Skill2Poster

Skill2Poster is a two-step workflow for creating academic conference posters from research papers. It is written primarily in Chinese but the concepts are straightforward: first use an AI image generator to quickly explore layout options, then rebuild the poster in HTML using real text, figures, and data from the actual paper. The repository provides two reusable agent skills that guide each step, along with example output images.

The motivation behind the split approach is that AI image generators are good at producing visually coherent poster layouts but tend to distort text, numbers, equations, tables, and QR codes. Using a generated image as the final product would give you a poster that looks polished but contains fabricated content. Skill2Poster treats the generated image as a design sketch only and then rebuilds everything accurately in HTML.

The first skill, paper-poster-imagegen, helps you construct a detailed prompt describing the poster's structure: section order, metric cards at the top, a thesis statement strip, and placeholder areas for real figures. The output is a visual mockup showing what the layout should look like. If the mockup has problems such as unreadable tiny text or invented charts, you adjust the prompt and regenerate before moving on.

The second skill, academic-poster-builder, uses that mockup as a reference and constructs the actual poster as an HTML file sized to A0 dimensions (the standard large-format size for academic conferences). All text, figures, tables, equations, logos, and QR codes come from the paper's own materials rather than from the image generator. The skill includes a quality checklist covering panel overflow, font readability, column alignment, and figure cropping.

The repository includes example images showing both the design reference and the finished HTML export. No scripts or application code are included; the skills are YAML-defined agent configurations intended to work with OpenAI-compatible tooling.