Figure Styling Skills
Resources for the Graph Maker Team's illustrator (diagram) and data_plotter (plot) agents. The leader writes aesthetic_guide: <style_id> into style_card.json; the producing agent then loads the matching style file listed below.
Available styles
| Style ID | File | Target | Figure class |
|---|---|---|---|
neurips_diagram |
styles/neurips_diagram.md | NeurIPS / top ML venues | Methodology / framework / pipeline diagrams |
neurips_plot |
styles/neurips_plot.md | NeurIPS / top ML venues | Statistical plots (bar, line, scatter, heatmap, …) |
How to use
- Leader sets
aesthetic_guide: "<style_id>"in{workdir}/inputs/style_card.json. - Sub-agent (illustrator / data_plotter) consults this skill index, then reads the style file whose id matches
aesthetic_guide. - Agent applies the guidance alongside
style_card.json. Priority chain for conflicts: user references > style_card.json > figure_styling/ > internal defaults.
If aesthetic_guide is custom or null, sub-agents do not load any file from this skill — they rely purely on style_card.json and their internal defaults.
Custom styles
Users can drop additional .md files into styles/ (e.g. nature_figure.md, ieee_figure.md, my_lab_style.md) following the same section structure as the NeurIPS guides. Set aesthetic_guide: "<new_style_id>" in style_card.json to activate.