Last30Days Skill is Matt Van Horn's open-source research skill for turning the past 30 days of posts, releases, and web sources into a linked brief.
To catch up on an open-source tool, ask Last30Days to find recent releases and issues on GitHub and see what users are discussing on Reddit and Hacker News. The brief puts release notes and community feedback together, with links back to the originals. Engagement signals such as upvotes and comments can affect how results are ranked. See the repository and examples.
| Source | What it needs |
|---|---|
| Reddit, Hacker News, GitHub, and Polymarket | Basic search works without an API key. You can start with these sources. |
| YouTube | You can add YouTube through yt-dlp. YouTube comments are listed separately in the configuration guide. |
| X | Use a signed-in browser session or an API. The official recent-search endpoint usually covers about a week by default. Longer coverage depends on your access. |
| TikTok, Instagram, and other optional sources | These can be added through ScrapeCreators. Usage limits and charges follow the provider's plan. |
Last30Days runs in tools that support skills, including Claude Code, Codex, and Cursor. After installing it, start with /last30days, enter a topic, and name the sources you want to use. The skill is MIT-licensed. Your host tool and any external services can have their own charges. See the source configuration guide and MIT License.
The topic goes to the network sources used in the run and to the web-search service. If a supported hosted model key is configured, the skill can also send the topic and shortlisted titles, snippets, and comments to that provider for reranking. A key set up for search can also enable reranking. Material from local files added as a corpus is scored locally. See the data handling and security documentation.
Last30Days Skill
Ready to try it? Open the official website.