image-generation-skill

Turn any research paper (PDF, arXiv URL, .docx, HTML, .tex export) into publication-ready figures via a 5-step pipeline — read the paper, write an in-depth résumé with a figure catalog, choose a render path (code for charts/diagrams/chat UIs; Nano Banana JSON for conceptual art), then generate PNGs. Handles three figure archetypes found in systems/ML papers: chat-UI mockups, architecture diagrams, and performance charts. Default output looks like the actual figures inside a real paper. Use when the user asks for paper figures, diagrams, schematics, system overviews, result charts, or illustrated conversation examples from a paper, abstract, or scientific PDF.