Реверс-инжиниринг промта по изображению

★ 7.1 · media-processing

Get Prompt from Image is a Claude Code skill that analyzes a user-provided reference image and reverse-engineers a high-fidelity prompt for AI image-generation tools. Reach for it when you want to recreate the look of an illustration you love, match the mood of a photograph, replicate a poster's visual language, or reproduce a product render's style without knowing how to phrase the original request. The skill identifies the medium type — photography, realistic 3D, anime, flat vector, mixed media, and others — then extracts 3–5 reproduction-critical visual anchors: composition, lighting, materials, color relationships, spatial layers, and mood. Those anchors are placed at the front of a positive prompt, while a negative prompt excludes medium confusion and common generation defects. Output arrives in both Chinese and English, ready to paste directly into an image-generation tool. The direction is image-to-text, not text-to-image. Brands, IP, and protected characters are translated into shape, silhouette, palette, and design language without relying on specific names. Works with photographs, illustrations, 3D renders, typography, logos, posters, and landscapes; not a substitute for OCR and not intended for plain content description.