What problem does it solve?
This Skill helps you generate publication-quality academic illustration images with correct layout, styling, and readable labels, avoiding the trial-and-error that usually makes diagram-generation unusable for papers.
Core Features & Use Cases
- Execution-grounded, multi-stage generation: plans the figure, optimizes layout, verifies CVPR/ICLR/NeurIPS-style constraints, generates a native raster via an MCP bridge, and then strictly reviews it.
- Strict acceptance criteria with iterative refinement: scores outputs (target score ≥ 9), rejects unclear or non-paper-ready figures, and loops with specific improvement instructions up to a capped number of iterations.
- Canonical integration artifacts: runs a required preflight step, finalizes to standardized outputs (including LaTeX include snippet and verification receipts), and verifies before reporting success.
- Use Case: turn a paper method description like an end-to-end pipeline into a clean architecture/workflow figure with correct arrow direction, hierarchy, and print-friendly typography.
Quick Start
Ask for a paper-ready workflow diagram and run the Skill’s required preflight, then generate the figure through the Codex image2 bridge and finalize with verification.