Scientific SchematicsCAUTION

Kosmos: An AI Scientist for Autonomous Discovery - An implementation and adaptation to be driven by Claude Code or API - Based on the Kosmos AI Paper - https://arxiv.org/abs/2511.02824

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: scientific-schematics
description: "Create publication-quality scientific diagrams, flowcharts, and schematics using Python (graphviz, matplotlib, schemdraw, networkx). Specialized in neural network architectures, system diagrams, and flowcharts. Generates SVG/EPS in figures/ folder with automated quality verification."
---

# Scientific Schematics and Diagrams

## Overview

Scientific schematics and diagrams transform complex concepts into clear visual representations for publication. Generate neural network architectures, flowcharts, circuit diagrams, biological pathways, and system diagrams using best-in-class Python libraries. **All diagrams are created as SVG/EPS files, stored in the figures/ subfolder, and referenced in papers/posters** - never embedded directly in LaTeX.

## Zero-Shot Diagram Generation Workflow

**Standard workflow for ALL diagrams:**

1. **Analyze requirements** - Identify diagram type and components
2. **Choose optimal library** - Select best tool for the specific diagram type
3. **Generate vector graphic** - Create SVG/EPS with proper spacing and layout
4. **Store in figures/** - Save to `figures/` subfolder with descriptive name
5. **Run quality checks** - Verify no overlaps, good contrast, proper resolution
6. **Reference in document** - Use `\includegraphics{figures/diagram_name.pdf}` in LaTeX

**Key principle:** Generate standalone vector graphics first, then integrate into documents.

## When to Use This Skill

This skill should be used when:
- Creating neural network architecture diagrams (Transformers, CNNs, RNNs, etc.)
- Illustrating system architectures and data flow diagrams
- Drawing methodology flowcharts for study design (CONSORT, PRISMA)
- Visualizing algorithm workflows and processing pipelines
- Creating circuit diagrams and electrical schematics
- Depicting biological pathways and molecular interactions
- Generating network topologies and hierarchical structures
- Illustrating conceptual frameworks and theoretical models
- Designing block diagrams for technical papers

## Best Libraries by Diagram Type

Choose the optimal library for your specific diagram type:

### Neural Network Architectures (Transformers, CNNs, etc.)
**Best library:** `graphviz` via Python's `pygraphviz` or `pydot`
- Excellent automatic layout algorithms
- Clean, professional appearance
- Perfect for layer stacks and connections
- Handles complex cross-connections well

**Alternative:** Custom `matplotlib` with careful positioning
- More control over exact placement
- Better for highly customized designs
- Requires more manual positioning

### Flowcharts and Process Diagrams
**Best library:** `graphviz` with `dot` or `flowchart` layout
- Automatic optimal positioning
- Standard flowchart shapes
- Clean arrow routing
- Minimal overlap issues

**Alternative:** `diagrams` library (for cloud/system architecture style)

### Circuit Diagrams
**Best library:** `schemdraw`
- Purpose-built for electrical circuits
- Extensive component library
- Automatic wire routing
- Professional engineering standard output

### Biological Pathways
**Best library:** `networkx` with custom rendering
- Graph-based pathway representation
- Algorithm-driven layout
- Flexible node/edge styling

### Block Diagrams and System Architecture
**Best library:** `graphviz` or `diagrams`
- Clean hierarchical layouts
- Automatic spacing
- Professional appearance

## Zero-Shot Examples for Common Diagram Types

### Example 1: Transformer Architecture (Neural Network)

Creating a Transformer encoder-decoder diagram like in "Attention Is All You Need":

```python
import graphviz
from pathlib import Path

def create_transformer_diagram(output_dir='figures'):
    """
    Create a Transformer architecture diagram.
    Zero-shot generation with automatic layout.
    """
    Path(output_dir).mkdir(exist_ok=True)
    
    # Create directed graph with TB (top-to-bottom) layout
    dot = graphviz.Digraph(
        'transformer',
        format='pdf',
        graph_attr={
            'rankdir': 'BT',  # Bottom to top (like the original paper)
            'splines': 'ortho',  # Orthogonal edges
            'nodesep': '0.5',
            'ranksep': '0.8',
            'bgcolor': 'white',
            'dpi': '300'
        },
        node_attr={
            'shape': 'box',
            'style': 'rounded,filled',
            'fillcolor': 'lightgray',
            'fontname': 'Arial',
            'fontsize': '11',
            'width': '2.5',
            'height': '0.5'
        },
        edge_attr={
            'color': 'black',
            'penwidth': '1.5'
        }
    )
    
    # ENCODER STACK (left side)
    with dot.subgraph(name='cluster_encoder') as enc:
        enc.attr(label='Encoder', fontsize='14', fontname='Arial-Bold')
        enc.attr(style='rounded', color='blue', penwidth='2')
        
        # Encoder layers (bottom to top)
        enc.node('enc_input_emb', 'Input Embedding', fillcolor='#E8F4F8')
        enc.node('enc_pos', 'Positional Encoding', fillcolor='#E8F4F8')
        enc.node('enc_mha', 'Multi-Head\nAttention', fillcolor='#B3D9E6')
        enc.node('enc_an1', 'Add & Norm', fillcolor='#CCE5FF')
        enc.node('enc_ff', 'Feed Forward', fillcolor='#B3D9E6')
        enc.node('enc_an2', 'Add & Norm', fillcolor='#CCE5FF')
        
        # Encoder flow
        enc.edge('enc_input_emb', 'enc_pos')
        enc.edge('enc_pos', 'enc_mha')
        enc.edge('enc_mha', 'enc_an1')
        enc.edge('enc_an1', 'enc_ff')
        enc.edge('enc_ff', 'enc_an2')
    
    # DECODER STACK (right side)
    with dot.subgraph(name='cluster_decoder') as dec:
        dec.attr(label='Decoder', fontsize='14', fontname='Arial-Bold')
        dec.attr(style='rounded', color='red', penwidth='2')
        
        # Decoder layers (bottom to top)
        dec.node('dec_output_emb', 'Output Embedding', fillcolor='#FFE8E8')
        dec.node('dec_pos', 'Positional Encoding', fillcolor='#FFE8E8')
        dec.node('dec_mmha', 'Masked S

CAUTIONgrade B · trust 89/100 Install with care. The audit found things worth knowing before you trust its output.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (3)

builder.add_resistor('R', '1kΩ', 'right')

builder.add_resistor('$R_1$', '1kΩ', 'right')

builder.add_resistor('$R_2$', '2kΩ', 'down')

Gates applied: no_behavioural_pass.