About this skill

Problem to Solve

E-commerce product images often fail because a model can generate an image, but not to platform rules: ozon white-background main images must have no text, logo, or watermark; taobao / pdd need 1:1 square images; language, subject size, scene people, and selling-point layout must also pass. If you ask a large model freely, you may get wrong dimensions, watermarked main images, inconsistent product shapes across the set, or prohibited absolute claims, making the assets unusable.

How the Skill Works

Gate: first checks whether the current model supports image generation; if not, it stops immediately instead of producing prompts or files.
Input parsing: recognizes ozon, taobao, and pdd; requires product dimensions and asks for missing platform, reference images, or size rather than inventing values.
Suite planning: plans a white-background main image, one infographic image, four scene images, a dimension image, and a selling-point image, separating pure-product scenes from people scenes.
Prompt orchestration: emits per-image prompts for image models, enforcing #FFFFFF background, 70%–85% subject coverage, no shadow, Russian for ozon, and Chinese for other platforms.
Compliance checks: validates language, subject size, dimensions, file size, shape consistency, prohibited claims, people appearance, and scene coverage.
Deliverable output: writes a {YYYYMMDD}_{platform}_{NNN}/ folder with images, 产品说明.md, or ozon 双语说明.txt containing title suggestions, selling points, and marketing keywords.

Boundaries

Best for e-commerce design and operations workflows where reference images, product dimensions, and platform are known. It is not a real-time official data scraper, price changer, automatic listing publisher, or photo-editing suite. Platform thresholds in reference.md are third-party summaries, so verify rules against the platform backend before launch.

Screenshots

Use Cases

Ops with multi-angle references needs Ozon 750×1000 white-background, Russian infographic, and bilingual listing assets.
Taobao sellers preparing sub-images need 1440×1440 scene, dimension, and Chinese selling-point images with prohibited-claim checks.
Designers given one reference and dimensions need eight product-image prompts, compliance checks, and a delivery folder layout.
Cross-border ops needs title suggestions, selling points, and keywords written into an Ozon bilingual .txt for listing.

Best For

Cross-border ops owners listing on Ozon need 750×1000 Russian-compliant product images and bilingual listing notes.
Taobao or PDD designers need multi-angle references turned into 1440×1440 white-background, scene, and Chinese selling-point images.
E-commerce content leads need checks for prohibited claims, watermarks, wrong dimensions, or inconsistent product shapes.
Multimodal application engineers need platform image rules packaged into a fixed prompt pipeline.