Figure Craft

Use when making a FIGURE / CHART / PLOT for a paper, deck, or presentation — in ANY language (R/ggplot2, Python/matplotlib, Julia/Makie). Governs the visual style: 16.5pt text throughout (axes, legend, title), no y-label (the title carries the y-axis meaning; paper figures get no headline title), concise axis labels, clean theme with colorblind-safe Paul-Tol palette, B&W-legible (shape/linetype redundancy), and a mandatory visual self-check (no clipped text, no overlapping labels). Covers DiD/event-study plots (dots + error bars, dashed treatment line), trend lines, stacked bars, distributions, scatter — and the layout each forces. At most 1-2 inside-canvas annotations that ENHANCE understanding (absolute counts on pct bars, pretreatment-level label on DiD); metadata (N, source, p-values) goes in LaTeX notes. Output: 5×3 in, saved to results/figures/. Fires on "make a figure", "plot this", "chart", "draw", "visualize" — AFTER the data is framed and verified.

lancegui Updated 2 repo stars
Causal Powers flow: any data, analysis, or figure request goes to question-framing, which forks to descriptive, reduced-form, structural, or predictive work, th

File contents

lancegui/causal-powers commit ca75d5f484

Frequently asked questions

npx get-skill add lancegui/figure-craft