What problem does it solve? Deciding which figures a mathematical modeling paper actually needs is hard: authors often produce redundant charts, fabricate unsupported diagnostics, or pick the wrong chart type for their model. This Skill identifies the models in a problem, report, or codebase and builds a minimal, evidence-backed figure chain for CUMCM and general modeling papers. ## Core Features & Use Cases - Model-to-figure mapping: Classifies modeling components (statistics, forecasting, evaluation, optimization, networks, dynamics, simulation, ML, sensitivity) and assigns each figure a defined evidence role with a priority score. - Evidence auditing: Verifies that required data (iteration histories, repeated runs, labels and scores, coordinates) actually exist before recommending a figure, and declares data gaps instead of fabricating results. - Bundled publication templates: Adapts 11 scibox-style Matplotlib templates (raincloud, ROC with CI, Taylor diagram, chord diagram, circular heatmap, SHAP combo, and more) by replacing synthetic data with real project data while preserving visual style and PNG/PDF/SVG export. - Use Case: Given a CUMCM solution using entropy-weighted TOPSIS plus a genetic algorithm, ask what figures to draw and receive a ranked plan covering weight plots, ranking sensitivity, convergence curves, and route maps with data sources, libraries, captions, and paper placement. ## Quick Start Ask the assistant to analyze your modeling report or code and recommend which figures the paper should include, using the model-figure-selector skill.
