What problem does it solve? Raw user requests for image generation are often vague or fragmented, causing image models to produce unstable or off-target results. This Skill converts those requests into complete, model-ready visual prompts before any image generation call is made. ## Core Features & Use Cases - Prompt Structuring: Transforms raw Chinese requirements into full visual instructions covering subject, scene, composition, camera angle, lighting, material, color, and mood. - Mode-Aware Rules: Applies distinct constraints for text-to-image, reference-guided, and image-to-image modes, defining what must stay unchanged versus what can be enhanced. - Multi-Reference Handling: Assigns explicit roles to each reference image (identity, composition, material, environment) instead of collapsing them into one vague constraint. - Use Case: A user asks for a Xiaohongshu-style cover image of a skincare product with two reference photos. The Skill locks the product identity from reference image 1, borrows lighting from reference image 2, and outputs a single final prompt with a clean title area and no rendered text artifacts. ## Quick Start Before generating an image, ask the assistant to optimize your idea into a final image prompt, for example: turn my request for a sunset product photo of this bottle into a ready-to-use generation prompt.
