What problem does it solve? Content pipelines generate scattered signals — post performance, niche research, operator feedback, agent failures — but no single agent connects them into strategy. This Skill reads all system intelligence and distills it into specific, actionable weekly direction for downstream content agents. ## Core Features & Use Cases - Performance Analysis: Parses LinkedIn post metrics (impressions, likes, comments, reposts) logged via Telegram and connects them back to idea types, hook styles, and topics. - Content Mix Balancing: Audits the last 4 weeks of drafts across opinion, how-to, and story formats, flagging any type exceeding 50% and recommending untried cross-niche formats. - Feedback-Driven Rule Updates: Detects patterns in operator image feedback (3+ similar notes) and updates brand-guidelines.md with concrete design refinements. - Use Case: Every week after research runs, the agent queries Supabase for fresh research, drafts, and feedback, then rewrites performance-patterns.md with directives like "this week MUST include a how-to using the 'I Tested This' format" and sends a Telegram summary. ## Quick Start Run bash agents/marketing_director/run.sh on the weekly planning day to synthesize all intelligence and update the content direction files.
