rouor/rouor-illustrations

ROUROU Illustrations is an AI agent skill for Codex and Claude Code that generates in-article illustrations for Chinese language writing, such as blog posts, WeChat articles, and Notion documents. Instead of producing generic infographics, it turns a single idea, process, structure, or state described in the text into a quiet, slightly odd, paper craft style picture. The visual identity is built around a recurring character called ROUROU, a warm off white crumpled paper figure with a dusty blue paperclip and a terracotta red stitch, drawn with charcoal black hand drawn lines on a pure white background. The rules are specific: the character takes up roughly 40 to 60 percent of the image, at least about 35 percent stays empty space, only one ROUROU appears per image, and each image expresses exactly one core action or idea rather than trying to cram in multiple concepts. Handwritten Chinese annotations are limited to three to six per image, used only when needed, and the color palette is a restrained set of seven muted tones with only 10 to 18 percent of the image in color. To use it, you copy the rourou-illustrations subfolder into either Codex's skills directory or Claude Code's skills directory, either at the user level or inside a single project. After installing, you invoke it inside a conversation by asking it to design and generate a set number of illustrations for a pasted article. The skill first checks whether the current session actually has an image generation or editing tool available. If it does not, it still produces a full shot list with prompts, reference image paths, and file names for every planned image, clearly marked as not yet generated. The included workflow has the skill first extract the article's core points and turning points, choose which ones are worth illustrating, produce a shot list, invent a paper based visual metaphor for each one, generate the images one at a time using whatever image backend is available in the current host, and check each result against a quality checklist before saving. The project is adapted from an earlier open source skill by another creator, with full attribution given in a separate notice file, and is released under the MIT License.