OpenLark

---
name: news-express
description: Use this skill when users ask for news updates, daily briefings, or what's happening in the world.
---

# News Express

## Overview

Use this skill when users ask for news updates, daily briefings, or what's happening in the world. Fetches news from reliable international and Chinese RSS feeds, **requiring no API keys**. Simply use the `web_fetch` tool to read RSS XML directly.

## Trigger Scenarios

- User asks "what's the news today", "latest updates", "daily briefing"
- User asks "what's happening domestically/internationally"
- User requests categorized news such as technology, finance, sports
- User asks for "morning briefing", "evening briefing", "news summary"

## RSS Feeds

### 🇨🇳 Domestic Sources

| Source | Category | URL |
|------|------|-----|
| 36Kr | Latest Flash News | `https://36kr.com/feed-newsflash` |
| Solidot | Technology | `https://www.solidot.org/index.rss` |
| SSPAI | Technology/Digital | `https://sspai.com/feed` |
| IThome | AI/Technology/Digital | `https://www.ithome.com/rss/` |
| Zhidx | Flash News/Headlines/AI/Robotics | `https://zhidx.com/rss` |
| ITJuzi | Finance | `https://www.itjuzi.com/api/telegraph.xml` |
| cnBeta | AI/Technology/Digital | `https://plink.anyfeeder.com/cnbeta` |
| Odaily | Flash News | `https://rss.odaily.news/rss/newsflash` |
| People's Daily | Headlines | `http://www.people.com.cn/rss/politics.xml` |

### 🌍 International Sources

| Source | Category | URL |
|------|------|-----|
| OpenAI | AI | `https://openai.com/news/rss.xml` |
| NPR | United States | `https://feeds.npr.org/1001/rss.xml` |
| Hacker News | Technology | `https://hnrss.org/frontpage` |
| PANews | Flash News | `https://www.panewslab.com/rss.xml?lang=en&type=NEWS&featured=true` |
| ArXiv | AI | `https://rss.arxiv.org/rss/cs.AI` |
| China Daily | English/International | `https://www.chinadaily.com.cn/rss/china_rss.xml` |
| TechCrunch | Technology | `https://techcrunch.com/feed/` |

## Workflow

### Step 1: Fetch RSS Content

Use the `web_fetch` tool to read RSS XML directly:

```
web_fetch(url="https://openai.com/news/rss.xml")
```

### Step 2: Parse Headlines

In the RSS XML structure, news headlines are within `<title>` tags, summaries within `<description>` tags, and links within `<link>` tags. Simply extract them from the markdown text returned by `web_fetch`.

### Step 3: Compile and Output

Organize by domestic and international categories, taking 6-8 items each, outputting each headline with a concise summary.

## Output Format

```
📰 [Latest News]
🗓️ [Date] · [Day of Week]

---

## 🇨🇳 Domestic

• [Headline 1] 
[Summary 1]
• [Headline 2] 
[Summary 2]
• [Headline 3] 
[Summary 3]
...

---

## 🌍 International

• [Headline 1] 
[Summary 1]
• [Headline 2] 
[Summary 2]
• [Headline 3] 
[Summary 3]
...

---
Data Source: RSS Feeds
```

## Important Notes

- **No API Key Required**: All data is obtained through public RSS feeds
- **Some domestic sources may require proper network access**: If access fails, automatically switch to backup sources
- **Content Timeliness**: RSS typically updates every 15-60 minutes
- **Language**: Domestic sources output in Chinese, international sources can be bilingual (Chinese/English)

---
name: xiaohongshu-image-auto
description: After users provide a title and body text, automatically completes the entire process: login detection, AI-generated images, content filling, AI-generated tags, tag activation, original content declaration, and publishing.
---

# Xiaohongshu Automatic Image-Text Post Publishing

## Feature Overview

After users provide a title and body text, automatically completes the entire process: login detection, AI-generated images, content filling, AI-generated tags, tag activation, original content declaration, and publishing. Only QR code login requires manual operation; all other steps are fully automated.

## Trigger Conditions

Trigger this skill when the user mentions any of the following keywords:
- Xiaohongshu image-text post publishing
- Publish Xiaohongshu image-text post
- Automatically publish Xiaohongshu image-text post
- Xiaohongshu image-text notes
- 小红书图文发布
- 发布小红书图文
- 自动发小红书图文
- 小红书图文笔记


## Use Cases
- Users have existing note content ready to publish on Xiaohongshu.

## Workflow
```
You provide content → Launch browser → Check login → AI generate images → Auto-fill → AI generate tags → Activate tags → Declare original → Publish note
            ↑
    Only QR code login requires manual operation
```

## Prerequisites
### Browser Environment
- Start the browser using host mode
- Open the Xiaohongshu Creator Platform: https://creator.xiaohongshu.com/publish/publish

### Account Login
- **Only manual QR code login is required**
- Automatically pauses and prompts when not logged in is detected
- Automatically continues execution after successful login

## Automated Execution Steps

### 1. Launch Browser and Check Login Status
```bash
browser start profile=openclaw target=host
browser navigate https://creator.xiaohongshu.com/publish/publish
browser snapshot refs=aria
```
**Login Detection Logic**:
- Check if the page contains a login form
- If not logged in, take a screenshot and prompt the user to scan the QR code
- Poll every 15 seconds automatically; continue after login

### 2. Switch to Image-Text Publishing Mode
```bash
browser act ref=<Upload Image-Text Button> kind=click
```

### 3. AI Automatically Generates Images
```bash
# Click the "Text to Image" button
browser act ref=<Text to Image Button> kind=click

# Enter image description
browser act ref=<Text Box> kind=type text="<User Title>"

# Generate image and select style (default "Basic")
browser act ref=<Generate Button> kind=click
browser act ref=<Style Button> kind=click
browser act ref=<Next Button> kind=click
```

### 4. Automatically Fill Title and Body Text
```bash
browser act ref=<Title Input Field> kind=type text="<User Title>"
browser act ref=<Body Input Field> kind=type text="<User Body Text>"
```
**Character Count Check**:
- Title exceeds 20 characters → Automatically truncate and prompt
- Body text exceeds 1000 characters → Automatically truncate and prompt

### 5. AI Generate Tags (if user does not provide)
```bash
# Analyze title and body text, generate 5-10 relevant tags
# Tag types: Core topics, Scene tags, Emotion tags
```
**Generation Rules**:
- Extract core keywords from the title
- Extract high-frequency words from the body text
- Supplement with popular related tags

### 6. Activate Topic Tags

**Key Rule**: Must click the first item in the pop-up recommendation list for the tag to become a searchable link

```bash
# Execute the following process for each tag
for each tag in tag list:
    # Step 1: Click the topic button to open the input field
    browser act ref=<Topic Button> kind=click
    
    # Step 2: Enter the tag name (including # symbol)
    browser act ref=<Topic Input Field> kind=type text="#tagname"
    
    # Step 3: Wait for the recommendation pop-up to appear (approx. 1 second)
    browser snapshot refs=aria
    
    # Step 4: Click the first item in the recommendation list (must!)
    browser act ref=<First Item in Recommendation List> kind=click
    
    # Step 5: Confirm the tag is activated (displayed as a blue link)
    # Continue to the next tag
```

**Activation Success Indicator**:
- Tag becomes a clickable blue link
- Format: `[Topic]#tagname`

**Important Notes**:
- ⚠️ Must click the recommendation pop-up; direct input is invalid
- ⚠️ Must click the first item to ensure tag standardization
- ⚠️ Each tag needs to be activated individually

### 7. Check Original Content Declaration
```bash
browser act ref=<Original Declaration Checkbox> kind=click
browser act ref=<Agree Checkbox> kind=click
browser act ref=<Declare Original Button> kind=click
```

### 8. Execute Publishing
```bash
browser act ref=<Publish Button> kind=click
browser snapshot refs=aria
```
**Publishing Success Indicator**: Page displays "Published Successfully" or redirects to the "Note Management" page

## Exception Handling

| Issue | Handling Method |
|------|----------|
| Login timeout | Take screenshot, poll automatically to detect login status |
| Title too long | Automatically truncate to 20 characters, prompt user to confirm |
| Publishing failed | Take screenshot record, prompt user to retry or escalate to manual |

## Key Considerations
### Account Security
- Only QR code login requires manual operation
- Recommended publishing interval ≥ 30 minutes
- Automatically check content compliance before publishing

### Content Compliance
- Title ≤ 20 characters, Body text ≤ 1000 characters
- Avoid absolute terms, medical efficacy claims, price inducement language

### Recommended Publishing Frequency
- New accounts: Daily or every other day
- Established accounts: Adjust based on follower activity times

## Quick Command Reference
```bash
browser navigate <URL>        # Open a page
browser snapshot refs=aria     # Get page elements
browser act ref=<Element> kind=click  # Click an element
browser act ref=<Element> kind=type text="Content"  # Input text
browser screenshot             # Take a screenshot
```

## Example
```
使用 xiaohongshu-image-auto 技能。汇总今日国内AI领域重要发布。
```