| name | handdrawn-tech-illustrations |
| description | Use when the user asks for Chinese handdrawn technical article illustrations, body images, WeChat covers, Xiaohongshu/Rednote covers, or technical explainer visuals from articles, notes, screenshots, outlines, or rough ideas. |
Handdrawn Tech Illustrations
Create final illustrations for Chinese technical articles and social covers: article body illustrations, concept visuals, WeChat covers, and Xiaohongshu/Rednote covers.
Default production uses the built-in image generation model. Crop, resize, and contact-sheet creation are only delivery helpers after the final image is generated.
Reference Map
Load only what the task needs:
references/style-dna.md: read before choosing a visual mode or writing prompts.references/content-planning.md: read when turning articles or outlines into a shot list.references/prompt-patterns.md: read before image generation.references/platform-specs.md: read when covers or fixed ratios are requested.references/production-workflow.md: read before saving outputs, resizing, or contact sheets.
Use assets/theme-tokens.json as the compact palette/style token file when prompt writing needs exact colors.
Output Modes
- Body illustration: 16:9 Chinese article illustration. Default for 正文配图.
- Article/WeChat wide cover: 21:9 cover. Use for 公众号主封面 or blog/article hero.
- WeChat square cover: 1:1 companion cover. Compose separately for the square surface.
- Xiaohongshu/Rednote cover: 3:4 cover illustration. Fill the portrait canvas with one strong visual hook.
- Shot list only: planning output with where each image goes, core idea, visual approach, text, and ratio.
Do not create editable document packages, reusable layout templates, or generated card systems inside this skill.
Workflow
-
Compress the request
- Decide the output mode, ratio, core idea, useful visible text, and any source assets.
- For articles or multi-image sets, pick only the few points that deserve images.
- If the user asks only for planning, return a shot list and stop.
-
Generate the image
- Use Friendly Handdrawn Tech.
- Read
style-dna.md and prompt-patterns.md before generation.
- Use one image generation call per final image. Put the ratio, main point, visual approach, composition, and Chinese visible text directly in the prompt.
-
Deliver
- Use the user-requested output path when one is provided; otherwise keep the generated image at the model-provided path and report it clearly.
- Copy or organize files only when useful for delivery.
- Resize or create a contact sheet only when useful for delivery.
Defaults
- Language: Simplified Chinese.
- Body illustration ratio: 16:9.
- Article/WeChat wide cover ratio: 21:9.
- WeChat square cover: 1:1, composed separately.
- Xiaohongshu/Rednote cover: 3:4.
- Body illustration text density: 3-6 short labels, notes, or short phrases when needed for clarity.
- Cover text density: concise main text plus 0-3 short labels.
- Background: near-white warm paper.
Non-Negotiables
- Do not copy unrelated card templates, editorial page systems, generated layout frameworks, or nested-card visual language.
- Do not package outputs into editable documents or multi-image deliverable bundles.
- Do not create dense whiteboards, business templates, SaaS dashboards, sticker decorations, random blobs, gradients, or heavy shadows.
- Do not invent facts, product details, metrics, release dates, or percentages.
- Do not let Chinese text become long, wrong, cramped, or unreadable.
- Do not crop supplied screenshots, product UI, code, or hardware details in a way that hides the evidence.
Final Response
Report image paths, image count, and ratios/dimensions when available. Show main images inline with absolute paths when useful.
