This is an early release preview. You may encounter bugs.
Last30days Skill logo
Unclaimed

Tool research

Last30days Skill

AI-agent skill that researches a topic across social platforms and the web and synthesizes an engagement-ranked brief

A+ 90/100 GitHub score ? This grade is derived from GitHub signals, not user votes. Open for the full breakdown.
No votes yet

01 / About

What Last30days Skill is.

/last30days is an agent skill that researches a topic, person, or company across social platforms, prediction markets, code hosts, and the web for a recent time window, then has the agent synthesize the findings into a single brief. Rather than ranking pages by search-engine relevance, it scores sources by engagement signals such as Reddit upvotes, X likes, YouTube and TikTok view counts, Hacker News points, and Polymarket odds, so the brief reflects what people are reacting to.

It runs as a Claude Code plugin and through the Agent Skills format on Codex, Cursor, Copilot, Gemini CLI, Claude Desktop, OpenClaw, and other hosts. Reddit, Hacker News, Polymarket, and GitHub work without configuration; a setup wizard on first run adds API keys or browser sessions for X, YouTube, TikTok, Instagram, LinkedIn, and other sources, and installs free CLIs for arXiv, Techmeme, and Digg. Sources you bring your own credentials for run in parallel, and the results are cross-referenced before the agent writes the summary.

Beyond a plain topic query, the skill has modes for specific tasks: a --hiring-signals flag turns a company's job pages into cited evidence of focus shifts, a discovery mode ("what's exploding in AI agents?") sweeps Reddit, Hacker News, Digg, and X to return velocity-ranked topics with momentum labels, and ticker or crypto queries automatically pull StockTwits sentiment. Comments are a default-on layer, with rank-based diversity so multiple quotes do not all come from one post.

Features

  • Source coverage: Reddit, X, YouTube transcripts, TikTok, Instagram Reels, Hacker News, Polymarket, GitHub, LinkedIn, Threads, Bluesky, Pinterest, StockTwits, Xiaohongshu, Perplexity, arXiv, Techmeme, Digg, and web results
  • Engagement scoring: items are ranked by upvotes, likes, views, points, and market volume rather than editorial placement
  • Keyless defaults: Reddit (via RSS and arctic-shift scraping with real scores and top comments), Hacker News, Polymarket, GitHub, arXiv, and Techmeme work with no API key
  • Setup wizard: a first-run flow unlocks authenticated sources and installs helper CLIs
  • People research: for a person, pulls PR velocity, top repositories, release notes, and recent posts and transcripts
  • Hiring signals: --hiring-signals cites current job listings as evidence of a company's direction
  • Discovery mode: returns 5–10 velocity-ranked emerging topics, each with cross-source numbers and a ready-to-run follow-up query
  • Relevance floor: filters viral off-topic posts so they cannot dominate a brief
  • Renderer-aware citations: output is formatted per host so citations read as a brief rather than a list of URLs
  • Multi-host support: Claude Code marketplace plugin, native Codex plugin, and npx skills add for other Agent Skills hosts

02 / Discussion CREDIBILITY-GATED

Discussion

Reading is open to everyone. Posting and voting need a verified identity or a GitHub grade of B or higher.

  • No discussions yet.

04 / Build

Build with Last30days Skill.

Browse the catalogue for frameworks, tools, and harnesses, each scored on real GitHub credibility.

Get Last30days Skill →

Browse the catalogue