Frontend & UI
image-to-code
Generates design images first, analyzes them, then builds matching frontend code.
What it is
An image-first workflow for visual web tasks. The agent generates section reference images, extracts typography, spacing, color, and components from them, then implements the site to match.
Elite website image-to-code skill for Codex. For visually important web tasks, it must first generate the design image(s) itself, deeply analyze them, then implement the website to match them as closely as possible. In C
Why it matters for design
- Code follows a real visual reference, so implementation does not drift into generic layouts.
- Each section gets its own large image, keeping text and spacing analyzable.
- Heroes stay under three headline lines and free of nested container stacks.
What it covers
- Mandatory image-first rule
- Generate enough images rule
- Deep image analysis requirement
- Hero minimalism rules
- Anti-drift implementation rule
- Anti-AI-slop rules
How to install
image-to-code
-
Clone the repo.
git clone https://github.com/Leonxlnx/taste-skill -
Point Codex at the skill file.
skills/image-to-code-skill/SKILL.md -
Then describe the UI you want. Codex follows the skill.
How to run it with Codex
-
State the section count; in Codex the skill generates one image per section.
-
Ask for a closer detail render when button or type detail stays unreadable.
-
Have it run the clarity check before it writes any implementation files.
Every plugin here is free, open source, and links to its real upstream source.
More on the codex-design list ↗Next step
Design with Open Design, without the setup
Install this plugin yourself, or run a whole curated design layer around Codex with Open Design. Bring your own key, own your output.