Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →jaechang-hits/scientific-visualization
Guide for choosing and creating scientific visualizations for publications and talks. Covers chart-type selection by data structure, color theory for accessibility/print, figure composition, journal formatting (Nature, Cell, ACS), and common pitfalls. Consult when visualizing data or preparing submission figures.
Passed 19 of 25 test cases with the skill, and 16 of 25 without it.
3 fixed16 passed either way6 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 3 test cases the skill fixed. Same model, same prompt, with and without the skill.
We have assembled a 4-panel figure (panels A through D). The draft uses lowercase red titles ('a)', 'b)') placed in the bottom-right corner of each panel box to save top margin space. What is the proper convention for panel identifiers?
| Checked | Without skill | With skill |
|---|---|---|
| Place uppercase bold letters in the top-left corner of each panel. | Missed | Passed |
The instructions your model reads when the skill loads.
Effective scientific visualization communicates data clearly, honestly, and accessibly. Poor chart choices, misleading axes, or inaccessible color palettes can obscure findings or introduce bias. This guide covers the full workflow of scientific figure preparation: from selecting the right chart type for your data structure through color theory, accessibility, and journal submission formatting requirements.
Every chart type is optimized for a specific data structure. Mismatches (e.g., pie charts for continuous distributions, bar charts for time series) hide structure and distort perception.
| Data Type | Recommended Chart | Avoid |
|---|---|---|
| Continuous distribution (1 group) | Histogram, violin plot, ridge plot | Bar chart with mean only |
| Continuous distribution (2–5 groups) | Violin + boxplot overlay, beeswarm | Grouped bar chart |
| Two continuous variables, correlation | Scatter plot, hexbin (large N) | Line chart without temporal order |
| Categorical counts / proportions | Bar chart (horizontal for long labels) | Pie chart (>4 categories) |
| Change over time (continuous) | Line chart | Bar chart |
| Change over time (sparse events) | Step chart, event raster | Connected scatter |
| Part-to-whole (≤5 parts) | Stacked bar, waffle chart | 3D pie chart |
| High-dimensional (>5 variables) | Heatmap (clustered), parallel coordinates | 3D scatter |
| Spatial data | Map, spatial heatmap | Bubble chart |
| Survival / time-to-event | Kaplan-Meier curve | Bar chart of median survival |
Color encodes information. Misused color introduces artifacts and fails readers with color vision deficiency (CVD; ~8% of males).
Sequential palettes encode ordered numeric data from low to high (e.g., expression level, concentration). Use perceptually uniform palettes: viridis, magma, cividis. These also print in grayscale.
Diverging palettes encode data with a meaningful midpoint (e.g., fold-change centered at 0, correlation from -1 to +1). Use RdBu, coolwarm, or vlag. Always ensure the midpoint maps to white/neutral.
Qualitative palettes encode unordered categories. Use Okabe-Ito (CVD-safe), tab10 (matplotlib default), or ColorBrewer qualitative palettes. Limit to ≤8 distinguishable colors; use shape or pattern as redundant encoding beyond that.
Color don'ts:
test cases passed, vs 16 / 25 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.