AI Agent Skill: Cross-Platform Social Search and 30-Day Synthesis
/last30days is an AI agent skill that aggregates the latest content from Reddit, X, YouTube, TikTok, Hacker News, and more into a 30-day briefing. It uses entity pre-research to identify key people, communities, and topics, then searches in parallel and scores by real engagement (upvotes, likes, money) rather than SEO. An AI synthesizes a cited, in-depth summary. Open-source (MIT), it supports Claude Code and 50+ agent frameworks. Ideal for engineers, PMs, and researchers needing a quick, grounded update before meetings or decisions.
/last30days is an AI agent skill that searches across Reddit, X (Twitter), YouTube, Hacker News, Polymarket, GitHub, and the web in parallel, then synthesizes the findings into a grounded briefing. Its core isn't replacing traditional search engines—it's letting a single agent penetrate multiple walled gardens simultaneously, ranking results by real engagement signals (upvotes, likes, views, money) instead of editorial picks or SEO scores.
/last30days 是一个 AI agent skill（技能），可跨 Reddit、X（Twitter）、YouTube、Hacker News、Polymarket、GitHub 等多个平台并行检索，并将检索结果综合成一份紧扣事实的简报。它的核心不在于替换传统搜索引擎，而让一个 agent 同时穿透多个封闭的社交网络，以真实的互动数据（赞、投票、播放量、赌注）作为排序依据——不再是编辑或 SEO 排名。
Every social platform is a walled garden. Google doesn't index Reddit comments or X posts; ChatGPT has a deal with Reddit but can't search X or TikTok; Gemini has YouTube but not Reddit; Claude has none natively. Manually tab-hopping across a dozen domains every day is unsustainable. Meanwhile, AI training data lags months behind what communities figure out in real time, making traditional search blind to the most current discoveries.
每个社交平台都是数据孤岛。Google 搜索不索引 Reddit 评论或 X 帖子；ChatGPT 虽有 Reddit 合作但无法搜 X 或 TikTok；Gemini 有 YouTube 但缺少 Reddit；Claude 没有任何原生社交数据。若逐个平台手工检索，一天的更新可能分布在十几个域名里，耗时极高。同时，AI 模型的训练数据总是滞后几个月，社群刚刚讨论出的新发现在传统搜索里完全看不到。
The engine first runs a pre-research step to resolve entities: X handles, subreddits, GitHub repos, YouTube channels, TikTok hashtags relevant to your topic. Then it searches all configured sources in parallel, scoring results by real engagement: upvotes, comments, views, Polymarket odds. Cross-source cluster merging groups the same story appearing on multiple platforms into one item. v3 adds a separate humor/virality judge that surfaces witty one-liners in a 'Best Takes' section.
/last30days 的核心是并行检索和实体理解。当你输入一个话题，引擎会先做预研究（pre-research），自动解析出相关的 X 账号、子版块、GitHub 仓库、YouTube 频道等实体。然后在所有已配置源上同时发起搜索，结果按点赞数、评论数、播放量、预测市场赔率等真实互动指标评分。引擎还具备同故事跨源聚类合并的能力——同一事件如果在 Reddit、X 和 YouTube 都有报道，会被合并为一个条目，而非碎片化输出。v3 版本还引入了「最佳评语」（Best Takes）独立评分机制，会对每条结果额外检测幽默与传播力。
Right after install, Reddit (with comments), Hacker News, Polymarket, and GitHub work with zero configuration. The first run triggers a 30-second setup wizard to unlock more sources: X via browser login, YouTube via brew install yt-dlp, TikTok/Instagram/Threads via ScrapeCreators API (100 free credits), Perplexity via OpenRouter. Other sources like Bluesky require an app password.
安装后，Reddit（含评论）、Hacker News、Polymarket、GitHub 即开即用，无需任何 API Key。首次运行 /last30days 会弹出一个 30 秒的设置向导，引导你配置 X（只需在浏览器登录）、YouTube（需要 brew install yt-dlp）、TikTok、Instagram、Threads、Bluesky 及 Perplexity 等源。每个源的支持方式不同：X 依赖本机浏览器 cookie，YouTube 使用 yt-dlp，TikTok 等通常需要 ScrapeCreators API Key（100 免费额度），Perplexity 走 OpenRouter 付费。
Multiple install paths are supported: Claude Code uses /plugin marketplace add mvanhorn/last30days-skill with auto-updates; other AI hosts (Codex, Cursor, Copilot, Gemini CLI, 50+ more) use npx skills add mvanhorn/last30days-skill -g; claude.ai web requires manual .skill file upload; OpenClaw users run clawhub install last30days-official.
项目提供了多条安装路径：Claude Code 用户可通过 /plugin marketplace add mvanhorn/last30days-skill 安装并由市场自动更新；其他 AI 宿主（Codex、Cursor、Copilot、Gemini CLI 等 50+ 平台）用 npx skills add mvanhorn/last30days-skill -g 全局安装；claude.ai 网页版可手动上传 .skill 文件；OpenClaw 用户则用 clawhub install last30days-official。
Just type /last30days <topic> in the agent chat. For example, /last30days Peter Steinberger returns a consolidated briefing covering recent X posts, Reddit threads, YouTube interviews, and GitHub PR activity—all with source citations. Advanced usage includes --emit=html for shareable offline HTML briefs, --competitors for automatic competitive analysis, --github-user for person-mode GitHub queries, and saying 'eli5' after a run to rewrite the briefing in plain language.
只需在 agent 对话中键入 /last30days <话题>。比如 /last30days Peter Steinberger 会输出此人在最近 30 天的 X 帖子、Reddit 讨论、YouTube 访谈、GitHub PR 记录的综合简报，并引用每条信息的来源。进阶功能包括：--emit=html 生成可分享的、离线可用的 HTML 简报文件；--competitors 自动发现并比较竞品；--github-user 启用 GitHub 人的模式（作者级查询）；在搜索结果后说 eli5 可一键将简报转为白话版。
Best for: pre-meeting person research, comparing tools/competitors, tracking breaking news across communities, checking real community opinions before a trip, and fast-learning a new domain with an agent-written prompt. Caveats: some sources require paid API keys (ScrapeCreators, OpenRouter) with limited free tiers; web search needs Brave Search API (2K free queries/month); TikTok/Instagram/Threads are pulled on-demand, not streamed; output quality depends on the underlying AI model; the project requires Python 3.12+, yt-dlp, and Node.js—not a zero-dependency tool.
适合场景：会前快速了解人物背景、比较竞品、追踪行业突发事件、旅行前查社区口碑、快速学习新领域（附带 agent 自写 prompt）。已知边界：部分源依赖付费 API（ScrapeCreators、OpenRouter），免费额度有限；web 搜索需要 Brave Search API Key（每月 2000 次免费）；TikTok、Instagram、Threads 等源在被搜索时才加载，非实时流式；输出质量受限于所选 AI 模型的推理能力；项目基于 Python 3.12+ 及 yt-dlp、Node.js 等依赖，不是纯零依赖方案。