research-figures

Generate scientific figures from papers, LaTeX manuscripts, patent drafts, method descriptions, and technical notes by routing requests to the correct backend. Use when Codex needs to create an editable architecture diagram, a publication-style method figure, a visual abstract, a teaser illustration, a patent figure, or a quantitative plot, and when backend selection, scientific content extraction, and figure verification matter as much as the final rendering.

Generate paper-style illustrations with the Banana/Jiekou text-to-image API when the user wants a visual abstract, teaser figure, concept illustration, method overview art, or publication-style research image from a paper abstract, method description, caption, or technical note. Use this skill for image-first scientific figures, not for precise charts, tables, or editable box-arrow diagrams.

Generate editable draw.io architecture diagrams from paper text, patent drafts, technical notes, or method descriptions. Use when the user wants a scientific or technical architecture figure as a real .drawio file for diagrams.net/draw.io, especially for papers, patents, proposals, system block diagrams, or model framework figures.