FIDELITY RECORD
Verified against real Stripe: every probe we run, and its result.
The Stripe drift battery checks selected implemented behaviors: each probe sends the same requests to aWorlds world and to real Stripe test mode, then compares the responses. A deleted customer returns 200 with a stub. An over refund is refused with Stripe's exact wording. A replay under the same idempotency key returns the first response. The full battery runs in every test run of the product. Against real Stripe it runs whenever a test key is present, and the newest run is what you see here.
This historical record predates the current beta candidate. It is evidence for the version shown below; refreshed candidate drift acceptance is still required. It does not establish Zendesk qualification.
LAST RUN
2026-09-28
PROBES
20
AS EXPECTED
20 OF 20
ALL OF THEM
MUST AGREE
16
DOCUMENTED GAPS
4
| PROBE | EXPECTED | RESULT |
|---|---|---|
| customer-crud-and-deleted-stub | MUST AGREE | AS EXPECTED |
| unknown-id-404-envelope | MUST AGREE | AS EXPECTED |
| pi-confirm-visa | MUST AGREE | AS EXPECTED |
| pi-confirm-declined-persists-failure | MUST AGREE | AS EXPECTED |
| over-refund-message | MUST AGREE | AS EXPECTED |
| charge-already-refunded | MUST AGREE | AS EXPECTED |
| refund-listed-on-charge | MUST AGREE | AS EXPECTED |
| idempotency-replay | MUST AGREE | AS EXPECTED |
| idempotency-param-mismatch | MUST AGREE | AS EXPECTED |
| idempotency-validation-not-reserved | MUST AGREE | AS EXPECTED |
| idempotency-validation-amount-zero | MUST AGREE | AS EXPECTED |
| pagination-has-more | MUST AGREE | AS EXPECTED |
| invoice-lifecycle-draft-open-void | MUST AGREE | AS EXPECTED |
| customer-balance-credit-applies | MUST AGREE | AS EXPECTED |
| invoice-pay-autocharge | DOCUMENTED GAP | AS EXPECTED |
| pi-capture-manual-refused | DOCUMENTED GAP | AS EXPECTED |
| price-trial-period-days | DOCUMENTED GAP | AS EXPECTED |
| balance-txn-invalid-currency | MUST AGREE | AS EXPECTED |
| balance-txn-second-currency-refused | DOCUMENTED GAP | AS EXPECTED |
| credit-other-currency-not-applied | MUST AGREE | AS EXPECTED |
Run 2026-09-28T08:12:30.072Z against real Stripe (test mode), using Worlds 0.9.0-beta.1. Source: 2026-09-28.json. Probes create only their own tagged objects in the test account and touch nothing else.
WHAT A GAP IS
A gap is a difference we keep on purpose and document: a parameter real Stripe accepts that Worlds refuses with a 400 that names it, so an agent relying on it finds out immediately instead of being graded on a guess. The battery fails if a gap closes without the docs saying so, just as it fails when an agreement drifts. The scope page lists what Worlds implements and what it deliberately leaves out.
Live checks compare implemented behavior and retain known gaps. Read the date, version and probe list when deciding what the record establishes. A passing battery is not complete Stripe API parity.