Guide · Open source
How to use ian-xiaohei-illustrations: install the Codex Skill and make article images
Updated · written with AI from the project's own documentation and checked against our catalogue data
ian-xiaohei-illustrations is a Codex Skill, a set of instructions that guides an AI agent to make 16:9 hand-drawn illustrations for Chinese articles, posts, blogs and methodology content. It uses a fixed style: white background, black line art, a small black character called Xiaohei, and a few red, orange or blue Chinese handwritten notes. This guide shows you how to install it, ask Codex for a shot list or finished images, and check the results.
Ian Xiaohei Illustrationshelloianneo/ian-xiaohei-illustrations · ★ 13k · MIT Install options & downloads Quick start
- 1Clone the repository. Run git clone on helloianneo/ian-xiaohei-illustrations and change into the folder.
- 2Copy the skill folder. Create the Codex skills directory if needed and copy the ian-xiaohei-illustrations subfolder into it.
- 3Start Codex with the skill. Begin your prompt with $ian-xiaohei-illustrations and paste the article text.
- 4Request a shot list. Ask for a plan first so you can choose which paragraphs deserve an image.
- 5Generate the images. Ask for 16:9 horizontal PNG images with a white background and a few short Chinese annotations.
- 6Review the output. Check background, Xiaohei's role, text accuracy and whether the image looks like a diagram template.
What the skill does and does not do
The skill is built for article body images. Its core aim is to find one judgment, process, structure, state or metaphor in an article and draw it as a memorable 16:9 explanatory picture.
The README says it is not a general illustration prompt and not a PPT infographic template. Default output is PNG files.
- Produces a 16:9 horizontal body image, usually a shot list of 4-8 images per article
- Saves final PNGs under assets/<article-slug>-illustrations/ in your workspace
- Does not produce PPTX, PDF, SVG, HTML or Canvas editable files by default
- Is not meant for commercial posters, brand artwork or long text-heavy infographics
Install the skill in Codex
You need Codex with skills support and Git. Clone the repository, then copy the inner ian-xiaohei-illustrations folder into your Codex skills directory. The README uses CODEX_HOME if it is set and falls back to $HOME/.codex otherwise.
Only the ian-xiaohei-illustrations subfolder needs to be installed. The root README, LICENSE, NOTICE and examples are for GitHub sharing.
git clone https://github.com/helloianneo/ian-xiaohei-illustrations.git
cd ian-xiaohei-illustrations
mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
cp -R ./ian-xiaohei-illustrations "${CODEX_HOME:-$HOME/.codex}/skills/"Call the skill in a Codex prompt
After installation, start your request with $ian-xiaohei-illustrations and describe the article and how many images you want. The README example asks for five images for a Chinese article.
If you only want planning, say so explicitly. You can ask for a shot list first, with the paragraph each image belongs after, the topic, the core meaning, the structure type, what Xiaohei is doing, and suggested Chinese labels.
- Planning only: tell Codex not to generate images yet and ask for a 5-image shot list
- Full generation: ask for 4 images, 16:9, white background, black line art, a few red, orange or blue Chinese notes
- Single concept: quote the one idea you want drawn and ask that Xiaohei carry the core action
- Editing: ask to remove a specific title or wrong text while keeping the rest unchanged
Use $ian-xiaohei-illustrations 先不要生图。
请分析下面这篇文章哪里值得配图，输出 5 张左右的 shot list。
每张图写清楚：放在哪段后、主题、核心意思、结构类型、小黑在做什么、建议中文标注词。
<粘贴文章>Check each image against the style rules
The skill's QA checklist looks at white background, empty space, Xiaohei's action, Chinese labels, whether the image feels non-PowerPoint, and whether it avoids copying the example images. The README notes that image models can produce typos, invented labels, style drift or an unwanted extra title.
If Chinese text has serious typos, reduce the number of labels and regenerate. Short labels are more reliable.
- Background is pure white with no paper texture, shadow or gradient
- Subject fills roughly 40%-60% of the frame
- Xiaohei does part of the core action; if removing it leaves the image unchanged, it is too decorative
- One core structure per image
- Example images are for calibration only, not composition templates
What people use it for
- Adding one metaphor image to a Chinese blog post about a decision or trust problem
- Turning a methodology article's workflow into a single hand-drawn state or before-and-after picture
- Producing a consistent set of 4-5 body images for a series of AI workflow posts in Chinese
- Redrawing a single concept sentence as one illustration for a Notion or Markdown article
Frequently asked questions
Is ian-xiaohei-illustrations free to use?
The repository is released under the MIT License, as stated in its README and LICENSE file. Check the LICENSE file in the repo for the full terms before you use it in a project.
What do I need to install ian-xiaohei-illustrations?
The install method is clone-and-build with two prerequisites: Codex with skills support and Git. You clone the repository and copy the ian-xiaohei-illustrations subfolder into your Codex skills directory.
Where does the output image get saved?
According to the README, final PNG files are saved to assets/<article-slug>-illustrations/ in your workspace. The default output is PNG only.
Can I edit the generated images as vector files?
Not by default. The README says the skill does not output SVG, HTML, Canvas or PPTX/PDF/Keynote files by default, so it is not the right choice if you need editable vector source files.
Can I use it to make PPT-style flowcharts or architecture diagrams?
No. The README states it is not a PPT infographic template and says it is not meant for complex architecture or flowcharts. Tools like Mermaid or Excalidraw are listed as comparable projects for diagram work.
Why does the text in my image have typos?
The README warns that AI image models can produce typos, invented labels and extra titles. Use fewer and shorter Chinese labels, then regenerate the image.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you write Chinese articles, posts or methodology content and want a consistent, quirky hand-drawn body image style inside Codex. Pick something else if you need commercial or brand illustration, PPT-style diagrams, or editable vector files.
Install options & downloads