What problem does it solve? Producing on-brand Movistar campaign pieces (email, display, OOH, Meta, Movistar+) manually leads to broken typography, hand-copied base64 assets, generic stock-like imagery, and unverified layouts. This Skill replaces that with a programmatic pipeline: brand assets as files, slot-based HTML templates, AI-generated photography with curated visual references, and a mandatory visual QA loop before delivery. ## Core Features & Use Cases - Slot-based assembly: Write HTML with slots like {{FONT_FACE_MIN}}, {{TOKENS_CSS}}, and {{LOGO_MARK}}; assemble.py injects real fonts, logos, and design tokens so the model never touches base64. - Reference-driven image generation: generate_image.py calls OpenAI gpt-image-2 with 2 Gold Standard references per image (Track A for pure photography, Track B for full pieces), enforcing brand-consistent light, casting, and composition. - Render and visual QA: render.py converts assembled HTML to PNG for a 14-point deterministic checklist (palette, logo placement, CTA type, sentence case, no overlaps, photo integration) with a maximum of 2 correction iterations. - Use Case: An Art Director agent receives a brief for a Movistar+ WOW banner. It reads 2-3 real reference pieces, builds the slot-based HTML, generates the key art with channel-specific Gold Standards, assembles, renders, verifies against the checklist, and delivers a client-ready PNG plus a contextual mockup. ## Quick Start Ask the agent to produce a Movistar email or banner for your campaign brief, and it will run the reference-first workflow, generate photography, assemble, render, and visually verify the final piece.
