Benchmark report
Task verified
4m 7s elapsed
The saved attempt met the documented success criterion.
94/ 100
Agent-Ready- SInstall
- SAuth
- SExecute
- ADocs
- BCurrent
Docs-page response field names (industry, companySize, founded) differ from the live payload (industries, staff.*, foundedYear), so the first run printed None and the client had to be rewritten and re-run.
Where the time went
4:08 agent time- Research0:25
- Attempt2:05
- Verify1:37
Judge reason
Saved code code/enrich_company.py POSTs to the documented operation https://api.richapi.ai/api/v1/enrich_company with an x-api-key header read from a private key file and body {"url":"https://www.linkedin.com/company/stripe/"}, matching the fetched docs page https://richapi.ai/data/company-enrichment (which states domains are not accepted). Transcript idx 18/21 and 46/49 show two live runs returning HTTP status: 200; saved responses are plainly live (distinct richapi_meta.request_id, credits decrementing 5021.94 -> 5020.94 -> 5019.94, signed LinkedIn CDN URLs), not mocked. outputs/enrich_company_stripe_response.json has name "Stripe", website "https://stripe.com", industries ["Computer Software"], staff.total 16694, staff.range 5001-10000, and the stdout file prints exactly those values; the curl run's stdout (16693) matches its own saved response. commands.sh reproduces the call with curl --config so the key never hits argv. A grep of all artifacts for the key and for access_token/bearer/x-api-key values is clean, and the responses contain no token field so no redaction was required.
Attempt record
- 0:54Answered from saved key: RichAPI API keyno wait
- —First success verified
Attempt artifacts saved.