# scientific-schematics

> Create publication-quality scientific diagrams using AI with smart iterative refinement and quality review — specialized in neural network architectures, system diagrams, flowcharts, biological pathways, and complex scientific visualizations.

**Use case**: Generate publication-quality scientific diagrams with AI

**Canonical URL**: https://agentcookbooks.com/skills/scientific-schematics/

**Topics**: claude-code, skills, science, scientific-writing

**Trigger phrases**: "create a scientific diagram", "draw this pathway", "generate a schematic", "make a flowchart for this method", "visualize this neural network"

**Source**: [K-Dense AI](https://github.com/K-Dense-AI/scientific-agent-skills/tree/main/scientific-skills/scientific-schematics)

**License**: MIT

---

## What it does

`scientific-schematics` is a Claude Code skill from K-Dense AI's [scientific-agent-skills repo](https://github.com/K-Dense-AI/scientific-agent-skills). It turns Claude into a scientific diagram generator that produces publication-quality figures using AI image generation with iterative quality refinement — specialized in neural network architectures, system diagrams, flowcharts, biological pathways, and other complex scientific visualizations that are hard to produce in standard plotting libraries.

A session produces a PNG or vector-format figure: the schematic is generated, reviewed against a quality threshold for the target document type (journal figure, poster, graphical abstract), and regenerated if quality is below threshold — only then is it returned.

## When to use it

Reach for it when:

- You need a graphical abstract for a manuscript (the `scientific-writing` skill calls this automatically)
- You want a clean flowchart or biological pathway diagram that would take hours in Inkscape or PowerPoint
- You're illustrating a novel neural network architecture or system design for a paper

When *not* to reach for it:

- Data plots (scatter, bar, histogram, heatmap) — use `matplotlib`, `seaborn`, or `scientific-visualization`
- General-purpose photos or illustrations — use `generate-image`

## Install

Copy the `SKILL.md` from K-Dense AI's [scientific-schematics folder](https://github.com/K-Dense-AI/scientific-agent-skills/tree/main/scientific-skills/scientific-schematics) into `.claude/skills/scientific-schematics/` in your project. Requires an OpenRouter API key for the AI image generation backend.

Trigger phrases: "create a scientific diagram", "draw this pathway", "generate a schematic", "make a flowchart for this method".

## What a session looks like

A typical session has three phases:

1. **Description and target.** Describe the diagram content and the target context (journal figure, poster, graphical abstract). Claude selects the appropriate output dimensions and style parameters.
2. **Generation and review.** The diagram is generated and reviewed against a quality rubric. If it falls below threshold, Claude iterates with refined prompts — you don't have to babysit the loop.
3. **Output delivery.** The accepted figure is saved to the specified output path with a recommended filename. For graphical abstracts, the aspect ratio (typically 1200×600px) is enforced automatically.

## Receipts

**Where it works well:**
- Graphical abstracts for manuscripts — the workflow/result summary format is well-suited to AI image generation, and the quality loop catches obvious failures
- Conceptual diagrams where artistic precision matters less than communicating a relationship or process clearly

**Where it backfires:**
- Diagrams requiring exact molecular structures, precise circuit layouts, or dimensioned technical drawings — AI image generation produces plausible-looking but not verified structures
- Iterative refinement adds API cost; very complex diagrams may cycle through multiple generations before meeting the threshold

**Pattern that works:** write a dense, specific natural-language description of the diagram rather than a vague prompt; more specificity produces higher first-pass quality and fewer regeneration cycles.

## Source and attribution

Originally authored by [K-Dense Inc.](https://github.com/K-Dense-AI). The canonical SKILL.md lives in the [`scientific-schematics` folder](https://github.com/K-Dense-AI/scientific-agent-skills/tree/main/scientific-skills/scientific-schematics) of their public scientific-agent-skills repository.

License: MIT. Install, adapt, and redistribute with attribution preserved.

This page documents the skill from a practitioner's perspective. For the formal spec and any updates, defer to the source repo.