last30days-skill

Product Introduction

1. Definition: The last30days-skill is an AI agent skill, a modular software component that equips AI assistants (like Claude, Codex, or Copilot) with the ability to perform automated, multi-source research. It functions as a real-time information synthesis engine, aggregating and analyzing data from over 15 disparate platforms including Reddit, X (Twitter), YouTube, Hacker News, Polymarket, GitHub, arXiv, and Techmeme.
2. Core Value Proposition: It exists to solve the critical problem of AI-native information overload and fragmented narratives. By bridging the "walled gardens" of modern platforms—each with its own API, authentication, and data format—it delivers a single, grounded, and scored summary of the latest 30 days of discussion, sentiment, and data around any topic, ranked by real human engagement rather than SEO.

Main Features

1. Parallel Multi-Source Aggregation: The skill's core engine fires concurrent queries across all configured data sources. It uses a pre-research "brain" to first resolve the correct handles, subreddits, and hashtags for a topic before any API call is made. Technically, it leverages asynchronous Python (asyncio) and a plugin architecture where each source (e.g., reddit_plugin, x_plugin) is a separate module handling authentication, rate-limiting, and data normalization into a common schema.
2. Engagement-Weighted Scoring & Synthesis: Raw data (posts, comments, videos, market odds) is not just collected but scored. The system assigns weight based on platform-native metrics: Reddit upvotes, X likes, YouTube view counts, TikTok engagement, and Polymarket trading volume backed by real money. An AI agent (the LLM) then acts as a judge, synthesizing these scored signals into a coherent brief, prioritizing information with high social proof.
3. Zero-Config & Guided Authentication: For immediate use, keyless sources like Reddit (via RSS/scraping), Hacker News, arXiv, and Techmeme work out-of-the-box. A first-run setup wizard (--preflight) guides users through unlocking premium sources like X and YouTube by helping extract browser cookies or set up API keys. The system includes a --doctor command for comprehensive diagnostics and repair instructions.
4. Advanced Output Modes & Context Awareness: Beyond standard summaries, it offers specialized modes: --hiring-signals analyzes company job pages for strategic insights; --github-user profiles developer activity from commits and PRs; --competitors auto-discovers and compares similar tools; and ELI5 mode simplifies complex summaries. It generates self-contained, shareable HTML briefs with renderer-aware citations for clean reading in AI agent interfaces.

Problems Solved

1. Pain Point: Traditional search (Google) and individual AI models (ChatGPT, Gemini) suffer from platform blindness and temporal lag. Google buries Reddit discussions and cannot search X posts natively. LLMs have outdated training data and lack real-time API access to critical social and community platforms, leading to incomplete or stale intelligence.
2. Target Audience: AI Researchers & Analysts tracking daily shifts in a fast-moving field; Content Creators & Journalists seeking the unfiltered community pulse on a topic; Product Managers & Founders conducting competitive analysis and market sentiment checks; Investors & Traders gauging real-world sentiment and prediction market odds; Professionals preparing for meetings with deep due diligence on individuals or companies.
3. Use Cases: Performing due diligence before a sales call or investment meeting; monitoring brand or product sentiment across developer (HN), consumer (Reddit), and professional (LinkedIn) circles; tracking the velocity of emerging tech trends (e.g., "AI agents") before they peak; getting a consolidated brief on current events, synthesizing news, social reaction, and prediction markets; learning a new technical concept quickly via community-voted best practices and tutorials.

Unique Advantages

1. Differentiation: Unlike Perplexity or traditional news aggregators that primarily search the web or curated news, last30days-skill searches people and their interactions. It treats a Reddit thread with 1,500 upvotes as a stronger signal than a low-traffic blog post, and Polymarket odds backed by $66,000 as harder data than pundit speculation. It doesn't replace these tools but layers their signals with social and crowd-wisdom data they cannot access.
2. Key Innovation: Its "bring your own keys" agent-centric model is the fundamental unlock. No single company can legally or technically aggregate all these sources at scale. However, an individual user can provide their own authentication tokens (cookies, API keys). The skill provides the orchestration framework that allows an AI agent to temporarily act as that user, querying all platforms in parallel, a feat impossible for any monolithic search engine or AI model today.

Frequently Asked Questions (FAQ)

1. Is last30days-skill free to use? Yes, the core skill is open-source (MIT license). Many sources like Reddit (via RSS/scraping), Hacker News, arXiv, and Techmeme are free and require no API keys. Access to platforms like X (Twitter), YouTube, and LinkedIn requires you to provide your own authentication (e.g., browser cookies), which the setup wizard helps you configure.
2. How does last30days-skill get real-time Reddit data without the official API? Following changes to Reddit's public API, the skill uses a hybrid approach: it leverages keyless RSS feeds for subreddit listings and employs dedicated tools like arctic-shift for scraping specific threads to obtain real upvote counts and top comments, ensuring access to unfiltered community sentiment without an API key.
3. Can I use last30days-skill to track stock or cryptocurrency sentiment? Yes. The skill auto-detects when a query is a stock ticker (e.g., TSLA) or cryptocurrency (e.g., Bitcoin) and activates the StockTwits source to pull in trader sentiment. This is combined with general social discussion and news for a rounded view.
4. How do I install last30days-skill on Claude.ai (web version)? For the Claude.ai web interface, you download the last30days.skill file from the GitHub repository's releases page. Then, in Claude.ai, go to Customize > Skills > + > Create skill > Upload a skill, and select the downloaded file. This manually loads the skill into your session.
5. What does the "engagement-weighted scoring" mean in practice? It means the AI synthesis prioritizes information based on measurable human interaction. For example, a YouTube comment with 2,000 likes under a relevant video will be ranked higher than a blog comment. A Polymarket prediction with 96% "Yes" probability and high trading volume is treated as a high-confidence data point. This creates a summary reflective of what people are actually engaging with, not just what is published.