← Portfolio /LAB 05 /UAT Sentinel
Open fullscreen ↗
← Portfolio /LAB 05
05
QA AUTOMATION

UAT Sentinel

QA scripts break the moment the UI changes. Which means they break constantly. I built an agent that tests in plain English — describe what it should do, it runs the test, flags what failed, and explains why. No code. No stale scripts. No waiting for a QA cycle.

L
THE LOOP
Goal written in plain English. Browser agent executes it — clicks, fills, checks, screenshots. Fails are flagged with exact reproduction steps. Evidence saved per run.

Six flows run at once: browse, search, product page, checkout, account. Each scored against established e-commerce UX benchmarks. Every result comes back as ship, needs review, or hold — with screenshots and a plain explanation of what failed.

A
THE ARTIFACT
A structured audit run: what was tested, what failed, reproduction steps, and a release recommendation. Human makes the go/no-go. Agent provides the evidence.

Screenshots, what went wrong, ranked by how serious it is — ready for the standup.

→
POTENTIAL APPLICATIONS
Built for engineering teams that ship fast and QA teams that can't scale scripts fast enough. Any product with a browser UI.
AGENTIC LAB · 05 / 06