pyang5166/gbro-cover-design — explained in plain English
Analysis updated 2026-05-18
Generate a cover image prompt for a WeChat public account article.
Create a Xiaohongshu style cover that keeps the same face across posts.
Pick from 10 preset composition styles instead of designing a cover from scratch.
Quickly produce a cover prompt for a video thumbnail or product announcement post.
| pyang5166/gbro-cover-design | affaan-m/jarvis | airbnb/viaduct | |
|---|---|---|---|
| Stars | 158 | 158 | 158 |
| Language | — | Python | Kotlin |
| Setup difficulty | — | hard | moderate |
| Complexity | 2/5 | 5/5 | 4/5 |
| Audience | writer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
gbro-cover-design is a skill for AI coding agents like Claude Code and Codex, or any agent that supports custom skills, that turns a written article into a ready to use image prompt for a cover image. A user gives the agent an article, the skill asks three rounds of questions, and it then outputs a single prompt that can be run through an image generation model to produce the cover. It also suggests a title automatically. The output image is always a 3:4 vertical portrait format, matching the cover size used on Chinese platforms like WeChat public accounts and Xiaohongshu. The skill is an improved version of an earlier MIT-licensed project called oh-my-cover-design. The changes described in the README are compressing the question flow down to three rounds, loading 10 composition style templates only as needed, adding a built-in library of example prompts used as few-shot references, fixing the image to a 3:4 aspect ratio with a defined safe area, and adding a first-time setup guide. On first use, the skill walks the user through setup without needing any API key, since the skill only produces a text prompt rather than generating the image itself. The user provides a clear front-facing photo of their face, saved as a default reference image, and confirms that the image generation model they plan to use supports multiple reference images, which is needed to keep the person's face consistent across generated covers. This configuration is saved so the setup questions are not repeated later. During normal use, the agent asks about composition style and cover title, then requests a face reference photo and any other material like product screenshots, then asks about expression, background tone, font, and font color, letting the model decide anything left unanswered. The skill includes 10 named composition styles, such as a dark gradient style, a flat solid color style, a product screenshot focused style, and a minimalist style with large empty space, each with its own prompt template and example. Installation is a plain git clone into the user's Claude Code skills folder, and the README notes the full repository, not just the main skill file, must be cloned because the style templates live in a separate folder. The project is released under the MIT license.
A Claude Code and Codex skill that turns an article into a ready-to-run image prompt for a 3:4 vertical article cover, with face consistency.
Free to use, modify, and distribute for any purpose, including commercial use, as long as you keep the copyright notice.
Mainly writer.
This repo across BitVibe Labs
double-check against the repo, no cap.