This connects Claude to a brute-force prediction system that tests millions of combinations of Trump social media signals against S&P 500 movements. It's tested 31.5M model combinations and maintains 551 surviving rules with a verified 61.3% hit rate across 566 predictions. You get MCP tools for checking daily signals, model performance rankings, and LONG/SHORT consensus predictions based on post timing, keywords like "RELIEF" or "TARIFF", and other features. The underlying system runs overnight searches, monitors Truth Social and X every 5 minutes, and exposes everything through JSON files and a REST API. Honestly, this is applied statistical pattern mining at scale. Whether the patterns hold up out of sample is the real question, but the data and methodology are all open for inspection.
npx skills add https://github.com/aradotso/trending-skills --skill trump-code-market-signals