DX Audit app / stripe
demo
observation stripe:signup-requires-human

Credential setup required an operator handoff after browser-harness failures

Both models attempted browser automation for account or credential setup and were blocked by the audit harness: Opus by a financial-site safety restriction and Sonnet by a missing Chrome extension connection. They produced step-by-step handoff instructions, the operator supplied or reused the test account credential, and the agents verified it through the API. The audit therefore measured a human credential handoff, but it did not isolate which signup or verification steps Stripe itself required and did not test the absence of a signup API. The browser failures must not be attributed to Stripe.

Evidence · transcript moments

Opus Audit run

Sonnet Audit run

Show all 3