scientific-schematics

SUMMARY

What it does

This Agent Skill generates publication-quality scientific diagrams from natural language descriptions. It uses the Nano Banana 2 AI image generation model via OpenRouter, and optionally reviews the output with Gemini 3.6 Flash. The skill implements an iterative refinement loop: it generates an image, reviews it against a quality threshold based on the target document type (e.g., journal, conference, thesis), and regenerates only if the quality is below the threshold, up to a maximum of 2 iterations. The output is a raster PNG image, with a review log (JSON) recording scores and critiques. The skill requires an OpenRouter API key and sends prompts and images to OpenRouter, so data leaves the machine. It is designed for creating diagrams such as neural network architectures, flowcharts, biological pathways, and system diagrams. The skill is part of the Scientific Agent Skills library by K-Dense Inc., licensed under MIT, and includes scripts and reference documentation.