home / skills / openclaw / skills / rss-ai-reader

rss-ai-reader skill

safe

This skill automates RSS subscriptions, generates Chinese summaries with LLM, and pushes updates to Feishu, Telegram, or Email.

npx playbooks add skill openclaw/skills --skill rss-ai-reader

Review the files below or copy the command above to add this skill to your agents.

Files (3)

SKILL.md

2.7 KB

---
name: rss-ai-reader
description: |
  📰 RSS AI 阅读器 — 自动抓取订阅、LLM生成摘要、多渠道推送！
  支持 Claude/OpenAI 生成中文摘要，推送到飞书/Telegram/Email。
  触发条件: 用户要求订阅RSS、监控博客、抓取新闻、生成摘要、设置定时抓取、
  "帮我订阅"、"监控这个网站"、"每天推送新闻"、RSS/Atom feed 相关。
---

# 📰 RSS AI Reader

自动抓取 RSS 订阅 → LLM 生成中文摘要 → 推送到 IM

## ✨ 核心功能

- 📡 自动抓取 RSS/Atom feeds
- 🤖 Claude/OpenAI 生成中文摘要
- 📬 多渠道推送：飞书、Telegram、Email
- 💾 SQLite 去重，不重复推送
- ⏰ 支持定时任务

## 🚀 快速开始

```bash
# 安装
git clone https://github.com/BENZEMA216/rss-reader.git ~/rss-reader
cd ~/rss-reader && pip install -r requirements.txt

# 配置（编辑 config.yaml）
cp config.yaml my_config.yaml
# 设置 feeds、LLM key、推送渠道

# 运行
python main.py --once              # 单次执行
python main.py                     # 启动定时任务
python main.py --stats             # 查看统计
```

## 📝 配置示例

```yaml
# RSS 订阅
feeds:
  - name: "Hacker News"
    url: "https://hnrss.org/frontpage"
    category: "tech"
  - name: "阮一峰周刊"
    url: "https://www.ruanyifeng.com/blog/atom.xml"
    category: "tech"

# LLM 配置
llm:
  provider: "claude"  # 或 "openai"
  model: "claude-sonnet-4-20250514"
  api_key: "${ANTHROPIC_API_KEY}"

# 推送到飞书
notify:
  feishu:
    enabled: true
    webhook_url: "${FEISHU_WEBHOOK}"
```

## 📬 支持的推送渠道

| 渠道 | 配置项 | 说明 |
|------|--------|------|
| **飞书** | `webhook_url` | 群机器人 Webhook |
| **Telegram** | `bot_token` + `chat_id` | Bot API |
| **Email** | SMTP 配置 | 支持 Gmail 等 |

## 🔧 命令行参数

```bash
python main.py [options]

--config, -c  配置文件路径 (默认: config.yaml)
--once        只执行一次
--stats       显示统计信息
--db          数据库路径 (默认: rss_reader.db)
```

## 💡 使用场景

1. **技术博客监控** — 订阅 HN、阮一峰、V2EX 等
2. **新闻早报** — 每天定时推送摘要到飞书群
3. **竞品监控** — 订阅竞品博客，自动摘要
4. **论文追踪** — 订阅 arXiv，AI 帮你筛选

## 📊 输出效果

飞书消息示例：
```
📰 Hacker News

**Why SQLite is Taking Over**

📝 SQLite 正在从嵌入式数据库扩展到更多应用场景。
文章分析了其在边缘计算、移动应用中的优势...

[🔗 阅读原文]
```

---

## ☕ 支持作者

- **GitHub Sponsors**: [@BENZEMA216](https://github.com/sponsors/BENZEMA216)
- **Buy Me a Coffee**: [buymeacoffee.com/benzema216](https://buymeacoffee.com/benzema216)

Overview

This skill automates subscribing to RSS/Atom feeds, generates concise Chinese summaries using Claude or OpenAI, and delivers updates to Feishu, Telegram, or Email. It deduplicates items with SQLite and supports scheduled polling so you receive only fresh, summarized content. The tool is lightweight, configurable, and designed for continuous monitoring or one-off runs.

How this skill works

The skill fetches configured RSS/Atom feeds on a schedule or on demand, parses new items, and checks a local SQLite store to avoid duplicate pushes. For each new item it calls the chosen LLM (Claude or OpenAI) to produce a Chinese summary and optional highlights. Finally it formats the result and sends the message to one or more configured notification channels (Feishu webhook, Telegram bot, SMTP email).

When to use it

You want automated daily or hourly news briefings in a team chat.
You need to monitor blogs, competitor sites, or arXiv feeds and get concise summaries.
You want deduplicated feed alerts to multiple channels (Feishu/Telegram/Email).
You prefer AI-generated Chinese summaries instead of raw article text.
You need a lightweight scheduler to poll feeds without maintaining a full pipeline.

Best practices

Limit feed frequency and use sensible polling intervals to avoid rate limits.
Configure meaningful feed names and categories to keep notifications organized.
Rotate and secure LLM API keys and webhook credentials; avoid embedding secrets in code.
Tune LLM prompts for the desired summary length and tone (bullet points or narrative).
Enable database backups for the SQLite file to preserve seen-item history.

Example use cases

Daily tech digest: send a morning summary of Hacker News and key blogs to a Feishu group.
Research monitoring: track arXiv categories and push summarized papers to your private Telegram channel.
Competitive intelligence: subscribe to competitor blogs and email summarized changes to product teams.
Personal reading: aggregate favorite blogs and receive one concise email with highlights each day.

FAQ

Which LLM providers are supported?

Claude and OpenAI are supported; choose provider and model in the configuration.

How does duplicate prevention work?

A local SQLite database stores seen item IDs/urls so items already pushed are skipped.