skillZs
★ LIVE SKILL TAGS ★
>>> LIVE SKILLS INDEX <<<
* OPEN SOURCE *
NO LOGIN, NO TRACKING
※ REAL INSTALL DATA ※
← back to all skills
alterlab-ieu/alterlab-academic-skills75 installs

alterlab-scientific-schematics

Creates publication-quality scientific diagrams with no API key - Claude draws the diagram as code (SVG, TikZ/PGF, schemdraw for circuits), renders it locally to PNG/PDF with exact, editable labels, then reviews the render itself against a document-type quality threshold (journal 8.5/10, poster 7.0, ...) and revises until it passes (max 2 iterations). Use when the request is for a rendered technical or scientific diagram — CONSORT/PRISMA flowcharts, neural-network architectures, system/block diagrams, biological pathways, circuits, or other complex scientific visuals. Optional AI raster generation (Nano Banana 2 with Gemini review, via OpenRouter) when OPENROUTER_API_KEY is set. For general photos, illustrations, or artwork use alterlab-generate-image; for infographics use alterlab-infographics; for editable text-based Mermaid diagrams use alterlab-mermaid. Part of the AlterLab Academic Skills suite.

How do I install this agent skill?

npx skills add https://github.com/alterlab-ieu/alterlab-academic-skills --skill alterlab-scientific-schematics
view source ↗

Is this agent skill safe to install?

  • Gen Agent Trust Hubpass

    This skill provides tools for creating publication-quality scientific diagrams (CONSORT/PRISMA flowcharts, neural networks, circuits) by rendering code (SVG, TikZ) locally or using an optional AI backend via OpenRouter. It uses standard local rendering tools and well-known Python libraries.

  • Socketpass

    No alerts

  • Snykpass

    Risk: LOW · No issues

What does this agent skill do?

Scientific Schematics and Diagrams

Overview

Scientific schematics turn complex concepts into clear visuals for publication. This skill draws them as code — SVG by default, TikZ for LaTeX manuscripts, schemdraw for circuits — renders the code locally, then reviews the rendered image itself against a document-type threshold and revises until it passes. No image model and no API key are involved.

Why code: every label and count is exactly the text you typed (an image model redraws text and can change an n=), the figure stays editable, it exports as vector PDF/SVG for the journal, and unpublished content never leaves the machine.

Quality thresholds by document type (Claude's review score, out of 10):

Document TypeThresholdDescription
journal8.5/10Nature, Science, peer-reviewed journals
conference8.0/10Conference papers
thesis8.0/10Dissertations, theses
grant8.0/10Grant proposals
preprint7.5/10arXiv, bioRxiv, etc.
report7.5/10Technical reports
poster7.0/10Academic posters
presentation6.5/10Slides, talks
default7.5/10General purpose

Review criteria, 0-2 points each: scientific accuracy, clarity and readability, label quality, layout and composition, professional appearance. At most 2 iterations.

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

Does NOT Trigger

ScenarioUse Instead
Diagram that should stay editable text in a README or Markdown doc (renders on GitHub, diffs in git)alterlab-mermaid
Photo, illustration, artwork, or slide/poster hero imagealterlab-generate-image
Infographic that presents statistics, a timeline, or a comparison for a general audiencealterlab-infographics
Plot of measured data (scatter, bar, line, heatmap)alterlab-matplotlib / alterlab-scientific-viz
Network graph whose layout is computed from real node/edge dataalterlab-networkx

Hand-off rule. When the request matches a row above, open your answer by naming that skill by its exact id (for example alterlab-mermaid) and saying in one line why it fits better, then stop applying this skill's procedure. Name it even when you also need to ask for missing input. If only part of the request is out of scope, name the skill for that part and do only the in-scope part. A brief general pointer is fine; the named skill does the work.

Workflow (keyless)

  1. Collect the content. List every node, arrow, label and count from the user's message and attached files. Copy numbers verbatim; check that they reconcile (CONSORT arms sum to the randomised total, each stage carries forward). If a required count is missing, ask for it — never invent one.
  2. Get the brief for the venue (guidelines, threshold, criteria): python scripts/generate_schematic.py --brief --doc-type journal
  3. Draw it as code, saved next to the figure (figures/consort.svg):
    • SVG for flowcharts, block diagrams, architectures and pathways. For CONSORT start from assets/consort_template.svg (replace every N_* token); building blocks and recipes for PRISMA, transformers, pathways and block diagrams are in references/README.md.
    • TikZ (standalone .tex) when the paper is LaTeX or labels need maths.
    • schemdraw for circuits (writes SVG); matplotlib when geometry comes from data.
    • Okabe-Ito colours plus a second cue (shape, dash, label); one <text> per line; font size on the root; no "Figure 1" or caption inside the image.
  4. Render and lint, versioned:
    uv run --python ">=3.11" --with cairosvg --with pillow python scripts/generate_schematic.py \
        figures/consort.svg -o figures/consort_v1.png --doc-type journal --width-in 7.2 --gray
    
    The output names the renderer and flags labels below 7 pt at print width, text that probably overflows its box, external references and blank renders. Fix every warning before reviewing.
  5. Review the render yourself. Read consort_v1.png (and consort_v1_gray.png), check each number and label against the source, score the five criteria, and record it:
    python scripts/figure_tools.py score figures/consort_review_log.json --kind schematic \
        --doc-type journal --image figures/consort_v1.png --source figures/consort.svg \
        --score 7.5 --critique "exclusion reasons overflow; arm boxes unequal heights"
    
    REVISE → fix the source and render consort_v2.png; DONE → deliver; STOP (2 iterations used, still below threshold) → deliver the best version and list what is still short. Never inflate a score to pass; the log is the audit trail.
  6. Deliver: the final PNG (300 dpi), a vector PDF/SVG (python scripts/figure_tools.py render figures/consort.svg -o figures/consort.pdf), the source file, and the review log, plus a suggested caption kept outside the image.

Output template

**Figure:** figures/consort_v2.png (300 dpi, 7.2 in) · vector: figures/consort.pdf · source: figures/consort.svg
**Review:** journal threshold 8.5 — v1 7.5 (overflowing exclusion reasons) → v2 9.0, DONE (figures/consort_review_log.json)
**Checked against:** trial_flow_counts.md — 500 assessed, 150 excluded (98/41/11), 350 randomised, 175/175 …
**Caption (for the manuscript):** CONSORT 2010 flow diagram of participant progress through the trial …

Optional: AI raster generation

When OPENROUTER_API_KEY is set (environment, .env, or --api-key), a description can be turned into a raster image by Nano Banana 2 (google/gemini-3.1-flash-image) and reviewed by Gemini 3.1 Pro Preview against the same thresholds:

python scripts/generate_schematic.py "CONSORT participant flow ..." -o figures/consort.png \
    --doc-type journal --backend ai

Without the key, --backend ai exits with status 2 and names the keyless route; it never falls back silently. Use it only when the user asks for an AI-generated image. The prompt leaves the machine (do not include confidential, clinical or unpublished details), and the image model draws every label itself, so compare each count and term with the source before use. If the review call fails the log records "review_skipped": true with a null score. Models can be overridden with ALTERLAB_IMAGE_MODEL / ALTERLAB_REVIEW_MODEL. Details: references/ai_generation_guide.md; worked prompts: references/generation_examples.md.

Best Practices Summary

Design for clarity over complexity, consistent styling, colorblind accessibility (Okabe-Ito, redundant encoding), ≥7-8 pt sans-serif type at print size, and vector output (PDF/SVG; 300+ DPI for raster). Integrate with LaTeX via \includegraphics{}, caption thoroughly, reference in text, and keep the figure source + outputs + review log under version control. The full design/technical/integration checklist is in references/submission_checklist.md.

Troubleshooting Common Issues

Renderer missing, overflowing text, small labels, detached arrowheads, LaTeX errors, and the optional AI backend's setup and generation problems are covered in references/troubleshooting.md and at the end of references/README.md.

Resources and References

Detailed References

  • references/README.md - Code-drawn workflow: choosing SVG/TikZ/schemdraw, SVG building blocks, CONSORT/PRISMA/transformer/pathway/block-diagram recipes, review loop
  • references/QUICK_REFERENCE.md - Commands, thresholds, print sizes, delivery checklist
  • references/diagram_types.md - Catalog of scientific diagram types with examples
  • references/best_practices.md - Publication standards and accessibility guidelines
  • references/submission_checklist.md - Pre-submission verification checklist
  • references/troubleshooting.md - Fixes for rendering, quality, accessibility, and AI-backend issues
  • references/ai_generation_guide.md - Optional AI backend: smart iteration, review rubric, Python API, CLI options
  • references/generation_examples.md - Optional AI backend: worked prompts

External Resources

Python Libraries

Publication Standards

Integration with Other Skills

This skill works synergistically with:

  • alterlab-scientific-writing - Diagrams follow figure best practices
  • alterlab-scientific-viz - Shares color palettes and styling
  • alterlab-latex-posters - Generate diagrams for poster presentations
  • alterlab-research-grants - Methodology diagrams for proposals
  • alterlab-peer-review - Evaluate diagram clarity and accessibility

Quick Reference Checklist

Before submitting diagrams, run through the full checklist (visual quality, accessibility, typography, publication standards, required quality verification, documentation/version control, and final integration) in references/submission_checklist.md.


Use this skill to create clear, accessible, publication-quality diagrams that effectively communicate complex scientific concepts: drawn as code, rendered locally, and reviewed against the venue's threshold before delivery.

Part of the AlterLab Academic Skills suite.

Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.

<a href="https://skillzs.dev/skills/alterlab-ieu/alterlab-academic-skills/alterlab-scientific-schematics">View alterlab-scientific-schematics on skillZs</a>