What problem does it solve? Creating consistent, well-composed cover images for articles requires design decisions about type, palette, rendering style, text density, and mood that most writers cannot make quickly or reproduce reliably. ## Core Features & Use Cases - Five-Dimension Customization: Combine 6 cover types, 11 color palettes, 7 rendering styles, 4 text levels, and 3 mood levels, with content-driven auto-selection when dimensions are omitted. - Multi-Backend Raster Generation: Route prompts to runtime-native image tools (Codex imagegen, Hermes image_generate) or external backends like baoyu-image-gen, with a hard rule against substituting SVG or HTML for raster output. - Persistent Preferences: Store watermark, default aspect ratio, output directory, quick mode, and backend pinning in an EXTEND.md file at project or user level. - Use Case: A blogger finishing a technical article asks for a cover image; the skill analyzes the content, recommends a conceptual type with a cool palette and digital rendering, confirms the choices, writes a reproducible prompt file, and generates a 16:9 cover.png. ## Quick Start Ask the assistant to generate a cover image for your article, optionally specifying flags like --palette cool --rendering flat-vector --aspect 16:9.