banner-creator

Banner-creator produces professional banners through AI image generation with an iterative, feedback-driven design process. It solves the common need for header, hero, cover, and social banners (GitHub README, Twitter/X header, LinkedIn banner, website hero, YouTube channel art) by guiding a structured workflow from requirements gathering to a final cropped, ready-to-use asset.

The workflow starts with a discovery step that gathers purpose, target ratio/size, style preference, content elements, and colors, waiting for user confirmation before generating. It then generates batches of variations (up to 20) at the widest 21:9 ratio using a companion 'nanobanana' AI image-generation skill (Gemini 3 Pro Image), builds a self-contained HTML preview page so the user can compare options, iterates on favorites based on feedback with a consistent naming scheme, and finally crops the approved banner to the target ratio and dimensions using a bundled Python crop script. It documents common prompt patterns (text, character, abstract/gradient, scene-based), supported aspect ratios, and per-platform sizes, and organizes all outputs into a dated .skill-archive directory. It also supports incorporating an existing logo image into a banner via image editing.

It targets developers, project maintainers, and creators who need branded banners quickly without a designer. The skill requires a GEMINI_API_KEY set in the environment (obtained from Google AI Studio) and the companion nanobanana skill. Use cases include creating a GitHub README banner matching an existing pixel-art logo, a Twitter header, or a website hero image. It ships with a crop script, preview template, format reference, and a worked example conversation.