How to use the Ian Xiaohei Chinese Article Illustrations agent skill

What is an agent skill and what does this one do?

An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches the skill's description. This skill is for Chinese-language writing. When you ask for illustrations, the agent reads your text, picks the key ideas worth visualizing, and proposes where images would help.

When you ask it to generate images, it creates one 16:9 black line-drawing image per key point on a clean white background. Each image has a small black Xiaohei character doing the central action and sparse red, orange or blue handwritten Chinese annotations. The skill also checks each image against a quality checklist and suggests regenerating when the character is only decorative, the image is cluttered, or the text has errors.

• Plans where a Chinese article needs illustrations and proposes a short shot list (default 4-8 images).
• Generates one standalone image per key point, not collages.
• Reviews results for decorative characters, clutter, PPT-style layouts and Chinese text errors.
• Saves final images to an assets folder when working inside a workspace.

Install the ian-xiaohei-illustrations skill

Download the ZIP of the skill from our page and unzip it. You can install it in one of two places. For personal use across projects, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder so it is available only there.

Alternatively, you can upload the skill in the Claude apps under Settings → Capabilities → Skills. Use the method that matches how you work; the skill files are the same either way.

• Personal install: ~/.claude/skills/
• Project install: the .claude/skills/ folder inside your project
• App install: Settings → Capabilities → Skills

Check what the skill needs before you start

The skill relies on the built-in image_gen image generation tool in your agent environment. Without it, the agent can still plan the shot list, but it cannot create the images. If you want the finished files saved, the agent also needs workspace file access so it can write to assets/.

The skill does not run scripts and does not use the network itself. Its files are instructions and reference documents, including a prompt template, a style guide, a character guide, composition patterns and a QA checklist.

• Required: built-in image_gen tool in the agent environment.
• Only if saving deliverables: workspace file access to write into assets/.
• Runs scripts: no.
• Uses network: no.

Trigger the skill with natural requests

You do not need special commands. Describe what you want in Chinese, and mention illustrations, images, or the style. The skill's description covers requests about articles, posts, blogs, Notion docs, workflow documents, methodologies, processes, and phrases like 正文配图, 文章插图, 配图建议 and shot list.

Ask for a plan first if you are not sure where images belong. Ask for generation once you have agreed on the shot list.

• Plan only: 分析这篇文章哪些地方需要配图，给我一个 shot list。
• Plan a Notion doc: 这篇工作流文档里哪些段落适合用小黑配图？
• Generate: 帮我生成这篇文章的第二张配图，主题是两个断点之间的承接。
• Revise: 这张图小黑只是站在旁边，帮我改成小黑参与核心动作。

Example workflow for a long article

Paste your article text, a link, a Markdown file or a Notion page. The agent first identifies the core claim, the sections that carry a turn in thinking, and the places where text alone is enough. It avoids spreading images evenly and prefers cognitive anchors such as a core judgment, a before-and-after comparison or an input-to-output loop.

After you approve the shot list, the agent generates one image per point. Each image is created separately, so keep the requests focused. If an image fails the checklist, ask for a regeneration or a local edit rather than a full redo.

• Keep articles short: 1-3 images. Long articles: rarely more than 9.
• Each image should show only one core structure.
• Limit Chinese annotations to a few short labels.

Save and organize the output

When you work inside a workspace, the agent copies the final images into a folder named after your article slug, with numbered file names. Original generated files are kept, and existing assets are not overwritten unless you explicitly ask for a replacement.

• Folder pattern: assets/<article-slug>-illustrations/
• File pattern: 01-topic-name.png, 02-topic-name.png
• Your report after generation should list how many images were made, the purpose of each, the save path, and which images are most reliable.

Safety and quality notes

The skill itself runs no scripts and makes no network calls, so its risks are mostly about output quality and what you share with the image tool. Review every image before publishing. Chinese text in generated images can contain errors, and the skill's own checklist treats text errors as a failure signal.

Do not assume an image matches an existing example; the skill is designed to avoid copying earlier compositions unless you ask for one.

• Check every Chinese annotation for typos before publishing.
• Keep only the images that pass the checklist: white background, Xiaohei performing the action, no title text in the top-left corner.
• Do not paste confidential text into a workspace you do not control.