What problem does it solve? Manually monitoring job boards, prices, news feeds, or repositories is repetitive and easy to abandon. This Skill guides the creation of a scheduled data collection agent that scrapes public sources, scores and summarizes items with a free LLM, stores results in a database, and learns from user feedback — all running free on GitHub Actions. ## Core Features & Use Cases - Three-Layer Architecture: Implements a Collect → Enrich → Store pipeline with pluggable source connectors (REST API, HTML scraping, RSS, Playwright for JS-rendered pages). - Batched AI Enrichment: Uses Gemini Flash with a four-model fallback chain and batching (5 items per call) to stay within free-tier rate limits. - Feedback Learning: Converts user accept/reject decisions stored in Notion into prompt bias so scoring improves over time. - Use Case: Build an agent that checks job boards every 3 hours, scores each listing against your resume, pushes matches to a Notion database, and refines its scoring as you mark items Applied or Rejected. ## Quick Start Build me an agent that monitors Hacker News for AI startup funding news and stores scored summaries in Notion on a daily schedule.
