Writing regression tests
The rules the agents follow when they write regression tests. Your team can follow them too.
One flow per test
Each test covers one thing a user does, from start to finish, and is linked to one manual test case.
Create your own data
Never depend on data that already exists. Create what the test needs through the API with a unique name, and remove it afterwards in reverse order. If clean up fails, the test fails.
Sign in through the API
Each user role signs in once before the run. Only one test should exercise the real login form.
Watch for hidden errors
Error messages, console errors and failed requests are collected during every test. A new one fails the test. Known issues are listed so they are reported without failing it.
Leave proof
Take a screenshot after every step, once its checks pass. In CI every test keeps a trace and a video.
Check what is inside files and emails
When a flow sends an email or produces a PDF, zip or CSV file, open it and check what is inside.
Use the shared steps
Reuse the steps in the shared library instead of writing new ones. If a step is broken, fix it in place so every test benefits.
Fire your QA backlog,
not your team
Let agents test every pull request, keep your regression suites green and send your team clear reports, videos and bug tickets.