Xiaohei Hand-Drawn Illustrations Skill
Turn judgments, flows, states, and metaphors in Chinese articles into 16:9 white-background hand-drawn quirky illustrations.
Evidence: The repository contains no code or executable files, only documentation and example images, so there are no signs of permission requests, data collection, or external interactions. Deductions: With no code, least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, and source attribution cannot be assessed; all criteria are unmet.
Evidence: The workflow and outputs described in README and SKILL.md are consistent, and example images match the style description. Deductions: Dependencies are not explicitly listed, only mentioning image models and Codex; failure messages are not provided, only QA checks are mentioned.
Evidence: README clearly states target audience, suitable and unsuitable scenarios, and provides multiple usage examples. Deductions: Environment fit is only for Codex, not other platforms; trigger precision is not explicitly defined, but examples provide trigger methods.
Evidence: Directory structure is clear, installation instructions are detailed, examples and FAQ exist, known limitations are listed in notes, and license is MIT. Deductions: No versioning or changelog, maintenance responsibility is only implied via author info, not explicitly stated.
Evidence: Output is PNG images, directly usable, examples show effect; marginal value lies in unique visual style and workflow. Deductions: Cost-benefit not quantified, no cost or performance data.
Evidence: Claims in README are consistent with example images and file structure, but no external verification. Deductions: No cross-source corroboration, facts and inferences not clearly separated.
- Repository contains only documentation and examples, no actual code, so functionality cannot be verified.
- Dependencies are not explicitly listed, which may affect reproducibility.
- No versioning or changelog provided, maintenance status unclear.
What does this agent do, and when should you use it?
Ian Xiaohei Illustrations is a Codex Skill that guides an AI agent to generate in-article illustrations for Chinese articles, posts, blogs, and Notion documents. It uses a default visual IP called 'Xiaohei': a solid black, white-dotted-eye, thin-legged, expressionless character. The style features a pure white background, black hand-drawn line art, generous whitespace, and sparse red/orange/blue handwritten Chinese annotations. The core methodology is to first extract cognitive anchors from the article, then choose a single judgment, process, structure, or metaphor per image, invent a low-tech, quirky but plausible physical metaphor, and make Xiaohei actively participate in the core action. Output is a 16:9 landscape PNG saved to the workspace's assets/<article-slug>-illustrations/ directory. The skill itself lives in the subdirectory ian-xiaohei-illustrations/, containing SKILL.md, agents/openai.yaml, and references/ with style guides and QA checklists.
The skill directs Codex to: read provided articles, Markdown, Notion content, or user topics; distill key insights and visualizable sections; output a shot list with theme, core meaning, structure type, Xiaohei's action, and suggested Chinese annotations; choose a structure type (e.g., Workflow, system partial, before/after contrast, conceptual metaphor); invent a physical metaphor; call an image model to generate 16:9 pure-white-background PNGs; finally verify against a QA checklist and save to assets/<article-slug>-illustrations/. It is triggered by 'Use $ian-xiaohei-illustrations' in Codex.
- A Chinese blogger writing articles needs consistent, memorable illustrations for each key section.
- A creator producing methodology content wants to visualize abstract concepts like 'trust is paved with evidence' as concrete metaphors.
- A user leveraging Codex for content production wants to reuse a consistent hand-drawn visual language without redesigning each time.
- An editor wants to plan 4-8 illustration ideas (shot list) before generating images, to review placements and themes.
- An author prefers a quirky, clean illustration style over PPT-style infographics, wanting a light, recognizable look.
What are this agent's strengths and limitations?
- Highly customized visual style with detailed style guides (SKILL.md, references/style-dna.md) and Xiaohei IP specifications, ensuring consistency.
- Focused on Chinese article illustrations, with optimization for Chinese text stability (recommends short text).
- Provides a structured workflow and QA checklist to reduce common image-generation issues (typos, hallucinated labels).
- Works only on Codex platform; not directly usable in other agent environments.
- Requires manual installation into Codex skills directory, involving cloning and copying steps.
- Image generation relies on external image models, may produce typos or style drift, and output is uneditable PNG.
How do you install or deploy this agent?
Clone the repository: git clone https://github.com/helloianneo/ian-xiaohei-illustrations.git; enter the directory: cd ian-xiaohei-illustrations; create the skills directory and copy the subdirectory: mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills" && cp -R ./ian-xiaohei-illustrations "${CODEX_HOME:-$HOME/.codex}/skills/". Requires Codex installed and CODEX_HOME set or default path.
How do you use this agent?
After installation, invoke in Codex: 'Use $ian-xiaohei-illustrations to generate 4 Xiaohei quirky illustrations for the following article.' Or for planning only: 'Use $ian-xiaohei-illustrations without generating images. Analyze where this article deserves illustrations and output a shot list of about 5 images.' Generated images are saved to the workspace's assets/<article-slug>-illustrations/ directory.
How does this agent compare with similar options?
The README mentions related projects: Ian Handdrawn PPT (hand-drawn PPT-style page generation skill), Awesome Claude Code Skills (curated Claude Code skills list), and Obsidian + Claude AI Second Brain (knowledge base guide), but does not explicitly compare alternatives.