What problem does it solve? Analysis code often produces plots that run cleanly but are unreadable in a paper or on a projector: clipped labels, overlapping ticks, tiny fonts, color-only distinctions that collapse in grayscale, and lines that falsely imply continuity between discrete estimates. This Skill enforces a consistent house style and a mandatory visual self-check so every figure is legible and honest. ## Core Features & Use Cases - Cross-language house style: Applies the same principles in R/ggplot2, Python/matplotlib, and Julia/Makie — clean theme, Paul-Tol colorblind-safe palette, 16.5pt fonts, no y-axis label (the title carries it), and concise axis labels. - Chart-type decision rules: Dots plus error bars for DiD/event-study estimates with a dashed treatment line, lines for continuous series, stacked bars for composition, and guidance for distributions and scatter plots. - Black-and-white legibility: Requires redundant shape, linetype, or hatch channels so figures survive grayscale printing and colorblind readers. - Use Case: After verifying an event-study regression, ask for a coefficient plot — the Skill produces a 5x3 inch PDF saved to results/figures/ with dodged points, a dashed treatment threshold at -0.5, B&W-safe shapes, and a rendered visual check for clipping and overlap. ## Quick Start Make an event-study coefficient plot from my verified regression results and save it to results/figures/ following the house style.
