The last30days-skill is an AI agent-driven research tool that aggregates and ranks real-time signals from over 25 disconnected platforms—including Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub, arXiv, Techmeme, Yahoo Finance, patents, and careers pages—into a single synthesized brief scored by human engagement metrics (upvotes, views, betting odds, PR velocity) rather than SEO. It operates via a v4 pipeline defined in skills/last30days/SKILL.md, installable globally via npx skills add mvanhorn/last30days-skill -g across 50+ agent hosts (Claude Code, Codex, Cursor, Gemini CLI, Grok, OpenClaw) or as an MCP bundle for Claude Desktop. Zero-config sources (Reddit, HN, Polymarket, GitHub, patents, careers) work immediately; first-run setup auto-installs free CLIs for arXiv, Techmeme, and Yahoo Finance in ~30 seconds. Authenticated sources require user-supplied keys (X_BEARER_TOKEN, ScrapeCreators, Perplexity, Brave Search) or browser cookies. The engine runs parallel searches, applies an AI judge to synthesize findings, and emits structured Markdown or versioned JSON (--emit=json). Key features include "Secret Signals" (patent filings and job postings flagged only when out-of-character for a company), quote lines with inline attribution, Polymarket odds with delta, Yahoo Finance 30-day performance vs S&P 500, discovery mode (--discover) for trending topics, SQLite-backed trend monitoring (--store), and a local research library with Atom feed export. Python 3.12+ required; macOS Keychain integration available for secrets.
Read original
github-trending/python