A setup skill that gives non-vision models image understanding by routing pictures through a vision API via script.