Taste Skill: Website Design Comps

Most skills in the Taste Skill collection steer agents away from generic front-end code. This one works differently: it writes no code and instead directs an image model to paint high-end website design comps — hero sections, landing pages and multi-section sites with strong typography, spacing and art direction.

It enforces one horizontal image per page section, so an eight-section landing page becomes eight separate renders rather than one cramped poster. Those images are meant as a visual target: you pass them to a coding agent, or to the collection’s image-to-code skill, to implement. It works with ChatGPT image generation, Codex’s image mode or any agent that can generate images, and a sibling skill does the same for mobile screens.