What problem does it solve? Turning an approved LinkedIn post into a scroll-stopping branded image normally requires a designer or manual template editing. This Skill automates that step: it reads an approved draft, selects the right HTML template, writes a visual headline, fills all template variables, and renders a LinkedIn-ready JPG. ## Core Features & Use Cases - Template Selection: Matches post type (stat, framework, comparison, opinion, quote, before/after) to one of six HTML templates such as editorial_data_card or split_comparison. - Headline & Variable Generation: Writes a scroll-stopping headline under 12 words and fills every template variable as JSON, enforcing brand rules like font sizes, accent color usage, and the three-zone layout. - Automated Rendering & Delivery: Saves template_name and template_vars to Supabase, runs scripts/render_template.py to produce a 1080x1350 JPG, and delivers the image to Telegram for human approval. - Use Case: After you approve a post draft in Telegram, the agent automatically designs the accompanying image, renders it locally with Playwright, and sends it back to you for review — with support for regeneration via feedback. ## Quick Start Run bash agents/social_media_designer/run.sh with a draft ID to design and render the image for that approved post.
