| id | figure_styling_skills_index |
| name | Figure Styling Skills Index |
| description | Aesthetic guidelines and output-type recipes for scientific figure production.
Supports lightweight default-agent use through SKILL.md + one outputType recipe,
with optional venue-specific style guides when requested.
|
Figure Styling
General scientific figure styling skill for default agents and graph-generation tasks.
This skill has two layers:
- OutputType recipes — self-contained recipes for default-agent use (
SKILL.md + one recipe file).
- Optional style guides — venue- or aesthetic-specific refinement, loaded only when requested.
Lightweight read rule: For normal use, read only this SKILL.md and one outputType recipe. Do not load styles/, quality/, input/, or triage/ unless explicitly needed.
Route by outputType
| outputType | Read | Rendering mode | Use for |
|---|
figure | figure.md | code-first | paper figures, data plots, multi-panel figures |
poster | poster.md | mixed | academic posters |
graphical-abstract | graphical-abstract.md | AI-first | graphical abstracts, TOC figures |
presentation | presentation.md | mixed | slide visuals |
flowchart | flowchart.md | AI-first | workflow diagrams, process diagrams |
Optional style guides
Read these only when the user explicitly requests a venue/style or a style_id is provided.
If a listed file is not present, continue with the outputType recipe and report the missing file.
| style_id | File | Use for |
|---|
neurips_diagram | styles/neurips_diagram.md | ML methodology diagrams (NeurIPS / ICML / ICLR / CVPR) |
neurips_plot | styles/neurips_plot.md | ML/statistical plots (NeurIPS / ICML / ICLR / CVPR) |
nature_figure | styles/nature_figure.md | Nature / Cell / Science-style scientific figures |
ieee_figure | styles/ieee_figure.md | IEEE journal / conference figures |
color_palettes | styles/color_palettes.md | Colorblind-safe palette reference (Paul Tol) |
Scientific visual language
Use concrete visual schemas instead of generic scientific prompts.
Prefer:
- Flat vector scientific illustration
- Clean white or very light background
- Short accurate labels
- Explicit entities and connections
- Editable-looking vector composition
- Journal-grade readability
- Clear module hierarchy
- Restrained scientific palette
Avoid:
- Generic AI art
- Decorative 3D or photorealistic glow
- Fake charts, fake axes, fake data in AI diagrams
- Dense tiny text or unreadable legends
- Abstract boxes without scientific meaning
- Childish cartoon style
- Heavy shadows or decorative gradients
Domain adaptation rule
Do not assume a fixed scientific domain.
When the user provides a domain, instantiate the visual schema with that domain's concrete entities, structures, labels, and relationships.
Examples (not defaults):
- 3D genomics: TAD, chromatin loop, boundary insulation, enhancer, promoter, CTCF/cohesin, Hi-C contact map, genome-browser tracks
- Cell biology: nucleus, mitochondria, ER, Golgi, membrane receptors, signaling pathway
- Immunology: T cell, antigen-presenting cell, receptor, cytokine, immune synapse
- Neuroscience: neuron, synapse, axon, dendrite, action potential
- ML/systems: encoder, decoder, attention, pipeline stage, data flow
These are domain instantiation examples, not default assumptions.
Custom styles
Additional .md style files may be added to styles/ following the existing file structure. Use the matching style_id when requesting a venue or aesthetic style.
