Benchmark report

Task verified

10m 50s elapsed

The saved attempt met the documented success criterion.

Experience score

88/ 100

Agent-Ready
  • SInstall
  • AAuth
  • SExecute
  • ADocs
  • ACurrent
Experience measures friction in the tested task. The verdict says whether it succeeded.
Agent profile for this run: Install S, Auth A, Execute S, Docs A, Current ASInstallAAuthSExecuteADocsACurrent
Agent profile · this run
Judge's note · what slowed the agent most
Official ZoomInfo doc hosts blocked by the proxy, plus a misleading 400 'not configured with any authorized scopes' on the body-param token variant, forced doc and auth probing.

Where the time went

10:51 agent time
  • Research0:40
  • Attempt6:42
  • Verify3:28
10m 50sTotal elapsed
4:49Time to first success
0:00Waited on the owner

Judge reason

Saved code code/zi_company_search.py + zi_client.py performs a live OAuth client-credentials token call then POST https://api.zoominfo.com/gtm/data/v1/companies/search; evidence/company_search_response.json records http_status 200, content_type application/vnd.api+json, meta.totalResults 402, links first/next/last and exactly 10 Company records. Filters were sent server-side in the request body (industryCodes=software + primaryIndustriesOnly=true, country=germany + locationSearchType=HQ, employeeRangeMin/Max 201/500, page[number]=1&page[size]=10) - all four confirmed in the saved OpenAPI spec docs/search-companies.md (CompanySearchRequestAttributes), so no post-filtering. All 10 results show country Germany and employeeCount 487-500; printed stdout matches the saved JSON name-for-name. ZoomInfo's lookup taxonomy (output/lookup_industries.json, 180 values) has no 'Software Development' value, so 'software' was used and separately confirmed: enrich stdout shows primaryIndustry contains 'Software' for all 10 via the documented Enrich endpoint (HTTP 200). Saved artifacts and output/token_response.redacted.json contain no token (access_token/id_token/refresh_token = [redacted]); scans found JWT-like text only in the vendor doc sample. One hygiene lapse: an early auth-variant probe (transcript line 148) echoed a token response truncated at 200 chars, exposing a 142-char JWT prefix (header + partial payload, no signature, unusable) into the session log only; the agent detected, disclosed and fixed it, and no deliverable is affected.

Attempt record

  1. 2:57Answered from saved key: ZoomInfo GTM API OAuth client credentialsno wait
  2. —First success verified

Attempt artifacts saved.