Your Chinese blog post, drawn by a blank-faced stick figure

What it does Ian Xiaohei Illustrations is a Codex Skill — a structured prompt-and-workflow package — that reads Chinese articles, blog posts, or Notion docs and generates 16:9 white-background PNG illustrations. It does not decorate; it translates a single cognitive anchor — a judgment, flow, state, or metaphor — into a hand-drawn visual explanation. The default visual identity is “Xiaohei,” a solid-black, blank-faced, thin-legged figure who must actively participate in the scene rather than stand in the corner like a mascot.

The interesting bit The skill treats illustration as argument, not ornament. It forces the agent to first extract a “cognitive anchor” from the text, invent a low-tech physical metaphor, and only then generate the image — a workflow designed to prevent the usual AI drift into generic clip-art. The aesthetic is deliberately anti-PPT: heavy whitespace, jittery thin black lines, and sparse red-orange-blue Chinese annotations on a stark white field.

Key highlights

• Outputs a shot list of 4–8 images per article, each focused on exactly one structural idea such as workflow, contrast, or metaphor.
• Enforces a strict visual IP: Xiaohei must be central to the action; if removing him leaves the composition intact, the prompt has failed.
• Includes a QA checklist to catch style drift, wrong Chinese characters, and accidental PPT-ification.
• Ships as a Codex Skill with reference documents covering style DNA, composition patterns, and prompt templates.

Caveats

• AI image models frequently hallucinate or misspell Chinese text; the README explicitly warns that shorter labels are more stable and typos may require regeneration.
• It is not a general-purpose illustration tool: the README rules out commercial brand work, complex architecture diagrams, children’s cartoons, and editable vector output.
• The example images are strictly for style calibration; the skill expects fresh metaphors invented per article, not copied compositions.

Verdict Worth a look if you publish Chinese knowledge-work content and want a consistent, lightly grotesque visual voice without touching PowerPoint. Skip it if you need polished brand illustrations, SVG source files, or diagrams dense with text and arrows.