deepelementlab/open-krowork/summarize

KroWork: Summarize

Fetch web content and use Claude's AI capabilities to summarize, analyze, or translate it.

How It Works

This skill combines KroWork's web scraping with Claude's AI abilities:

1. KroWork handles content fetching and preprocessing (cleaning, deduplication, extraction)
2. Claude handles the AI part (summarization, analysis, translation)

No external LLM API is needed — Claude IS the AI engine.

Workflow

Step 1: Understand the Request

The user input $ARGUMENTS may contain:

• A URL to summarize
• A description of what analysis they want (summarize, key points, translate, compare, etc.)
• Multiple URLs to compare

Determine the action:

• Summarize: "帮我总结这篇文章" → Provide a concise summary
• Key points: "这篇文章的重点是什么" → Extract and list key takeaways
• Translate: "翻译成中文/英文" → Translate the content
• Analyze: "分析这个产品的优缺点" → Deep analysis
• Compare: "对比这两个页面" → Compare multiple sources
• Monitor + Summarize: "每天帮我总结这个网站的更新" → Set up monitoring

Step 2: Fetch and Preprocess Content

Use krowork_preprocess_content to get clean, structured content:

• url: The target URL
• max_length: Adjust based on need (default 8000, increase for deep analysis)

This returns:

• title: Page title
• description: Meta description
• key_points: Headings extracted from the page
• clean_text: Noise-free main content
• stats: Content statistics

For RSS feeds, use krowork_scrape_rss instead.

Step 3: Apply AI Analysis

Using the preprocessed content, apply Claude's AI capabilities:

• For summaries: Read the clean_text and produce a structured summary
• For analysis: Combine key_points with clean_text for deeper insights
• For translation: Translate the clean_text to the requested language
• For comparison: Fetch multiple URLs, then compare side by side

Step 4: Present Results

Format the AI output clearly:

• Use headers and bullet points for readability
• Include the source URL
• Note the content date if available
• Highlight key findings

Step 5: Suggest Follow-up

• "Create a monitoring app" → /krowork:create
• "Track this page for changes" → krowork_monitor_page
• "Save as data source" → /krowork:datasource

Tips for Best Results

• For long articles, increase max_length to get more content
• For news sites, use krowork_scrape_rss for structured feed data
• For comparison, fetch all URLs first, then analyze together
• The key_points from preprocessing help Claude focus on the right sections