Depends a lot on the application tech stack. A generic tool is usually the worst idea. Further, I would recommend writing tests along the axis of minimum change.