article-illustration

name	article-illustration
description	Generate article illustrations in multiple styles — watercolor art, flat illustrations, technical diagrams, infographics, and more. Use when agent needs covers, inset illustrations, atmospheric dividers, blog banners, or technical diagrams for articles. Supports text-only generation, reference-image-plus-text style guidance, and prompt-only dry-run review.

Article Illustration

Use this skill to create illustrations for articles — covers, inset imagery, dividers, banners, and diagrams.

Quick workflow

1. Identify what you need:
   • WeChat cover (公众号头条封面, 2.35:1)
   • Body inset illustration (for literary essays or tech articles)
   • Atmospheric section divider
   • Blog banner / hero
   • Technical diagram (architecture, process, knowledge card)
2. Pick a style profile:
   • auto (default): choose from the brief instead of forcing a house style
   • editorial-atmospheric: textured contemporary editorial illustration, good for essays and cultural observation
   • modern-guochao-editorial: restrained modern guochao / Chinese cultural editorial style
   • cinematic-editorial: film-still inspired editorial image with grounded light and scene detail
   • watercolor-illustration: fine-art watercolor, use only when the article truly wants a soft painterly mood
   • flat-illustration: editorial flat illustration, soft look
   • sketchnote: hand-drawn notebook feel
   • flat-tech-infographic: technical infographic with clean modules, arrows, bilingual labels
   • soft-tech-diagram: subtle dashed containers, light academic style
   • repo-architecture-clean: crisp codebase and ownership diagrams
3. Pick a size:
   • wechat-cover-hd for WeChat headline covers (auto-crops to 1080x460, 2.35:1)
   • portrait-hd for mobile-first inset illustrations
   • blog-banner for blog hero images
   • doc-hd / doc-2k / doc-4k for landscape document illustrations
   • auto for provider-chosen default
4. If the user supplied reference images, treat them as style guidance only.
5. If the user wants to inspect the prompt first, use --dry-run.
6. For live generation, run the bundled script with uv run.

Script

Use uv run for Python scripts:

# Literary essay illustration, auto style selection
uv run scripts/generate_article_illustration.py \
  --title "Snowy Window Scene" \
  --brief "A body inset illustration of snow falling outside a warm-lit classroom window..." \
  --size wechat-cover-hd

# Technical diagram
uv run scripts/generate_article_illustration.py \
  --title "RAG Pipeline Overview" \
  --brief "A technical infographic showing ingestion, chunking, embedding, retrieval..." \
  --style-profile flat-tech-infographic \
  --language zh-en \
  --size doc-hd

Use --mode reference+text --reference-image <path> when the new image should follow the style of one or more reference images.

Use --dry-run when you want the exact prompt and parameters without calling the API.

Default output settings:

• --style-profile auto (default; resolves from the brief)
• --language zh (default; use zh-en for bilingual labels)
• --size auto for best compatibility with GPT image providers
• --quality auto for high-fidelity default behavior

Size presets:

• wechat-cover-hd (1792x1024, auto-crops to 1080x460 2.35:1) — WeChat headline cover
• portrait-hd (1024x1536) — mobile-body inset
• blog-banner (2048x1152) — blog hero / desktop banner
• 9:16 (1024x1792) — full phone portrait
• doc-hd (1536x1024) — landscape document
• doc-2k (2048x1152) — high-res landscape
• doc-4k (3840x2160) — 4K landscape
• auto — provider-chosen default

Authentication

• Read OPENAI_API_KEY from the current shell first.
• Read OPENAI_BASE_URL from the current shell second.
• If OPENAI_API_KEY is missing for a live generation run, prompt temporarily in an interactive shell or tell the user to export the environment variable.
• Do not write credentials to disk.

Model selection

• Use gpt-image-2 for official OpenAI and for https://aihubmix.com/v1.
• Use openai/gpt-image-2 for https://api.ofox.io/v1.
• Default back to gpt-image-2 unless a provider requires a namespaced model name.

References

• Read references/style-profiles.md when selecting or explaining styles.
• Read references/prompt-patterns.md when the requested illustration type needs prompt refinement.
• Browse references/styles/ for ready-to-use style presets (editorial-pencil-sketch, hand-drawn-tech-editorial, technical-diagram, and more). Treat them as creative references, not mandatory templates — adapt or ignore them based on the brief.
• Read references/openai-image-workflow.md when debugging auth, request modes, or output behavior.
• Read references/openai-official-links.md when you need the official OpenAI docs for GPT Image, Images API, or Python examples.