Test set
Ten real inputs with known answers, run every time you change anything.
A test set is a small collection of real inputs where you already know the right answer. Run it each time you change the prompt, the model or the tool. If the score drops, roll back.
One good example proves nothing, because the model varies from draw to draw. Ten or twenty items show whether a change helped or hurt in a minute.
Find your way around
Learn it in the course, copy it from a playbook, see it work in a case study.