The "last30days-skill" repository, developed by mvanhorn, is an innovative AI agent skill designed to provide comprehensive, up-to-date summaries on any given topic by aggregating information from a diverse range of online platforms. Its primary function is to act as a research assistant, synthesizing information from sources like Reddit, X (formerly Twitter), YouTube, Hacker News, Polymarket, and the broader web to deliver a concise and grounded overview of a subject within the last 30 days. The skill distinguishes itself by prioritizing information engagement metrics like upvotes, likes, and real-world financial backing (Polymarket odds) over traditional editorial or SEO-driven rankings.
The core feature of the skill is its ability to search multiple platforms in parallel, a capability that sets it apart from conventional search engines. It leverages the unique data available on each platform, such as Reddit comments, X posts, YouTube transcripts, and Polymarket predictions, to offer a more holistic and nuanced understanding of a topic. The skill's architecture allows it to bypass the limitations of individual platforms, which often operate as walled gardens with their own APIs and authentication requirements. Users can bring their own API keys and browser sessions to unlock access to these platforms, enabling the AI agent to gather and synthesize information from disparate sources.
The skill’s main purpose is to provide users with a rapid and comprehensive understanding of a subject, particularly in rapidly evolving fields like AI. It aims to keep users informed about the latest developments, trends, and community discussions. The repository's creator built it to stay ahead of the curve in the AI space, recognizing that the most current information often appears first on platforms like Reddit and X. The skill has evolved beyond this initial purpose, now serving as a valuable tool for various applications, including pre-meeting research, tracking product launches, comparing tools, understanding global events, and planning trips.
The skill's v3 update introduced several significant improvements. Intelligent search is a key enhancement, allowing the engine to understand the topic and identify relevant sources, including specific individuals, subreddits, and hashtags, before initiating the search. This pre-research capability, developed by [@j-sperling], significantly improves the accuracy and relevance of the results. The "Best Takes" feature highlights humorous, witty, and viral content, adding a layer of social context to the summaries. Cross-source cluster merging consolidates information from multiple sources on the same topic, preventing redundancy. Other enhancements include single-pass comparisons, auto-discovered competitor comparisons, a GitHub person-mode for analyzing individuals' activities, and an "ELI5" mode for simplified explanations.
The skill is designed to be easily integrated into various platforms, including claude.ai, Claude Code, OpenClaw, and Gemini CLI. Users can install the skill through the Claude marketplace, OpenClaw, or by cloning the repository. The skill requires users to bring their own keys for certain platforms, such as X, YouTube, and others, to access their content. The repository emphasizes its open-source nature, with an MIT license, no tracking, and a commitment to keeping research data private. The project is built with Python 3.12+, yt-dlp, Node.js, and the ScrapeCreators API. The repository also includes a detailed changelog and a star history chart, demonstrating its development and popularity. The skill's ability to synthesize information from diverse sources, combined with its focus on social engagement metrics, makes it a powerful tool for staying informed in today's dynamic information landscape.
