterface_enrichcompany.md (path /data/v1/companies/enrich, operationId EnrichInterface_enrichCompany, server https://api.zoominfo.com/gtm). output/enrich_http_status.txt records '200 / application/vnd.api+json'; output/enrich_response.json is a JSON:API document with data[0].type=Company, id=352810353, meta.matchStatus=FULL_MATCH, name='Stripe', website='www.stripe.com', domainList including stripe.com, primaryIndustry ['Software','Financial Software'], employeeCount 8000, employeeRange '5,000 - 10,000'. Transcript tool result at the run step shows the same values printed live (console also saved at output/run_console.txt), so printed values match the saved response; the company was identified by domain as the docs allow. Credentials came from /usr/local/bin/benchmark-request (files under /tmp/benchmark-control, read only inside code) and were never echoed; output/token_response.redacted.json has access_token/id_token/refresh_token replaced with "[redacted]" and keeps token_type, expires_in and scope unchanged. A repo-wide grep in the transcript found no real JWT in any saved artifact — the only eyJ... match is the public documentation's own truncated placeholder in docs/docs_client-credentials-flow.md, and run_console.txt contains only the literal word 'Bearer' as token_type metadata. No mocking or hard-coded data: response depth (employeeGrowth points, naicsCodes, 12,270 contacts, noisy domainList) is consistent with a live ZoomInfo record.

Attempt record

  1. 2:05Answered from saved key: ZoomInfo GTM API OAuth client credentials (Client Credentials Flow)no wait
  2. —First success verified

Attempt artifacts saved.