---

✨ What the templates look like (real renders)

Every image below is a real output of a built-in FigureForge template — not a mockup: pick a skeleton, swap in your data, export SVG / PNG / TIFF / PDF / PPTX in one click.

Advanced charts — omics-, cohort- and project-management-level figures, also out of the box:

Bioinformatics atlas — dot-clustered heatmap, correlation matrix, ring heatmap, Mantel test, GSEA, PCA confidence ellipses, marginal-density scatter and other high-frequency bioinformatics figures, in NPG-style colors:

Extra atlas (new) — OncoPrint, waterfall, alluvial, ridgeline, Bland-Altman agreement, immune-composition stacked area, forest plot, Manhattan plot: 8 high-frequency clinical and omics figures:

Plus atlas (new) — chord diagram, UpSet, network graph, ternary plot, clustered heatmap, treemap, sunburst, hexbin density, parallel coordinates: 9 advanced compositions:

Ultra atlas (new) — Circos ring diagram, QQ plot, KM survival curves with risk table, raincloud, GO lollipop, Venn diagram, Gantt chart, radial dendrogram, sequencing-coverage tracks:

Max atlas (new) — scatterplot matrix, MA plot, pathway dot plot, circle packing, bump chart, 2D contour density, dumbbell plot, diverging stacked bars, calendar heatmap:

Nova atlas (new) — Sankey flow, chord correlation diagram, mosaic plot, PCA biplot, population pyramid, ring heatmap, dual-axis combo, pseudo-3D response surface, ECDF with KS test:

Apex atlas (new) — funnel plot, Taylor diagram, arc diagram, icicle chart, beeswarm, cycle plot, horizon chart, bullet chart, waffle chart:

🧪 FigureForge — figure-making without code

Open online (nothing to install): jing1312.github.io/nature-figure-skill/figureforge/

Core capabilities

• Infinite canvas (VibePaper-style workspace): generate a chart onto the canvas with one sentence; arrange chart / image / sticky-note / text node cards freely and drag ports to create reference links; the asset plaza ships 9 atlases with 83 panels ready to grab; double-click a card to refine it in the editor and changes write back automatically.
• 22 publication-grade templates: bars (superplot raw-dot overlay), clustered heatmap (row/column dendrograms + diverging scale), volcano plot, forest plot, KM survival curves, violin, box plots, multi-panel composites… all pre-set to submission standards with error bars, significance brackets, confidence bands and legend placement.
• 16 curated palettes: Tableau 10, Tol Muted, Economist, Morandi, macaron, Candy pink-blue-green-yellow-purple, Berry blue-purple-pink family, and more; series-linked recoloring — change one color and every data point, legend entry and colorbar in that series updates in sync, with one-step undo for the whole group.
• Figma-grade editing: smart guide snapping (Alt temporarily disables), Shift+click / marquee multi-select, Ctrl+G grouping, corner-handle resize, arrow-key nudging, full undo/redo; multi-selection pops up an align / distribute toolbar, the format painter copies styles in one click, and imported external SVGs are split into per-element editable shapes (nested transform groups hit precisely).
• Bitmap editing kit: drag in or Ctrl+V paste a screenshot onto the canvas, then crop by double-click, rotate 90°, flip, and adjust brightness/contrast/saturation.
• Day & night themes: ☀ / 🌙 one-click toggle with a full-screen starfield transition; 14 workspace backgrounds to choose from.
• Session autosave: edits persist to the browser in real time — close the page and pick up where you left off; projects can also be exported as .json.
• Multi-format export: SVG (text stays editable) / PNG / TIFF (with dpi metadata, submission-ready) / PDF (vector, selectable text) / PPTX, at 1×–4× scale and 300/600 dpi.

🤖 nature-figure — Agent-skill figure workflow

Load SKILL.md into your agent (Claude Code, Codex, etc.) and figure-making turns from "one-off scripts" into "an auditable production line":

figure contract → backend gate → single-backend render → layout & QA → independent export → delivery checklist

The workflow enforces three hard rules, mapped to the three biggest sources of figure rework:

① Independent figure-set export — three figures are three figures; never silently merge them into one. The whole set shares fonts, axis wording, color semantics and statistical precision.

def save_pub_set(fig, out_dir, stem, dpi=600):
    fig.savefig(out_dir / f'{stem}.svg',  bbox_inches='tight')  # primary: editable
    fig.savefig(out_dir / f'{stem}.pdf',  bbox_inches='tight')  # TrueType embedded
    fig.savefig(out_dir / f'{stem}.tiff', dpi=dpi, bbox_inches='tight')  # print
    plt.close(fig)

② Non-occluding annotation — legends and stat boxes always stay out of data-dense regions; stat strips are frameless and quieter than the data:

ax.legend(loc='center left', bbox_to_anchor=(1.02, 0.5), frameon=False, fontsize=6)
add_stat_strip(ax, 'Pearson r = 0.83, n = 46')   # built-in helper

③ Semantic color — color is a scientific language: assign roles first (primary method / control / threshold / background), then encode by significance; a user-provided color list is a "palette direction", not a sequence to apply in order.

🚀 Quick start

Browser (zero install): open the online FigureForge, pick a template and start dragging.

Hand it to your agent:

Please install this Agent Skill: https://github.com/jing1312/nature-figure-skill
Review the repository contents and SKILL.md first, then install the whole directory into your current agent's Skills directory.
When done, report the actual installation path and verification result.

Manual install (copy the whole directory — the skill depends on references/ and other files):

git clone https://github.com/jing1312/nature-figure-skill.git
cd nature-figure-skill
mkdir -p ~/.codex/skills
cp -R skills/nature-figure ~/.codex/skills/    # main skill, full extensions

See install.md for details.

📦 Deliverables

Not "one bitmap" — a submission-grade file set for every figure:

figure-01.svg    ← primary: text stays editable, journals can reword directly
figure-01.pdf    ← TrueType embedded, submission companion
figure-01.tiff   ← 600 dpi, meets most journals' print requirements
figure-01.png    ← preview / slides

🗺 Roadmap

• [x] 22 publication-grade templates + 16 curated palettes
• [x] Day/night themes / bitmap editing kit / series-linked recoloring / full undo-redo
• [x] Multi-line text editing (tspan line breaking)
• [x] Vector PDF export (svg2pdf.js + jsPDF, text stays selectable)
• [x] Radar, effect-size funnel, Gantt timeline and other new figure types (shipped with the 9 atlases / 83 panels in the asset plaza)
• [ ] Bar hatch / dot-pattern fills (print-friendly black & white)
• [ ] FigureForge ↔ Python pipeline loop (editor SVG → save_pub_set one-click PDF/TIFF backfill)
• [ ] Journal spec presets: 89 / 183 mm column-width checks, minimum font-size checks
• [ ] Color-vision deficiency simulation preview
• [ ] GitHub Actions: template XML validity + syntax gates

📚 Research skill pack bundled with this repo

🔍 Derivative transparency

git clone https://github.com/Yuan1z0825/nature-skills && git -C nature-skills checkout f3941a1
git clone https://github.com/jing1312/nature-figure-skill && git -C nature-figure-skill checkout 11fc2b8
diff -r nature-skills/skills/nature-figure nature-figure-skill/skills/nature-figure

📄 License & acknowledgments

• The FigureForge editor and this repo's documentation and template extensions © jing1312, released under the MIT license.
• The nature-* skill content is derived from Yuan1z0825/nature-skills © Yuan Yizhe and contributors, MIT license; original copyright notice preserved in LICENSE.
• Issues and PRs are welcome — or make your next submission figure with the online editor.