What problem does it solve? Choosing an art direction for a game character, environment, or prop is hard when you can only imagine it in words. This Skill produces 3-4 deliberately divergent rough concept images of the same subject so you can react to real visuals and lock a direction before committing to a final asset. ## Core Features & Use Cases - Batch Variant Generation: Issues 3-4 parallel image generation calls where each variant differs along one explicit axis such as palette, era, stylization, silhouette, or lighting. - Exploration vs Final Disambiguation: Asks a routing question up front so users who want a single polished portrait are handed off to the character-portrait skill instead. - Prior Concept Search: Checks existing assets via summer_search_assets before generating, offering to riff on prior work with img2img or start fresh. - Use Case: You are designing a swamp witch and are unsure of the mood. The Skill generates four painterly concepts varying only in palette (warm autumn, cold moonlit, sickly green, monochrome ash), you pick one, and the choice is recorded in GameSoul.md for future asset skills. ## Quick Start Ask your AI agent to show you 3-4 art direction options for your character or environment, for example: "Give me 4 concept variants of the goblin warrior varying the silhouette."
