Scientific Schematics and Diagrams

Scientific Schematics is the diagram engine behind K-Dense's Claude Scientific Writer (MIT). Other skills in the pack, such as literature reviews, grants, and posters, call it to produce figures. You describe a diagram in plain language; the skill generates it with Nano Banana 2, has Gemini 3.6 Flash review it, and regenerates only if the review score is below the threshold for your document type. Earlier listings named Nano Banana Pro and Gemini 3 Pro; version 1.4 uses the newer models.

Key Features

• Document-type thresholds: journal 8.5/10, conference, thesis, and grant 8.0, preprint and report 7.5, poster 7.0, presentation 6.5. Higher bars cost more attempts but aim for cleaner output.
• Smart iteration: at most two iterations. If the first image passes, it stops, which saves API calls on simple figures. A review log records scores, critiques, and early stops, and outputs are versioned.
• Scientific defaults: white background, high contrast, labels of at least 10 pt, sans-serif type, and the colorblind-friendly Okabe-Ito palette are applied in every prompt.
• Specialties: CONSORT and PRISMA flowcharts, neural network architectures such as Transformers, biological pathways, circuit diagrams, system and IoT block diagrams, and conceptual frameworks.

Use Cases

• A CONSORT participant-flow diagram for a clinical manuscript.
• A Transformer encoder-decoder figure for an ML paper or slides.
• A signaling-pathway graphic for a poster, generated at the lower poster threshold.

Pricing and Access

The skill is free (MIT). Every generation and review call goes through OpenRouter using your OPENROUTER_API_KEY, so cost depends on OpenRouter's current pricing for those models and on how many iterations run.

Getting Started

1. Install the plugin in Claude Code: /plugin marketplace add https://github.com/K-Dense-AI/claude-scientific-writer, then /plugin install claude-scientific-writer.
2. export OPENROUTER_API_KEY='your_api_key_here' (keys at openrouter.ai/keys).
3. Run python scripts/generate_schematic.py "CONSORT participant flow diagram with 500 screened, 150 excluded, 350 randomized" -o figures/consort.png --doc-type journal.

Limitation: the output is a raster image from a generative model, so labels, arrows, and numbers can be wrong even after a passing review. The skill's own best-practice list recommends vector PDF or SVG for publication, which this workflow does not produce. Always proof the figure against your data.

Frequently Asked Questions

Does it need coding or templates?

No. A specific text prompt is the input; the skill's guide shows what a good prompt includes (type, components, flow, labels).

What models does it use?

The scripts default to google/gemini-3.1-flash-image for generation and google/gemini-3.6-flash for review, both via OpenRouter.

Alternatives

• Excalidraw Diagram Skill: editable hand-drawn style diagrams.
• Mermaid Visualizer Skill: text-defined diagrams that stay exactly as written.
• Generate Image Skill: general illustrations and photos rather than technical figures.

Conclusion

Scientific Schematics is a fast way to get a clean draft figure when drawing tools feel like overkill. For figures where every label must be exact, use a deterministic tool like Mermaid, or redraw the AI draft. Browse more tools in the skills hub.