Turn a written WeChat article into a set of up to ten hand-drawn style images with Chinese text embedded directly in the artwork.
Generate portrait-format illustration covers sized for Xiaohongshu or short-video posts from a written script.
Export complete image-generation prompts from the skill to paste into a separate AI image tool when your agent cannot generate images directly.
Convert a personal retrospective into a series of images using the Juju character to visually organize key ideas.
Copy the skill directory into your AI agent's skills folder and point the agent at the SKILL.md file to activate.
Juju Content Illustrations is a skill for AI agents that converts written content into hand-drawn style illustration sets. The project is in Chinese and is designed for Chinese content creators who publish on platforms like WeChat public accounts, Xiaohongshu, and X (formerly Twitter). The character at the center of the visual style is Juju, described as a white Bichon dog who acts as a content organizer: circling key ideas, separating relationships, and lighting up paths through complex material. The skill works by taking an article, tutorial, method explanation, or personal retrospective and turning it into one or more images. You paste your content into the agent, choose a target aspect ratio and platform, and the skill decides how many images to generate, up to ten. Options include landscape formats for article headers, tall portrait formats for short-video covers, and square formats for social media. Each image is meant to explain a single idea rather than cramming everything into one graphic. Chinese text is embedded directly into the artwork, not added as a caption afterward. The visual style is consistent across outputs: white or near-white background, black light line art, low-saturation accent colors that vary by content tone, and the Juju character performing some organizing action in each scene. The README describes a concept called a parallel world: a low-technology, touchable paper setting that fits the content, such as a photography practice space, a method-sorting desk, or a repair workshop for retrospectives. The parallel world is reinvented per piece of content rather than reused from a fixed template. If the AI agent you are using can generate images directly, it does so. If not, the skill outputs complete image-generation prompts you can copy into another tool. Installation is done by copying the skill directory into your agent's skills folder and pointing the agent at the SKILL.md file. The project is released under the MIT license.
← dososo on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.