Generate high-quality, distinctive thematic illustrations for technical articles, without the AI plastic feel.
Core Principles
Reject typical AI ailments:
- Prohibited: plastic feel, light pollution, glowing brains, neon overload, over-rendering
- Pursue: physical texture, photographic realism, the purity of a specific art movement
Workflow
Step 1: Analyze → Choose style → Write prompt
- Extract core elements
- Article theme (technical field, specific product/concept)
- Emotional tone (objective analysis/humanistic narrative/hardcore depth/future outlook/philosophical reflection)
- Visual imagery (the core concepts in the article that can be visualized)
-
Match style Choose according to the emotional tone, see details in references/styles.md：
- Rationality and precision → Swiss design, isometric, patent, plotter, ASCII
- Humanities and narrative → stencil printing, tech noir comic, Da Vinci manuscript, collage, printmaking
- Industrial power → brutalism, stippling, X-ray, tiles, cyanotype
- Digital native → 1-Bit pixel, glitch art, low poly, synthwave, LiDAR
- Aesthetics and vision → thermal imaging, paper sculpture, solarpunk, Bauhaus, modern ink wash
-
Construct prompt
[风格名称 + 核心视觉特征]. [文章核心意象的具体描述]. [材质/纹理细节]. [色彩方案]. No text, no watermark, no plastic texture, no lens flare. -
Output for user confirmation
- Recommended style (Chinese and English names)
- Style introduction (1-2 sentences explaining what this style is)
- Reason for selection (why it suits this article)
- English prompt
- Chinese prompt
Step 2: Generate image
Environment detection and execution:
Case A: MCP image-generation tool available
If the mcp-image:generate_image tool exists in the environment, call it directly after user confirmation:
{
"name": "mcp-image:generate_image",
"parameters": {
"prompt": "[英文提示词]",
"aspectRatio": "16:9",
"imageSize": "2K",
"purpose": "Tech article hero image"
}
}
Case B: No MCP image-generation tool
Output the complete prompt (bilingual Chinese-English) + prompt structure explanation, in the following format:
📷 Generation prompt
English version (recommended for Midjourney / DALL-E / Stable Diffusion):
[完整英文提示词]
Chinese version (for reference understanding):
[中文翻译]
🔍 Prompt structure analysis:
| Component | Content | Function |
|---|---|---|
| Style declaration | [e.g.: Algorithmic plotter art] | Locks in the overall visual language, avoiding AI's default "generic tech style" |
| Core imagery | [e.g.: neural network topology] | Turns the article's abstract concepts into concrete visualizable imagery |
| Material/texture | [e.g.: fine black ink lines on textured white paper] | Gives a physical texture, eliminating the "plastic feel" of AI images |
| Atmosphere word | [e.g., Mathematical precision, computational beauty] | Fine-tune the mood to guide the model's understanding of the visual character |
| Negative constraints | [e.g., No text, no watermark, no plastic texture] | Eliminate common AI image issues (light pollution, watermarks, fake text) |
💡 Suggested modifications:
- Change style: Replace the "style statement" section; refer to the 25-style library
- Change imagery: Modify the "core imagery" to fit the specific content of your article
- Adjust atmosphere: Add or remove "atmosphere words" to change the image mood (e.g., add "dramatic contrast" to increase impact)
🛠 Suggested image generation tool parameters:
| Tool | Recommended settings |
|---|---|
| Midjourney | --ar 16:9 --v 6 --style raw |
| DALL-E | Landscape, HD mode, Natural style |
| Stable Diffusion | CFG 7-9, Steps 30+ |
💡 Tip: If you need to generate images directly in Claude, you can configure an MCP tool that supports image generation. After configuration, using this skill again will automatically call image generation, with no need to manually copy the prompt.
Style quick reference table
| Mood | Recommended style | Typical scenarios |
|---|---|---|
| Strategic analysis | Swiss design, Bauhaus | Industry trends, macro reports |
| Technical architecture | Isometric, X-ray | System design, in-depth breakdown |
| Hardware review | Patent, flat lay, X-ray | Chip analysis, supply chain |
| Algorithm principles | Plotter, modern ink wash | Mathematical models, neural networks |
| Code/security | ASCII, printmaking | Geek culture, cybersecurity |
| Industry observation | Stencil printing, collage | Warm analysis, conflict |
| Future prediction | Tech noir comics, synthwave | Autonomous driving, future cities |
| Concept innovation | Da Vinci manuscripts | From 0 to 1, prototype design |
| Infrastructure | Brutalism, cyanotype | Data centers, cloud computing |
| Character close-up | Pointillism | CEO interviews, business leaders |
| Metaverse/VR | Low poly, LiDAR | Spatial computing, digital twins |
| System failure | Glitch art | AI hallucinations, bug analysis |
| Energy efficiency/heat | Thermal imaging | Cooling reviews, market heat |
| Hierarchical structure | Papercraft | Software stacks, databases |
| Green tech | Solarpunk | Carbon neutrality, sustainability |
| Eastern strategy | Modern ink wash | Internet of Everything, Zen tech |
Aspect ratio selection
| Purpose | Recommended ratio | Description |
|---|---|---|
| WeChat Official Account cover | 16:9 | Standard landscape |
| Website article header image | 16:9 or 21:9 | Wider format has more impact |
| Social Media | 1:1 | Instagram, Weibo |
| Vertical/Mobile | 9:16 | Stories, mobile wallpapers |
| PPT Cover | 16:9 | Presentation standard |
Detailed Style Reference
For the complete descriptions of all 25 styles, keywords, use cases, and representative prompt templates, see references/styles.md。
Full Example
Input: A technical article about large model inference optimization
First-step output:
Recommended style: Algorithmic Plotter Art
Style introduction: Art that simulates a mechanical plotter driving a pen tip with mathematical formulas. It features precise interlacing of fine lines, moiré interference patterns, and the physical texture of pure black-and-white ink lines on paper.
Reason for selection: Inference optimization is essentially the ultimate pursuit of mathematics and algorithms. The mathematical beauty of the plotter style and its moiré visuals perfectly match the core concept of "optimization."
English prompt:
Algorithmic plotter art. A generative art piece featuring dense intersecting fine black ink lines forming a neural network topology with flowing optimization gradients, creating geometric moiré interference patterns on textured white paper. Mathematical precision, computational beauty. No text, no watermark, no plastic texture, no lens flare.Chinese prompt:
算法绘图仪艺术。一幅生成艺术作品，由密集交叉的精细黑色墨线构成神经网络拓扑结构，流动的优化梯度形成几何摩尔纹干涉图案，绘于有纹理的白纸上。数学的精确，计算的美感。无文字，无水印，无塑料质感，无镜头光晕。Once confirmed, I will generate the image for you (or provide detailed image generation tool parameters).
User confirmation → Second-step output (when there is no MCP):
📷 Generation Prompt
English version:
Algorithmic plotter art. A generative art piece featuring dense intersecting fine black ink lines forming a neural network topology with flowing optimization gradients, creating geometric moiré interference patterns on textured white paper. Mathematical precision, computational beauty. No text, no watermark, no plastic texture, no lens flare.
Chinese version:
算法绘图仪艺术。一幅生成艺术作品，由密集交叉的精细黑色墨线构成神经网络拓扑结构，流动的优化梯度形成几何摩尔纹干涉图案，绘于有纹理的白纸上。数学的精确，计算的美感。无文字，无水印，无塑料质感，无镜头光晕。
🔍 Prompt Structure Analysis:
| Component | Content in this example | Function |
|---|---|---|
| Style declaration | Algorithmic plotter art | Locks in the "Algorithmic Plotter Art" visual language |
| Core imagery | neural network topology with flowing optimization gradients | Transforms "inference optimization" into "network topology + gradient flow" |
| Material/Texture | fine black ink lines on textured white paper | Ink lines + paper texture, eliminating the digital plastic feel |
| Visual effect | geometric moiré interference patterns | Moiré enhances mathematical beauty |
| Atmosphere words | Mathematical precision, computational beauty | Emphasize a rational, precise emotional tone |
| Negative constraints | No text, no watermark, no plastic texture, no lens flare | Exclude common AI image issues |
💡 Modification suggestions:
- If the article leans more toward "performance improvement," the imagery can be changed to "accelerating data streams through optimized pathways"
- If you want it more abstract, remove "neural network" and keep purely the interlacing of lines
🛠 Image generation tool parameters:
- Midjourney:
--ar 16:9 --v 6 --style raw - DALL-E: landscape, HD, Natural style