Landry Bodjona
Back to the datasheet

LB-07 QA Engineer

A browser agent that tests the company’s web shop from a plain-English goal, files bug reports, and writes the Playwright test that proves each bug.

LB-07

QA Engineer Evaluation board

Live

An AI agent tests a web shop the way a QA engineer would. Switch on some of the shop's six bugs, give a goal in plain words or pick a curated one, and watch the agent plan the test, run it in a sandboxed browser, file bug reports and write a Playwright test. The test is kept only if it fails with the bugs on and passes with them off.

The browser runs in a sandbox that can reach the staging shop and nothing else. Only Chromium is installed there, so the second engine is Chromium wearing Firefox's user agent: it catches a shop that treats Firefox differently, not a bug in Firefox's own engine.

Test the shop

Choose a curated test run

The goal, in English, as the agent reads it

Buy two bags of Ethiopia Guji with the coupon WELCOME10 and check the cart says 2 items and Total €26.10.

Bugs switched on

  • The coupon counts twice

A correct run comes toKept

A live run uses one of your two test runs today and makes two to eight model calls. A curated goal is the owner's, so it is not screened.

Your own test run

Switch on the bugs of your choice and say what to test in your own words. Your goal is screened for instructions to the agent, and the plan can use only a closed set of steps inside the shop.

Bugs to switch on

They reach the shop in a token the service signs, which the agent can neither read nor change.

Plain words, 3 to 300 characters. For example: buy two bags of Brazil Cerrado and check the total.

0 of 300 characters

The Scope

Start a run and its trace appears here, step by step.