image-prompt-designer

# Image Prompt Designer Builds AI image prompts using Strategy A (shared Art Direction Prefix for visual consistency across 4 images). ## Inputs - `workspace/{task_id}/image_slots.json` - `projects/{slug}/brand-identity.json` (colors, founded, mission) - `projects/{slug}/brand-config.json` (voice_pair, industry, target_locale) - `workspace/{task_id}/angle.json` (format_id, hook) - `references/image/style-presets.md` — 12 visual styles - `references/image/format-style-mapping.md` — 24 format × style mapping - `references/image/negative-prompts.md` — universal + category negatives ## Output `workspace/{task_id}/image-prompts.json` — ⛔ `prompts` MUST be a JSON **array** of objects (each with a `slot_id`), NEVER an object keyed by slot_id. The dict-keyed shape silently broke chart-render, real-photo sourcing, the placeholder lint, and the publish gate on 2026-06-29; the schema-validate hook now blocks a dict-shaped write. ```json { "art_direction": { "visual_style": "editorial documentary photography, National Geographic 2026 aesthetic", "color_signature": "primary #FF8C00, secondary #2E5B7A, accent natural earth tones", "lighting": "golden hour or soft natural daylight", "mood": "competent, focused, aspirational", "camera_specs": "35mm lens, f/4-f/5.6", "universal_negatives": "no text overlays, no watermarks, no AI face tells..." }, "prompts": [ { "slot_id": "cover", "subject": "Single male angler casting a fly fishing rod over...