What problem does it solve? When starting a new game project, the visual style is often undecided, leaving AI agents to guess at art direction. This Skill structures that decision by asking the user up to 7 targeted questions and persisting the answers as written constraints, so downstream art generation follows the human's stated preferences instead of assumptions. ## Core Features & Use Cases - Two-round structured interview: Asks 4 questions about fundamentals (camera viewpoint, pixel art vs high resolution, palette size, outlines) then 3 about finish (saturation, post-effects, mood references), using AskUserQuestion with an "omakase" (leave it to AI) escape hatch. - Constraint recording: Writes only the axes the human actually chose into the "この画面の画風" section of AGENTS.local.md, always leaving a dated trace comment so status checks stop prompting. - Unattended-mode handling: In CI or sub-agent contexts where no one can answer, it proceeds without fabricating preferences, recording only the trace comment. - Use Case: When make status prompts for /style-interview on a new game project, run it to capture the user's art direction before handing off component and color decisions to /visual-dict. ## Quick Start Ask me the style-interview questions to define the visual style for my new game and record my answers in AGENTS.local.md.
