Skip to content
Back to search

lets-go-rss

SkillWorks with Claude CodeGitHub

轻量级全平台 RSS 订阅管理器。一键聚合 YouTube、Vimeo、Behance、Twitter/X、知识星球、B站、微博、抖音、小红书的内容更新,支持增量去重和 AI 智能分类。

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/albedo-tabai--lets-go-rss/badge.svg)](/listings/albedo-tabai--lets-go-rss)

HTML

<a href="/listings/albedo-tabai--lets-go-rss"><img src="/api/listings/albedo-tabai--lets-go-rss/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Install | 安装

# Core dependencies | 核心依赖
pip install httpx yt-dlp

Basic Usage | 基本使用

# Add subscriptions | 添加订阅
python3 scripts/lets_go_rss.py --add "https://www.youtube.com/@MatthewEncina"
python3 scripts/lets_go_rss.py --add "https://vimeo.com/xkstudio"
python3 scripts/lets_go_rss.py --add "https://www.behance.net/yokohara6e48"

# Update all | 更新全部
python3 scripts/lets_go_rss.py --update --no-llm

# Digest mode (1 item per account) | 摘要模式(每账号 1 条)
python3 scripts/lets_go_rss.py --update --no-llm --digest --skip-setup

# Read cached report (bot push) | 读取缓存报告(Bot 推送)
python3 scripts/lets_go_rss.py --status

# List subscriptions | 查看订阅
python3 scripts/lets_go_rss.py --list

---

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 204 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. ---
  2. name: lets-go-rss
  3. description: 轻量级全平台 RSS 订阅管理器。一键聚合 YouTube、Vimeo、Behance、Twitter/X、知识星球、B站、微博、抖音、小红书的内容更新,支持增量去重和 AI 智能分类。
  4. ---
  5. # Let's Go RSS
  6. 全平台 RSS 内容聚合工具,支持增量更新、去重、AI 分类。
  7. ## 快速使用
  8. ### 添加订阅
  9. ```bash
  10. python3 scripts/lets_go_rss.py --add "https://www.youtube.com/@MatthewEncina"
  11. python3 scripts/lets_go_rss.py --add "https://vimeo.com/xkstudio"
  12. python3 scripts/lets_go_rss.py --add "https://www.behance.net/yokohara6e48"
  13. ```
  14. ### 更新全部(耗时操作,建议用 crontab 后台跑)
  15. ```bash
  16. python3 scripts/lets_go_rss.py --update --digest --skip-setup
  17. ```
  18. ### 读取缓存报告(Bot 推送用,瞬间返回)
  19. ```bash
  20. python3 scripts/lets_go_rss.py --status
  21. ```
  22. ### 查看订阅
  23. ```bash
  24. python3 scripts/lets_go_rss.py --list
  25. python3 scripts/lets_go_rss.py --stats
  26. ```
  27. ### 查看全量概览(所有订阅最新一条)
  28. ```bash
  29. python3 scripts/lets_go_rss.py --overview
  30. ```
  31. ### 健康体检 & 自动修复
  32. ```bash
  33. python3 scripts/lets_go_rss.py --doctor # 只打印状态
  34. python3 scripts/lets_go_rss.py --auto-fix # 自动重启挂掉的 RSSHub、清僵尸锁
  35. ```
  36. ### Playwright tier(绕过平台反爬)
  37. `RSS_PLAYWRIGHT_PLATFORMS` 控制哪些平台走 skill 自管的 Chromium(Playwright)—— 用真浏览器访问,继承"正常用户"反爬声誉。
  38. **默认已为 Bilibili 开启**(B 站无需登录,直接工作)。Twitter / 小红书平台强制要求登录态。
  39. #### Agent 应主动编排的一次性登录流程
  40. 当 `--doctor` 或摘要报告显示 Twitter/XHS 源因 `auth` 失败时,**agent 应主动提议用户让你帮他完成登录**,而不是让用户自己去跑命令。标准套路:
  41. 1. `AskUserQuestion` 问用户:要不要现在登录 Twitter / 小红书(多选)
  42. 2. 对用户勾选的每个平台,运行:
  43. ```bash
  44. python3 scripts/lets_go_rss.py --login <twitter|xiaohongshu>
  45. ```
  46. 这一条命令会依次:① 打开可见 Chromium 到登录页 → ② 等用户关窗 → ③ 自动拉一条该平台已有订阅验证登录成功 → ④ 自动把该平台写进 `.env` 的 `RSS_PLAYWRIGHT_PLATFORMS`
  47. 3. 最后跑一次 `--doctor` 或一轮冒烟,确认失败数下降
  48. 用户看到的步骤极简:**只需要在弹出的 Chromium 里登录自己的账号,登完关窗,其他全部由 agent 搞定**。
  49. ## Bot 推送最佳实践
  50. **问题**:`--update` 需要 30-60 秒抓取全部订阅,Bot 定时任务可能超时。
  51. **方案**:抓取和推送解耦——crontab 提前跑更新,Bot 只读缓存文件。
  52. ### 稳定命令(推荐)
  53. ```bash
  54. # 后台更新(内置超时参数 + 并发防重入锁)
  55. ./scripts/run_update_cron.sh
  56. # Bot 推送只读缓存
  57. ./scripts/run_status_push.sh
  58. ```
  59. ```bash
  60. # crontab -e
  61. # 每 2 小时的 55 分更新(提前 5 分钟准备好数据)
  62. 55 */2 * * * cd /path/to/lets-go-rss && ./scripts/run_update_cron.sh >> /tmp/rss_cron.log 2>&1
  63. # Bot 在整点读缓存推送(瞬间完成)
  64. 0 */2 * * * cd /path/to/lets-go-rss && ./scripts/run_status_push.sh
  65. ```
  66. Bot 只需调用 `--status`,该命令直接读取 `assets/latest_update.md` 并输出内容,无需网络请求、无需等待。
  67. ## 平台支持
  68. | 平台 | 主策略 | 备注 |
  69. |------|--------|:----:|
  70. | Vimeo | Native RSS | ✅ |
  71. | Behance | Native RSS | ✅ |
  72. | YouTube | yt-dlp / Atom feed fallback | ✅ |
  73. | 微博 | RSSHub → Chrome session fallback | ⚠️ 需本机 Chrome 开启 remote debugging |
  74. | 抖音 | RSSHub | ⚠️ 当前不稳定 |
  75. | B站 | RSSHub | ⚠️ 需配置 |
  76. | 小红书 | Chrome session primary | ⚠️ 需本机 Chrome 已登录且开启 remote debugging |
  77. | Twitter/X | Syndication API | ✅ |
  78. | 知识星球 | pub-api (公开) | ✅ |
  79. ## 安装依赖
  80. ```bash
  81. # 基础(YouTube + Vimeo + Behance + Twitter syndication + Zsxq pub-api)
  82. pip install httpx yt-dlp anthropic
  83. # 配置环境变量:把 ANTHROPIC_API_KEY 写进 .env(已 git-ignore)
  84. cp .env.example .env && $EDITOR .env
  85. # 中国平台(B站/微博/抖音/小红书 RSSHub 路由)
  86. # 方式 A(推荐):skill 自管的 RSSHub,首次 start 自动 npm install
  87. python3 scripts/rsshub_manager.py start
  88. # 方式 B:若已经跑了 ready-cowork 的 :1200 RSSHub,skill 会自动作为 fallback
  89. # 方式 C:自己起 Docker rsshub 并通过 RSSHUB_BASE_URL_FALLBACK 指过去
  90. # 可选:启用真实 Chrome 会话抓取(微博 / 小红书推荐)
  91. # 1) 打开 Chrome
  92. # 2) 访问 chrome://inspect/#remote-debugging
  93. # 3) 勾选 Allow remote debugging for this browser instance
  94. ```
  95. ### RSSHub 自管与升级
  96. 首次 `rsshub_manager.py start` 会:
  97. 1. 在 `<skill>/node_modules/rsshub` 下 `npm install rsshub@latest`(~50MB,首次 1-2 分钟)
  98. 2. 如未装,自动下载 puppeteer 要求的 Chrome for Testing(约 200MB 到 `~/.cache/puppeteer`)
  99. 3. 启动 `scripts/rsshub_worker.mjs` 监听 :1201,写 pidfile 到 `assets/.rsshub.pid`
  100. 后续 `run_update_cron.sh` 在每次 cron 前都会调 `start-if-needed`(最多等 10 秒),保证 :1201 健在。每周一早晨 06:xx 额外调 `update` 拉最新 RSSHub(后台跑,不阻塞该轮 cron)。
  101. ```bash
  102. # 常用运维
  103. python3 scripts/rsshub_manager.py status
  104. python3 scripts/rsshub_manager.py restart
  105. python3 scripts/rsshub_manager.py update # npm install + restart
  106. python3 scripts/rsshub_manager.py logs # tail worker log
  107. ```
  108. ## Bot 汇报规范(⚠️ 必须严格遵守)
  109. 当 Bot 需要推送 RSS 更新时,**只需执行一个命令,然后原样转发输出**。
  110. ### 完整流程(仅 2 步)
  111. ```
  112. 步骤 1: 运行命令
  113. python3 scripts/lets_go_rss.py --status
  114. 步骤 2: 把命令输出原封不动地作为你的回复发送
  115. ```
  116. **就这么简单。不需要任何额外处理。**
  117. ### 输出格式说明
  118. `--status` 命令会输出类似以下格式的纯文本(自动生成,不需要 Bot 构造):
  119. ```
  120. 📡 RSS 增量更新 | 2026-02-21 18:23 | 3 个账号有新内容
  121. 🆕 📺 影视飓风 02-18 03:00
  122. [【4K限免】你的新设备能顶住吗?](https://t.bilibili.com/1170572725010300960)
  123. 🆕 🐦 歸藏(guizang.ai) 02-14 17:15
  124. [Tweet by @op7418](https://x.com/op7418/status/2022721414462374031)
  125. ```
  126. 仅显示有新内容的账号,无更新时显示"暂无新更新"。若需查看所有订阅的全量状态,使用 `--overview`。
  127. ### ❌ 禁止行为
  128. - ❌ **不得重新排版**:不可以按平台分组、加表格、加标题 `#` 层级
  129. - ❌ **不得分多条消息**:所有内容必须在一条消息内发送
  130. - ❌ **不得删除/修改链接**:标题中的链接不可去掉或替换
  131. - ❌ **不得添加前言后语**:不要加"以下是 RSS 更新"等多余文字
  132. - ❌ **不得执行 --update**:推送时只读缓存,不做抓取
  133. ### ⏸️ 暂无更新时的处理
  134. 当 `--status` 输出中显示"暂无新更新"或类似表述时,**只需回复一句话**:
  135. ```
  136. RSS 暂无新更新 ✅
  137. ```
  138. **不需要列出各账号的最新内容**,直接说暂无更新即可。
  139. ---
  140. ## 输出文件
  141. | 文件 | 说明 |
  142. |------|------|
  143. | `assets/latest_update.md` | 增量更新报告(`--status` 读取,仅含变动账号) |
  144. | `assets/full_overview.md` | 全量概览(`--overview` 读取,含所有账号) |
  145. | `assets/feed.xml` | 标准 RSS 2.0 XML |
  146. | `assets/summary.md` | 统计摘要 |
  147. | `assets/subscriptions.opml` | OPML 订阅导出 |
README.md · README · 221 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Let's Go RSS 🛰️ 全平台 RSS 订阅管理器
  2. > **AI-Powered Universal RSS Subscription Manager | AI 驱动的全平台 RSS 订阅管理器**
  3. A lightweight RSS aggregator designed to work as a **Claude Skill** inside AI-powered IDEs and agents. Add subscriptions from 7 platforms(YouTube, Vimeo, Behance, Bilibili, Weibo微博, Douyin抖音, Xiaohongshu小红书), auto-update with deduplication, and get digest reports — all through simple CLI commands that your AI assistant can run for you.
  4. 一个轻量级 RSS 聚合工具,设计为 **Claude Skill** 在 AI IDE 和 Agent 中运行。支持 7 个平台(YouTube, Vimeo, Behance, Bilibili, Weibo微博, Douyin抖音, Xiaohongshu小红书)的订阅管理、自动更新去重、智能摘要推送——通过简单的命令行指令,让你的 AI 助手自动完成。
  5. ---
  6. ## 💡 Why This Project? | 为什么做这个?
  7. ### The Problem | 问题
  8. We are trapped inside algorithms. The average internet user spends **2 hours 23 minutes per day** on social media (DataReportal 2024), of which **80-90% is passive zombie scrolling**. Research from UPenn shows that only **30 minutes** of daily social media use is genuinely beneficial — every second beyond that has diminishing or even negative returns.
  9. 我们被困在算法里。全球网民平均每天花 **2 小时 23 分钟**刷社交媒体(DataReportal 2024),其中 **80-90%** 是无意识被动浏览。宾夕法尼亚大学研究表明,每天社交媒体使用超过 **30 分钟**后的每一秒,边际效益都在递减甚至变为负值。
  10. ### The Science | 科学依据
  11. - **Dunbar's Number (150)** — Oxford anthropologist Robin Dunbar proved that humans can maintain at most ~150 stable relationships. Indiana University's analysis of Twitter confirmed: even users following thousands of accounts only actively interact with **100-200 people**. Following more than 150 accounts means you're consuming data streams, not maintaining relationships.
  12. - **Miller's Law (7±2)** — Cognitive psychology tells us our working memory holds ~7 items. The brain can deeply process only **5-9 quality sources per day**. Hundreds of subscriptions trigger decision fatigue, forcing your brain into shallow scanning mode.
  13. - **邓巴数(150)**—— 牛津大学人类学家邓巴证明,人类最多维持约 150 段稳定社交关系。印第安纳大学对 Twitter 的大数据分析确认:即使关注数千人,活跃互动圈依然卡在 **100-200 人**。超过 150 个关注,你消费的是数据流,而非社交关系。
  14. - **米勒定律(7±2)**—— 认知心理学表明,短时记忆容量约 7 个单位。大脑每天能深度消化的高质量信源通常不超过 **5-9 个**。关注几百个账号的结果是"决策疲劳",大脑放弃深度处理,转为浅层扫描。
  15. ### The Coming Storm | 即将到来的风暴
  16. With AIGC's marginal cost approaching zero, the internet is heading toward a reality where **90%+ of content is AI-generated**. The "Dead Internet Theory" is becoming fact. Social platforms are splitting from "Social Media" into "Recommendation Media" — AI feeds content, humans secrete dopamine. Real human connection is retreating into private, verified circles (**Dark Forest socialization**).
  17. 随着 AIGC 边际成本趋零,互联网正走向 **90% 以上内容由 AI 生成**的现实。"死互联网理论"正在成真。社交平台正从"社交媒体"裂变为"推荐媒体"——AI 负责投喂,人类负责分泌多巴胺。真人社交正撤退至私密的、经过验证的小圈子(**黑暗森林化**)。
  18. ### The Solution | 解决方案
  19. **Take back control.** Stop handing your attention to "guess what you like" algorithms. Build your own information moat:
  20. **夺回控制权。** 别再把注意力交给"猜你喜欢"。建立你自己的信息护城河:
  21. > 🎯 Curate ≤150 accounts → 📡 Let RSS pull updates → 🤖 Let AI filter noise → ☕ Reclaim your 2 hours
  22. >
  23. > 🎯 精选 ≤150 个关注 → 📡 让 RSS 拉取更新 → 🤖 让 AI 过滤噪音 → ☕ 夺回你的 2 小时
  24. **From FOMO to JOMO** — embrace the joy of missing out. 99% of information is noise. Your attention is the last scarce resource in the age of AI.
  25. **从 FOMO 到 JOMO** —— 拥抱「错过的快乐」。99% 的信息都是噪音。在 AI 时代,你的注意力是最后的稀缺资源。
  26. <details>
  27. <summary>📚 References | 参考文献</summary>
  28. 1. Dunbar, R. I. M. (1992). *Neocortex size as a constraint on group size in primates.* Journal of Human Evolution, 22(6), 469–493.
  29. 2. Gonçalves, B., Perra, N., & Vespignani, A. (2011). *Modeling Users' Activity on Twitter Networks: Validation of Dunbar's Number.* PLoS ONE, 6(8), e22656. (Indiana University)
  30. 3. Miller, G. A. (1956). *The Magical Number Seven, Plus or Minus Two.* Psychological Review, 63(2), 81–97.
  31. 4. Hunt, M. G., Marx, R., Lipson, C., & Young, J. (2018). *No More FOMO: Limiting Social Media Decreases Loneliness and Depression.* Journal of Social and Clinical Psychology, 37(10), 751–768. (UPenn)
  32. 5. Kemp, S. (2024). *Digital 2024: Global Overview Report.* DataReportal / We Are Social / Meltwater.
  33. 6. GWI (2024). *Social Media Trends Report.* GlobalWebIndex.
  34. </details>
  35. ---
  36. ## 🤖 Designed for AI IDEs | 为 AI IDE 设计
  37. This Skill is built to be used with AI-powered coding environments:
  38. 本 Skill 设计为配合以下 AI 编程环境使用:
  39. - **[Claude Code](https://claude.ai/code)** — Anthropic's AI coding agent (recommended)
  40. - **[Cursor](https://cursor.sh)** — AI-first code editor
  41. - **[Windsurf](https://codeium.com/windsurf)** — AI-powered IDE by Codeium
  42. - **[OpenClaw](https://github.com/nicepkg/openclaw)** — Open-source Claude Code alternative
  43. Just share this repo's URL with your AI assistant, and it will read `SKILL.md` to understand how to manage your RSS subscriptions automatically.
  44. 只需将本仓库 URL 分享给你的 AI 助手,它会读取 `SKILL.md` 并自动帮你管理 RSS 订阅。
  45. ---
  46. ## ✨ Features | 功能特性
  47. | Feature | 功能 | Description |
  48. |---------|------|-------------|
  49. | 📡 7-Platform Support | 7 平台支持 | YouTube, Vimeo, Behance, Bilibili, Weibo, Douyin, Xiaohongshu |
  50. | 🔄 Incremental Updates | 增量更新 | SQLite-based dedup, only fetches new content |
  51. | 📋 Digest Mode | 摘要模式 | `--digest` shows latest 1 item per account |
  52. | 🤖 AI Classification | AI 分类 | Optional Claude-powered topic categorization |
  53. | 📰 Standard Output | 标准输出 | RSS 2.0 XML + Markdown reports |
  54. | ⏰ Schedulable | 可定时 | Works with crontab for automated updates |
  55. ---
  56. ## 🚀 Quick Start | 快速开始
  57. ### Install | 安装
  58. ```bash
  59. # Core dependencies | 核心依赖
  60. pip install httpx yt-dlp
  61. ```
  62. ### Basic Usage | 基本使用
  63. ```bash
  64. # Add subscriptions | 添加订阅
  65. python3 scripts/lets_go_rss.py --add "https://www.youtube.com/@MatthewEncina"
  66. python3 scripts/lets_go_rss.py --add "https://vimeo.com/xkstudio"
  67. python3 scripts/lets_go_rss.py --add "https://www.behance.net/yokohara6e48"
  68. # Update all | 更新全部
  69. python3 scripts/lets_go_rss.py --update --no-llm
  70. # Digest mode (1 item per account) | 摘要模式(每账号 1 条)
  71. python3 scripts/lets_go_rss.py --update --no-llm --digest --skip-setup
  72. # Read cached report (bot push) | 读取缓存报告(Bot 推送)
  73. python3 scripts/lets_go_rss.py --status
  74. # List subscriptions | 查看订阅
  75. python3 scripts/lets_go_rss.py --list
  76. ```
  77. ---
  78. ## 🏗️ Architecture | 架构
  79. ```
  80. ┌──────────────────────────────────────────────────┐
  81. │ Tier 1: Native RSS (zero dependency) │
  82. │ Vimeo / Behance → httpx reads RSS directly │
  83. ├──────────────────────────────────────────────────┤
  84. │ Tier 1b: yt-dlp (pip install) │
  85. │ YouTube → yt-dlp extracts metadata │
  86. ├──────────────────────────────────────────────────┤
  87. │ Tier 2: RSSHub Proxy (optional Docker) │
  88. │ Weibo / Douyin / Bilibili / XHS → local RSSHub │
  89. └──────────────────────────────────────────────────┘
  90. ```
  91. ## 📊 Platform Support | 平台支持
  92. | Platform | Method | Dependency | Ready? |
  93. |----------|--------|------------|:------:|
  94. | YouTube | yt-dlp | `pip install yt-dlp` | ✅ |
  95. | Vimeo | Native RSS | `httpx` | ✅ |
  96. | Behance | Native RSS | `httpx` | ✅ |
  97. | Weibo 微博 | RSSHub | Docker | ⚠️ |
  98. | Douyin 抖音 | RSSHub | Docker | ⚠️ |
  99. | Bilibili B站 | RSSHub | Docker | ⚠️ |
  100. | Xiaohongshu 小红书 | RSSHub | Docker | ⚠️ |
  101. ---
  102. ## 🇨🇳 Chinese Platforms Setup | 中国平台配置
  103. For Weibo, Douyin, Bilibili, and Xiaohongshu, you need a self-hosted [RSSHub](https://docs.rsshub.app/):
  104. 使用微博、抖音、B站、小红书需要自建 [RSSHub](https://docs.rsshub.app/):
  105. ```bash
  106. docker run -d --name rsshub -p 1200:1200 diygod/rsshub:chromium-bundled
  107. export RSSHUB_BASE_URL="http://localhost:1200"
  108. # Optional: tighter network timeout for bot timeout limits
  109. export RSS_HTTP_TIMEOUT="10"
  110. export RSS_HTTP_RETRIES="2"
  111. export RSS_XHS_TIMEOUT="6"
  112. export RSS_XHS_RETRIES="1"
  113. export RSS_YTDLP_TIMEOUT="12"
  114. ```
  115. ---
  116. ## 📂 Project Structure | 项目结构
  117. ```
  118. lets-go-rss/
  119. ├── SKILL.md # Claude Skill entry point | AI 技能入口
  120. ├── README.md # This file | 本文件
  121. ├── requirements.txt # Python deps | Python 依赖
  122. ├── scripts/
  123. │ ├── lets_go_rss.py # Main entry | 主入口
  124. │ ├── rss_engine.py # Core engine | 核心引擎
  125. │ ├── scrapers.py # Platform scrapers | 平台爬虫
  126. │ ├── database.py # SQLite manager | 数据库
  127. │ ├── classifier.py # AI classification | AI 分类
  128. │ ├── rss_generator.py # XML generation | XML 生成
  129. │ ├── report_generator.py # Markdown reports | 报告生成
  130. │ ├── run_update_cron.sh # Stable update command | 稳定更新命令
  131. │ └── run_status_push.sh # Stable status command | 稳定推送命令
  132. └── assets/ # Runtime data (gitignored) | 运行时数据
  133. ```
  134. ## ⏰ Scheduled Updates | 定时更新
  135. ```bash
  136. # Recommended stable commands
  137. cd /path/to/lets-go-rss && ./scripts/run_update_cron.sh
  138. cd /path/to/lets-go-rss && ./scripts/run_status_push.sh
  139. # crontab -e — update at :55 every 2 hours, push at every even hour
  140. 55 */2 * * * cd /path/to/lets-go-rss && ./scripts/run_update_cron.sh >> /tmp/rss_cron.log 2>&1
  141. 0 */2 * * * cd /path/to/lets-go-rss && ./scripts/run_status_push.sh
  142. ```
  143. The engine now uses `assets/.update.lock` to prevent overlapping update jobs.
  144. ## 🤝 AI Classification (Optional) | AI 分类(可选)
  145. ```bash
  146. pip install anthropic
  147. export ANTHROPIC_API_KEY="your-key"
  148. # Update with AI classification | 使用 AI 分类更新
  149. python3 scripts/lets_go_rss.py --update
  150. ```
  151. ## License
  152. MIT

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.