What problem does it solve? Research reports and direction discussions often hide weak evidence behind aggregate bars, fabricated error bars, or misleading chart defaults. This Skill produces honest, reproducible figures and diagrams that encode evidence without distortion, and validates every figure before it is embedded in a report. ## Core Features & Use Cases - Data Charts and Evidence Tables: Select the right form (histogram, paired plot, small multiples, reliability diagram) from the analytical question, with strict rules against fabricated uncertainty, truncated axes, and misleading defaults. - Argument and Process Diagrams: Draw implementation-chain diagrams, related-work alignment maps, risk maps, and decision diagrams using Mermaid or self-contained inline SVG with a fixed semantic color palette. - Figure Validation and Retention: Run a final review checklist (axis labels, colorblind safety, caption completeness, render verification) and classify each artifact as diagnostic, evidence, or explanatory with explicit retention rules. - Use Case: After an experiment run, hand off a statistical summary and receive a reproducible matplotlib script plus PDF/PNG figure whose caption states the comparison, observational unit, and uncertainty semantics. ## Quick Start Ask the assistant to plot the baseline-expected-actual comparison for your latest experiment run as a publication-ready figure with an honest caption.
