Verdict. I use it as a second way of sampling what’s happening, next to whatever my own feeds show me. Engagement ranking picks what got attention, so I treat the output as a list of candidates and never as a finding.
what it is
An agent skill that searches more than 20 platforms, limited to the last 30 days, and writes a cited brief ranked by engagement instead of editorial picks. Sources include Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub, arXiv, Techmeme, LinkedIn, Bluesky and Digg. MIT licensed. It claims 2,700+ tests.
# Claude Code
/plugin marketplace add mvanhorn/last30days-skill
# any Agent Skills host (Codex, Cursor, Copilot, Gemini CLI)
npx skills add mvanhorn/last30days-skill -gNeeds Python 3.12+. The core sources (Reddit, HN, Polymarket, GitHub, arXiv, Techmeme, Bluesky) need no API keys. TikTok, Instagram and LinkedIn need a ScrapeCreators key. Perplexity or Brave can improve the synthesis.
Output goes to Markdown briefs with citations, JSON, self-contained HTML, or a SQLite database for watchlists and trend monitoring.
/last30days what's exploding in AI agents?
/last30days --hiring-signals <company>
how I use it
Weekly research sweeps and keeping an eye on competitors. The Polymarket and GitHub sources give me a signal I don’t get anywhere else.
what I found
- The keyless core made it free to try. The only cost was disk.
- The X and social sources work by pulling cookies out of your Chromium browsers. That’s a real permission, so understand it before you turn those sources on.
- Briefs land in
~/Documents/Last30Days/by default, outside your notes. I keep it that way and only copy the finding over.