What problem does it solve?
It helps you turn a research request into publication-quality academic figures by running a multi-stage generation and enforcing strict visual verification, so your diagrams meet conference-style requirements.
Core Features & Use Cases
- Multi-stage figure generation: Uses a Claude-led planning step plus Gemini-assisted layout optimization and CVPR/NeurIPS style verification, then renders the final diagram.
- Strict visual review loop: Performs mandatory, itemized checks (especially arrow directions, block content correctness, readability, and print-friendliness) and iteratively refines until quality is high (target score ≥ 9).
- Supports common paper figure types: Best for architecture diagrams, method illustrations, and conceptual figures where you need clean, academic visuals (not photo-realistic images or statistical plots).
Quick Start
Use the paper-illustration skill to generate a CVPR-style pipeline diagram by providing a description of the components, data flow, and visual constraints.