What problem does it solve? Generating new images or video that match an existing visual style is hard because a caption per image does not generalize. This Skill analyzes a set of reference images and produces a structured style guide that captures what is consistent across the set, so future generations match the look even for subjects never shown in the references. ## Core Features & Use Cases - Multi-stage style analysis: A seven-stage procedure covering intake, per-image observation, style-versus-incident separation, palette rules, character construction rules, and handoff with confidence labels. - Structured markdown output: Writes files such as STYLE.md, palette.md, rendering.md, lighting.md, characters.md, extrapolation.md, negatives.md, and manifest.json to ~/.gen-ai/projects/style/<slug>/ for reuse by other skills. - Extrapolation rules: Derives construction rules so unseen subjects (a pig from a tiger, grass from a tree) can be rendered in the same style, with established/provisional/inferred/unknown confidence labels. - Use Case: A user drops twelve screenshots from a cartoon and asks for a style guide; the Skill produces a guide that other skills like video-prompt-engineer or text-to-visual consume to generate matching new content. ## Quick Start Build a style guide from these reference images so I can generate new artwork in the same style.
