last30days-skill – Open-Source Cross-Platform AI Agent for Real-Time Comment Research

1. What is last30days-skill

last30days-skill is an open-source AI Agent research skill that enables agents to automatically scrape real discussions from major overseas social platforms over the past 30 days and synthesize them into reports with citations. The tool deeply crawls Reddit comments, X top replies, YouTube full transcripts, TikTok interactions, and Polymarket odds, sorting them by real human engagement. last30days-skill supports intelligent topic parsing, cross-source story clustering, automatic competitive analysis, and shareable HTML briefs, helping users quickly grasp the true community consensus and latest developments on any topic.

Technical Positioning and Domain: last30days-skill belongs to the intersection of natural language processing and information retrieval, specifically positioned as a research skill for AI Agents focused on social media. It differs from traditional search engines or summarization tools by concentrating on extracting deep discussion data (comments, replies, transcripts, etc.) from multiple social platforms and sorting them by human engagement signals, providing users with research reports based on real community consensus. Its uniqueness lies in combining AI Agent capabilities with multi-platform data aggregation to achieve automated, continuous social intelligence collection, filling the gap in traditional search for in-depth comment area mining.

Development Background: This tool was created and open-sourced by developer mvanhorn under the MIT license, aiming to address the issues of fragmented information and the prevalence of SEO-optimized content. As social platforms have become the main venues for real discussions, but remain siloed, users struggle to obtain a global perspective across platforms. last30days-skill emerged to address this, allowing researchers and developers to freely customize and deploy it via open-source, promoting democratization in social data research. Its development was driven by the pursuit of "real signal" – measuring information value through human behaviors such as likes, comments, and bets, rather than SEO rankings.

Core Value: The core value of last30days-skill lies in providing "community consensus driven by real signals." It uses human engagement metrics such as upvotes, likes, views, and Polymarket odds to filter out low-quality SEO-optimized content and present the true community discussion热度. Users no longer need to manually browse multiple platforms to obtain comprehensive reports with cited sources, significantly improving the efficiency and quality of information retrieval. For scenarios such as product research, trend analysis, competitive comparison, and background investigations on individuals, this tool offers more in-depth and reference-worthy primary information than traditional search methods.

Technical Features: The tool employs an intelligent entity parsing engine (v3), which automatically identifies topic-related entities (such as X accounts, GitHub repositories, subreddits, TikTok hashtags) before searching, enabling precise search results. It supports cross-source story clustering, merging discussions about the same event across different platforms into a single entry to avoid duplication. It includes ELI5 mode, automatic competitive analysis, and a shareable dark-mode HTML brief, balancing depth with usability. Open-sourced under the MIT license, it contains no tracking or telemetry, and all research data is stored locally, ensuring user privacy.

2. Key Features

• Cross-platform Deep Aggregation: Simultaneously search Reddit, X, YouTube, TikTok, HN, Polymarket, GitHub, Bluesky, and more than a dozen other platforms. It not only captures post titles but also dives deeper to extract comments, replies, full video transcripts, and prediction market odds, providing in-depth data that goes far beyond the surface-level information of search engines.
• Authentic Popularity Ranking: Results are ranked based on human engagement signals such as upvotes, likes, views, and Polymarket odds, ensuring that content with high genuine discussion volume is prioritized. This effectively counters SEO spam and makes rankings reflect true community interest.
• Smart Pre-research: After entering a topic, the system automatically parses related entities (such as X accounts, GitHub repositories, subreddits, TikTok hashtags) and performs precise searches on these entities, significantly improving result relevance and avoiding noise from broad searches.
• Cross-source Clustering and Merging: When the same news or story appears on different platforms, it is automatically merged into a single entry to avoid redundancy. It displays discussion perspectives and popularity differences across platforms, allowing users to get a comprehensive view of the event in one place.
• Shareable HTML Briefing: Generates a self-contained HTML file in dark mode, with all references and clustering results embedded within. It can be directly embedded into Slack, email, or Notion, making it convenient for team collaboration and long-term archiving without relying on online services.
• GitHub Personal Mode: Automatically switches to author-based queries for person-related topics, displaying data such as PR merge rates, star counts, and activity updates. This is ideal for technical research on individuals, providing a more authentic activity profile than traditional resumes.
• ELI5 Mode: With one click, complex reports can be rewritten in plain, easy-to-understand language, reducing the reading barrier. This is suitable for non-expert audiences to quickly grasp core conclusions, expanding the tool's application scenarios.
• Competitor Auto-comparison: Supports the --competitors parameter, automatically identifying and performing parallel comparisons with top competitors, generating multi-dimensional comparison reports to assist in product selection and market analysis decisions.

3. How to Use

1. Environment Requirements and Installation: This tool runs as an AI Agent skill and requires a host environment for support. For Claude Code users, run /plugin marketplace add mvanhorn/last30days-skill and /plugin install last30days; for other hosts (such as Open Interpreter), run npx skills add mvanhorn/last30days-skill -g for a global installation. Ensure that Node.js (v16+) and npm are installed on your system.

2. First Run and Basic Search: After installation, type /last30days <topic> to initiate the first search. With zero configuration, the tool automatically works with four data sources: Reddit, HN, Polymarket, and GitHub, without requiring additional API keys. The tool will automatically parse the topic and return preliminary results.

3. Unlock More Data Sources: To enable additional platforms such as YouTube, TikTok, and Instagram, configure the corresponding credentials as prompted: log in to X browser (for scraping X replies), install yt-dlp (for YouTube transcripts), and set up ScrapeCreators or a Perplexity API key. Specific steps will be guided after the first run, and users can configure them progressively as needed.

4. Generate and Share Reports: After completing the search, simply ask /last30days <topic>, give me a shareable HTML brief to generate an offline HTML report in dark mode. The report includes all referenced sources and clustering results and can be shared directly. Subsequent deep questions can be asked within the same session, and the AI will answer based on the fully scraped context without requiring repeated searches.

4. Pros and Cons Analysis

5. Comparative Analysis with Similar Tools

Selection Recommendations: If the requirement is to deeply aggregate real discussions across multiple social platforms, especially to uncover first-hand information from Reddit, X, and YouTube comment sections, last30days-skill is the only open-source tool dedicated to this specific scenario, making it ideal for technical researchers and data enthusiasts. Its unique content depth and real-signal ranking mechanism are hard to replace among competitors, but it requires users to have a certain level of technical configuration ability.

For users who prioritize quickly obtaining web summaries and academic resources, Perplexity excels in citation accuracy and coverage breadth, and does not require local deployment. Gemini Advanced's deep research mode can generate structured reports, making it suitable for scenarios requiring multimodal interaction, although its capabilities for social data mining are weaker. Microsoft Copilot integrates well with the Office ecosystem and is suitable for everyday light searches. Overall, last30days-skill has unique advantages in social data depth and open-source freedom, but its configuration complexity and coverage limitations are notable drawbacks. Users should choose based on their technical background and research objectives.

6. Editor's Summary

last30days-skill stands out with its distinctive technological innovation: its intelligent entity parsing engine (v3) automatically identifies topic-related entities before search, significantly improving retrieval accuracy; the cross-source clustering and merging algorithm effectively addresses information redundancy; and the real-time popularity ranking mechanism uses human engagement as its core metric, fundamentally countering SEO pollution. These design choices make it unique in the field of social data mining, offering deeper and more authentic community consensus information compared to traditional search engines and general-purpose AI search tools.

In terms of practical value, this tool provides direct assistance to product managers, market analysts, technical researchers, and developers. In scenarios such as product selection, trend tracking, background investigations on individuals, and analysis of hot events, last30days-skill can significantly reduce information collection time and provide first-hand data from comment sections that traditional search methods cannot reach. The HTML briefs it generates are convenient for team collaboration and knowledge retention, further enhancing work efficiency.

The target user base is clearly defined: it is suitable for mid-to-high-level users with a certain technical background who frequently conduct social data research. While the configuration requirements may be somewhat high for ordinary consumers, the ELI5 mode and shareable reports reduce the understanding cost of the results. In terms of future development potential, if it can expand to more social platforms (especially Chinese platforms) and simplify the configuration process, it has the potential to become a standard tool for social intelligence analysis. Additionally, as the AI Agent ecosystem matures, last30days-skill, as a skill plugin, can integrate with more hosts and expand its application boundaries.

7. Application Scenarios

• Pre-event Person Research: Input the target person's name, and automatically aggregate their posts, code submissions, and community discussions from platforms such as X, GitHub, and Reddit over the past 30 days, generating a more authentic dynamic profile than LinkedIn. Useful for investment due diligence, partner evaluation, and background checks on speakers.
• Learning New Technical Concepts: Input emerging concepts such as "loop engineering" to obtain first-hand community discussions, diverging opinions, and best practices, avoiding reliance solely on second-hand information. Helps technical professionals quickly grasp the latest trends and community consensus.
• Product Selection Comparison: Input comparisons such as "OpenClaw vs Hermes" to generate a multi-dimensional comparison report based on real user feedback, GitHub star trends, and community controversies. Aids in technical decision-making and reduces the influence of advertising and marketing content.
• Tracking Hot Events: Input topics such as "Kanye West's new album" or international conflict issues to aggregate news, prediction market odds, YouTube in-depth analysis, and reactions from social platforms, gaining comprehensive insights. Suitable for media monitoring and sentiment analysis.
• Consumer and Travel Decisions: Input items such as "Universal Epic Universe" or electronic products to obtain the latest community updates on queue times, repair issues, real user experiences, and cost-performance discussions. Helps consumers make decisions based on genuine user feedback.

8. FAQ

Q: Is the last30days-skill plugin paid?
A: It is completely free and open source, using the MIT license, and can be used without charge. However, some advanced data sources (such as YouTube transcripts, TikTok data) require configuring third-party APIs or tools, and these services may have their own pricing policies. Users are responsible for any associated costs.

Q: Does it support searching for topics in Chinese?
A: The tool itself does not restrict language, but the main social platforms it covers (such as Reddit, X, YouTube) are primarily in English, with relatively fewer Chinese discussions. The results for Chinese topics may not be as ideal as for English ones. It is recommended to use it in combination with tools specifically designed for Chinese platforms.

Q: How to configure the YouTube transcription feature?
A: You need to install the yt-dlp command-line tool and ensure it is available in your environment variables. The tool will provide installation prompts on its first run. For specific installation instructions, refer to the official yt-dlp documentation, which supports Windows, macOS, and Linux.

Q: How is data privacy protected?
A: All research data is stored locally on your machine. The tool does not include tracking or telemetry features. As an open source project, users can review the code themselves to ensure there is no data leakage. The generated HTML report is fully offline and does not rely on any external servers.

Q: Can I customize data sources or add new platforms?
A: Currently, data sources are pre-configured, but since it is an open source project, developers can fork the code and add new data scrapers. Future versions may support plugin extensions, and community contributors can participate in development.

Q: Does the generated HTML report include all references?
A: Yes, the report includes all links to the comments, replies, and transcriptions that were collected, with the source platform indicated for easy verification and traceability. Users can directly click the links to navigate to the original content.

9. Project Links

• GitHub Repository: https://github.com/mvanhorn/last30days-skill