Attack lab

Get an approved receipt from the loop above (or on the demo page), then try to stretch it. Each box sends the receipt with one thing changed.

Same approval, different invoice

Expected: refused: the approval covers a different action

Same approval, different agent

Expected: refused

Same approval, with a body

Expected: refused: the body hash differs

Same approval, second time

Expected: the first mint succeeds, the second is refused: this approval was already used

Open-request attacks (self-approval by case, a person listed twice, misleading paths) are blocked when a request is created. The public demo only opens its own two scenarios, so those are shown on the evidence page from the engine test run.

FIRME AI CORPORATION · CAIN-42 is a Byzantine trust fabric. Numbers marked "our own check" were run by us, not by an independent party. Nothing on these pages is a certification.