A skill for visual QA of agent-built UI
I gave my podcatcher app a major UI polish this week, ahead of the public beta.
The annoying thing, for this kind of work, is that even with a skill like /impeccable, agents often present work with all sorts of small visual bugs: misaligned elements, insufficient margins, button labels that wrap onto two lines. The issues are obvious if you just look, with your eyes.
/visual-review is a skill that makes my agents look. They inspect screenshots of the views they’re working on, at all breakpoints. This step—plus some programmatic inspection—lets them catch many visual bugs before I see them.
