You already have the requests. You ran them once, found a bug, fixed it, and moved on. The gap in most CI pipelines is not detection β it is that nobody ever re-runs the request that proved the bug in the first place.
This post is about a small harness that fixes that: take the requests your testing produced, replay them against a second tenant's session, and fail the build when the response says "yes" to a question it should have said "no" to.
The shape of the test
A cross-tenant authorization bug has a signature that is easy to express as a test:
An object exists for tenant A (invoice 4821, project 91, report 17).
Tenant B holds a valid session.
Tenant B requests A's object by ID.
The response is 200 with A's data instead of 403 / 404 .
Steps 1β3 are the hard part to write by hand and the trivial part to automate.β¦