scientific-schematics_skill

Overview

This skill creates publication-quality scientific diagrams, flowcharts, and schematics using Python tooling (graphviz, matplotlib, schemdraw, networkx). It focuses on neural network architectures, system diagrams, and standard methodological flowcharts, producing vector outputs (SVG/EPS/PDF) saved in a figures/ folder with automated quality verification. Diagrams are generated as standalone vector files ready to be referenced in papers and posters.

How this skill works

I inspect your diagram requirements (type, components, layout direction, and cross-connections), pick the optimal library, and generate a vector graphic with careful spacing and styling. Files are written to figures/ with descriptive names and rendered in multiple formats (PDF, SVG, EPS) where applicable. A lightweight quality check verifies layout issues (overlaps, contrast, resolution) before you insert the figure into your document.

When to use it

• Designing neural network architecture diagrams (Transformers, CNNs, RNNs, attention maps).
• Drawing system architecture, data flow, or cloud topology diagrams for papers or talks.
• Creating methodology flowcharts (CONSORT, PRISMA, data-processing pipelines).
• Producing circuit and electrical schematics with schemdraw.
• Visualizing biological pathways, gene/protein interaction networks, or network topologies.

Best practices

• Generate standalone vector files first (SVG/EPS/PDF) and reference them from LaTeX rather than embedding images inline.
• Choose graphviz for automatic layout; use matplotlib/schemdraw for fine-grained custom control when needed.
• Name files descriptively and store them in figures/ to keep manuscript assets organized.
• Set rankdir and clusters to reflect hierarchy or sequence; prefer orthogonal splines for clean arrow routing.
• Run automated quality checks: overlap detection, color contrast, and intended export resolution before finalizing.

Example use cases

• Produce a Transformer encoder-decoder figure rendered as figures/transformer_architecture.pdf, .svg, and .eps for a conference paper.
• Generate a CONSORT-style participant flowchart saved as figures/consort_flowchart.pdf for a clinical methods section.
• Create a CNN layer-stack diagram for a model architecture figure in a journal submission.
• Build circuit schematics with schemdraw and export EPS for high-fidelity reproduction in print.
• Visualize a signaling pathway as a networkx graph exported to SVG for interactive web preview and PDF for print.