Paper Illustration

Generates publication-ready figures for academic papers through a five-stage loop: Claude parses your request, Gemini refines the layout and checks conference style compliance (CVPR/NeurIPS color palettes, arrow conventions), renders via Paperbanana, then Claude scores the result 1-10 and forces iteration until it hits 9 or better. The opt-in style-ref flag lets you match caption length and figure density to a reference paper without copying visual design. Useful when you need architecture diagrams or method illustrations that actually pass the aesthetic bar of top-tier ML conferences. Max five iterations before it gives up, outputs to figures/ai_generated/. Requires GEMINI_API_KEY.

Install to Claude Code

npx -y skills add wanshuiyin/auto-claude-code-research-in-sleep --skill paper-illustration --agent claude-code