What problem does it solve? Creating promotional images for a restaurant (LINE banners, menu photos, logos, seasonal key visuals) involves more than just generating a picture: the output must pass EXIF/metadata removal, malicious SVG checks, and Japanese advertising regulation review (Pharmaceutical Affairs Law, Premiums and Representations Act, food labeling) before it can be safely published. This Skill turns a vague request like "make an autumn banner" into a controlled pipeline from purpose definition to human-approved publishing. ## Core Features & Use Cases - Purpose-first generation: Confirms use case (LINE asset, store logo, menu photo, site banner), direction (season, mood, target customers), and constraints before generating, then pulls store context and brand profile to keep prompts consistent. - Multi-candidate workflow with safety gates: Produces at least 3 candidates via the nano banana / Gemini image connector, requires human selection, then enforces EXIF removal, mXSS/malicious SVG inspection, and size/format checks before upload. - Regulation and approval control: Flags NG claims under Japanese advertising and food labeling laws for human review, uploads only via signed URLs (create_line_message_asset_upload_url) or update_store_logo, and never auto-publishes. - Use Case: A store manager says "I want an autumn key visual for LINE." The Skill checks the image connector, drafts prompts from the store's brand profile, generates 3 directional candidates, runs safety and regulation checks on the chosen one, uploads it via the signed URL, and stops at human publish approval. ## Quick Start Ask the assistant to create an autumn-themed LINE banner for the store and have it walk through candidate generation, safety checks, and upload preparation.
