Academic Plotting

Turns paper sections and experiment results into conference-ready figures without the usual matplotlib hell. You paste in your methods paragraph or a results table, and it extracts what needs to be visualized. For architecture diagrams it generates them via Gemini in one of three visual styles (hand-drawn sketch, modern minimal, or icon-rich). For data it auto-picks chart types and renders with matplotlib/seaborn using proper academic styling. The style guide is opinionated, which is good because most ML paper figures look identical. Handles the tedious stuff like consistent colors, readable fonts, and making "your method" stand out in comparisons. Built for NeurIPS/ICML submissions where figure quality actually matters.

Install to Claude Code

npx -y skills add orchestra-research/ai-research-skills --skill academic-plotting --agent claude-code