What problem does it solve? Research figures often ship with cluttered layouts, unreadable labels, or inconsistent styling because plotting code is trusted without inspecting the rendered output. This Skill enforces a disciplined visual standard and a mandatory self-review pass so milestone charts, paper figures, and appendix figures communicate one clear claim. ## Core Features & Use Cases - Surface classification: Distinguishes connector milestone images, paper main figures, appendix figures, and internal review plots, each with its own export rules (PNG vs vector PDF/SVG plus preview). - Academic style contract: Applies a bundled muted Morandi Matplotlib style (assets/deepscientist-academic.mplstyle) with rules for spines, grids, legends, colormaps, and chart selection by research question. - Mandatory render-inspect-revise loop: Requires opening the rendered figure, checking a readability checklist, revising, and re-exporting before treating any figure as final. - Use Case: After an experiment campaign, produce a double-column paper figure comparing your method against baselines, inspect the rendered PNG, fix legend overlap and tick density, then export PDF plus PNG preview and record it in the figure catalog. ## Quick Start Use the figure-polish skill to turn my experiment results CSV into a clean paper-ready line chart and review the rendered output before finalizing it.
