dvampire/autogenesis/figures-diagram

Flowchart and structure-diagram generation

Recognize requests for flowcharts, structure diagrams, and concept diagrams, then produce a professional drawing prompt for an image-generation AI.

1. When to use

Diagram types handled here

Diagrams that do NOT belong here

Use the figures-python skill for:

• Line charts, bar charts, pie charts, and other data charts
• Scatter plots, heatmaps, and other statistical plots
• Any chart that requires exact numeric values

2. Checklist

• [ ] Confirm this is a flow/structure diagram, not a data chart
• [ ] Collect the diagram information (purpose, elements, relationships)
• [ ] Confirm style preferences and label language
• [ ] Generate the prompt
• [ ] Recommend an image-generation tool

3. Prompt templates

3.1 General template

Draw a publication-quality [diagram type] for an academic paper, showing "[topic]".
 
[Overall layout]
- Direction: [top-to-bottom / left-to-right / radial from center]
- Background: white
- Style: professional academic
- Resolution: 450 DPI
 
[Content]
[Describe every element and relationship in detail]
 
[Color scheme]
- [Color 1]: [what it marks]
- [Color 2]: [what it marks]
 
[Style requirements]
- All text clearly legible
- Arrow direction unambiguous
- Add the necessary annotations

3.2 Flowchart template

Draw a publication-quality flowchart for an academic paper, showing "[process name]".
 
[Process steps]
Step 1: [name]
- Box color: [color]
- Content: [content]
 
Step 2: [name]
- Relationship to step 1: [arrow description]
...
 
[Color scheme]
- Blue  (#4E79A7): data-processing steps
- Orange (#F28E2B): core algorithm steps
- Green  (#59A14F): output / result steps
 
[Style requirements]
- Rounded rectangles
- Arrowed connectors
- Bold borders on key steps

3.3 Neural-network architecture template

Draw a publication-quality neural-network architecture diagram for an academic paper,
showing "[model name]".
 
[Network structure]
Input layer: [shape, how it is drawn]
 
Hidden layer 1: [layer type]
- Parameters: [kernel size, channel count, ...]
- Color: [color]
 
[continue with further layers...]
 
Output layer: [shape, activation function]
 
[Color scheme]
- Blues:   convolution layers
- Greens:  pooling layers
- Oranges: fully connected layers
- Purples: attention layers
 
[Annotation requirements]
- Label every layer with its name and parameters
- Label input and output dimensions
- Mark the novel components with a red dashed box

Set the label language (English or Chinese) explicitly in the prompt so the generated diagram matches the manuscript.

4. Recommended image-generation tools

Guidance:

• Prefer Nano Banana: designed for academic figures
• Use DALL-E 3 for complex structures: it follows long descriptions better
• Use Midjourney when you need to iterate: it supports variant generation

5. Color references

Nature / Science style

Blue:      #2E86AB
Magenta:   #A23B72
Orange:    #F18F01
Brick red: #C73E1D
Yellow-green: #95C623

Minimal academic style

Dark blue:  #1D3557
Mid blue:   #457B9D
Light blue: #A8DADC
Off-white:  #F1FAEE
Brick red:  #E63946

How to apply the colors

• Input / output: blues
• Core processing: oranges, with a bold border
• Auxiliary steps: grays
• Novelty markers: red dashed lines
• Success / completion: greens

6. Conversation flow

When the user asks for a flowchart

"I'll generate the drawing prompt for you. First, a few questions:

1. Purpose: what should this flowchart show?

2. Main steps: which key steps does it include?

3. Step relationships: is it linear, or are there branches and loops?

4. Label language: English or Chinese labels?

5. Target venue: what kind of journal are you submitting to?"

After generating the prompt

"Here is the drawing prompt — copy it into your image-generation AI:

---

[prompt text]

---

Recommended tool: Nano Banana

If the result needs adjusting, tell me what to change."