nature-figure

Name: nature-figure
Author: mubuky

github.com/mubukyDesign & UI136 installsUpdated 13w ago

About nature-figure

A skill for creating evidence-based scientific figures for researchers and scientific communicators. It emphasizes clear visual arguments, faithful data representation, backend selection, and validation of exported figures.

Key features

01Define figure claims and evidence
02Select a drawing backend by project context
03Map data, statistics, and uncertainty
04Export vector and raster figures
05Validate source and rendered outputs

Use cases

01Create a graphical abstract from supplied evidence
02Plan multi-panel scientific figures
03Adapt a template to compatible data
04Check figure integrity and readability

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/mubuky/nature-skills --skill nature-figure

After installing, type/nature-figurein Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

.gitignore

manifest.yaml

README_EN.md

README.md

agents/

openai.yaml

assets/

README.md

evals/

evals.json

references/

api.md

asset-adaptation.md

backend-selection.md

chart-types.md

common-patterns.md

demos.md

design-theory.md

figure-contract.md

figure-legend-conventions.md

nature-2026-observations.md

qa-contract.md

r-template-index.md

r-workflow.md

template-catalog.md

tutorials.md

scripts/

nature_figure_backend.py

plot_templates.py

validate_figure.py

static/core/

contract.md

stance.md

static/fragments/backend/

python.md

r.md

Details

From the project README

Nature Scientific Figures

Build the visual argument from supplied evidence. A graphical abstract or schematic remains a draft visual and must not invent mechanisms or results.

Route

Read manifest.yaml and its core contract. Select the backend in this order:

explicit user choice;
existing source file, project dependencies, or established project backend;
saved user preference;
ask only when Python and R are both viable and the choice would materially affect integration or maintainability.

If the user asks you to choose, use references/backend-selection.md and proceed. Use one backend as the source of truth for drawing, layout semantics, and export. Neutral viewers, image/PDF/SVG inspection, and deterministic validators may perform QA without becoming a second drawing backend. Persist a backend preference only when the user explicitly asks to make it the future default.

Figure contract

Before code, define:

one-sentence conclusion and intended reader;
evidence chain and role of every panel;
source-data fields, transformations, statistics, and uncertainty;
target journal dimensions, formats, fonts, and accessibility;
likely reviewer misreadings and integrity risks.

Load only the selected backend fragment and the references required by the archetype, chart, template adaptation, or export. Prefer the provenance-recorded chart-atlas/gallery examples or user code; do not copy an external paper figure.

Build and validate

Choose an archetype and a clear hero panel.
Map data and statistics before styling. Preserve missingness, sample sizes, units, and uncertainty.
Use restrained hierarchy, typography, and color; verify grayscale and color-vision robustness when distinctions matter.
Adapt templates by semantics, never by relabelling incompatible data.
Export the required vector/raster products under the task's output directory.
Resolve SKILL_DIR as the directory containing this SKILL.md, load references/qa-contract.md, run python3 "$SKILL_DIR/scripts/validate_figure.py" SOURCE, and inspect rendered outputs at final size with an appropriate neutral viewer.

Integrity and completion

Do not invent values, mechanisms, labels, institutional marks, or statistical significance. Do not hide data or alter analysis to improve appearance. Finish only when source and exports agree, labels and legends are readable, panel claims are evidence-supported, and final-size visual QA passes when a rendered deliverable is in scope and inspectable; otherwise state clearly that visual QA was not run.

Comments

Markdown supported