SkillInstalls cleanly
Design mockup images before you build
imagegen-skills
Four skills that guide image generation for website mockups, app screens and brand boards, then build a website that matches the images you picked.
- Works with
- Codex CLI, Claude Code
- Good for
- Graphic designers, Web developers
What it does
Before building a website or an app, designers often make picture mockups first, so everyone can see the look before any code exists. These skills help an AI image tool make those mockups well, instead of the same generic "AI look" every time.
There are four skills:
imagegen-frontend-webmakes one image per section of a landing page.imagegen-frontend-mobilemakes matching app screens.brandkitmakes brand boards with logo ideas, colours and fonts.image-to-codedoes the whole job: it makes the design images, studies them, and then builds a website that matches.
The skills are written instructions only: rules about layout variety, readable text and one consistent colour palette. They contain no scripts, call no image service of their own and need no key. The images come from your assistant's own image generator.
That is why they work best in Codex. OpenAI's Codex docs say its built-in image generation uses the gpt-image-2 model and counts toward your normal Codex usage limits. Images use those limits 3 to 5 times faster on average than other requests. If you set an OpenAI API key, Codex can make images through the API, and then you pay API prices instead.
Claude Code has no built-in image generator. There you can make the images in another tool and give them to Claude Code.
The web skill makes many images on purpose: six for a landing page and eight for a full website when you do not give a number.
Who it is for
People who want to see and pick a design as pictures before the assistant builds the page or app, and who use an assistant that can make images, such as Codex.
Example prompts
- "Use imagegen-frontend-web to design a landing page for my yoga studio, six sections."
- "Make three app screens for a habit tracker with imagegen-frontend-mobile, light and dark."
- "Use brandkit to propose two logo directions and a colour palette for my coffee shop."
- "Use image-to-code: design the page images first, then build the site to match."
How to install
/plugin marketplace add Leonxlnx/imagegen-skills/plugin install imagegen-skills@imagegen-skillsHow to use it
In Codex, ask: Use imagegen-frontend-web to make design images for a landing page for my bakery. Then use image-to-code to build the page from them.
Other assistants
Codex CLI (from the README). Start a new Codex session afterwards so the skills load:
codex plugin marketplace add Leonxlnx/imagegen-skills
codex plugin add imagegen-skills@imagegen-skillsSkills CLI, for any assistant (from the README). Leave out --skill to install all four, or swap in another skill name:
npx skills add https://github.com/Leonxlnx/imagegen-skills --skill "imagegen-frontend-web"We have not tested the Codex route.
Check it works
- In Claude Code, type
/and look for/imagegen-skills:image-to-codeand the three other skills. - In Codex, start a new session after the install. Name the skill in your request, for example "Use imagegen-frontend-web to ...". OpenAI's docs say you can also type
$imagegento call Codex's own image generator directly.
Common problems
- The skills do not show up in Codex — the session started before the install — start a new Codex session, as the README says.
- You installed these from taste-skill before — the skills moved — install them from this repo; the install names did not change (README).
- No images appear in Claude Code — Claude Code has no built-in image generator — make the images in an image tool and give them to the assistant (our reading).
- Your Codex limit runs out quickly — each image uses more of your limit than a normal request — ask for fewer sections, or set an OpenAI API key so API prices apply (Codex image generation docs).
What to watch out for
- Needs an assistant that can make images: the README says three of the skills produce images only and image-to-code needs an agent that can both generate images and write code; Claude Code has no built-in image generator, so there you would hand the images over from another tool (our reading)
- Uses up your limits faster: the web skill makes one image per page section (six for a landing page by default, per SKILL.md), and OpenAI says image generation in Codex uses included limits 3 to 5 times faster on average
- One-person project: the skills moved here from taste-skill in October 2026; check the repo for updates
Licence and author
- Licence
- MIT
- Author
- Leonxlnx
- Source code
- github.com/Leonxlnx/imagegen-skills
How we tested it
We installed it on an empty test computer on 22 Oct 2026. Installed and started in a clean test system. Not yet used in a real AI session.
Not run: "codex plugin marketplace add Leonxlnx/imagegen-skills" (not a form the sandbox runs). Not run: "codex plugin add imagegen-skills@imagegen-skills" (not a form the sandbox runs). No image was generated. Not yet used on real work.
Technical details of the test
- Date
- 22 Oct 2026
- Method
- Intake sandbox run on 2026-10-22 in a clean container with no account: cloned the repository and checked its skill, agent and plugin files, ran "npx -y skills --help"; installed imagegen-skills@imagegen-skills with the Claude Code command line.
Checked again on 22 Oct 2026
Alternatives
- UI/UX Pro Max — design rules for building the interface in code, no image step.
- Frontend Design — Anthropic's plugin that helps Claude write distinctive page code directly.
- brand-identity — brand strategy and guidelines in text.
FAQ
Does it make the website too?
Only image-to-code does. The other three make images.
Which assistant should I use?
One that can make images and write code, such as Codex, for image-to-code.
Why so many images? The web skill makes one image per page section on purpose, so each section can be built accurately.
Is it free? The skills are free (MIT licence). The images count toward your assistant's plan limits, or cost API prices if you use an API key.
Sources
- github.com/Leonxlnx/imagegen-skills
- github.com/Leonxlnx/imagegen-skills/blob/main/skills/imagegen-frontend-web/SKILL.md
- developers.openai.com/codex/image-generation