The offer says 20% off.
The cart says otherwise.
Merchants ship storefront changes faster than they can verify every shopper journey. A banner can promise a discount while the actual bag silently charges full price.
One merchant decision.
One clear path.
K9 compares the same first-order goal across control A and challenger B, then asks Kane to verify the real browser cart before the merchant acts.
Simulation is a signal. Kane’s browser result is the proof.
Kane reads what
the shopper sees.
The coding agent used Kane to generate a cart test, reviewed its assertions, ran it, read the failure, repaired the store, and replayed the same test.
- DESIGNApprove generated
_test.mdfor the discount journey - RUNCheck the displayed discount line and cart total in Chrome
- ACTUse the sealed evidence pack to guide and verify the repair
Red run.
Repair. Green rerun.
The reviewed challenger test found the offer missing from the bag. After the fix, the same test saw the correct discount and total. A later two-test suite passed for challenger and control.
Turn a proven journey
into a release habit.
The focused discount journey is verified. Next, connect Kane credentials to CI, require the suite check on pull requests, and expand coverage to more commerce promises.
Current proof: same challenger test failed, then passed; control and challenger suite passed 2/2. The CI merge rule is not active yet.