Lesson 10 — OpenClaw Summarize Skill: One-Command Summaries of Any Web Page, PDF, or YouTube Video (2026)
Goal: Install the Summarize Skill so OpenClaw can summarize any web page, PDF document, YouTube video, or audio file with a single command.
What Is the Summarize Skill?
Summarize is ClawHub's #2 most downloaded Skill (221k downloads). Built on the summarize CLI tool, it supports the following content types:
| Content Type | Examples |
|---|---|
| Web page URL | Any webpage, news article, technical documentation |
| PDF file | Local PDF or an online PDF link |
| YouTube video | Automatically extracts captions and summarizes |
| Images | Screenshots, charts, scanned documents |
| Audio files | mp3, wav, meeting recordings |
Key advantage: No API Key required — it runs on the models you've already configured in OpenClaw, at zero extra cost.
Step 1: Install the Dependency CLI Tool
The Summarize Skill depends on the summarize CLI. Install it first:
npm install -g @steipete/summarize-cliVerify the installation:
summarize --version
# Should output a version number like 1.2.0Step 2: Install the Skill
In WebChat or Telegram, send:
/install @steipete/summarize
Verify:
pnpm openclaw skills list
# summarize should appear in the listStep 3: Summarize a Web Page
Send any URL with a summarize instruction:
Summarize this article for me: https://example.com/article
Or use the slash command:
/summarize https://example.com/article
Sample output:
Key Points
- OpenClaw's native MCP support lets it connect to 500+ tools, making it the self-hosted AI framework with the broadest integration
- The new version reduced latency by 40%, mainly from gateway-layer optimizations
- The team plans to release native Windows support in Q2
Conclusion: This is a developer-experience-focused release with solid upgrade value for production environments.
Step 4: Summarize a Local PDF
Pass the PDF path to AI:
Summarize this PDF: ~/Downloads/report.pdf
Or summarize an online PDF (pass the URL directly):
/summarize https://arxiv.org/pdf/2501.12345
Ideal use cases:
- Quickly skim an academic paper to decide if it's worth reading in full
- Extract key clauses from a contract or report
- Compress lengthy technical documentation into action points
Step 5: Summarize a YouTube Video
Just send the YouTube link:
/summarize https://www.youtube.com/watch?v=xxxxx
The Skill will automatically extract video captions (supports English and other languages) and generate a structured summary.
What about videos without captions?
If a video has no captions, the Skill will first transcribe the audio using Whisper before summarizing (requires the Whisper Skill from Lesson 23).
Step 6: Summarize Image Content
Send an image path or URL:
What does this screenshot say? Give me a summary: ~/Desktop/meeting-screenshot.png
Good for:
- Quickly reading content from phone screenshots saved to your desktop
- Extracting table or chart content from scanned PDFs
- Pulling information out of product manual images
Step 7: Summarize an Audio File
/summarize ~/Downloads/meeting-recording.mp3
The Skill will transcribe first, then summarize, outputting:
- Main meeting topics
- Key decisions and conclusions
- Action item list
Customizing the Summary Style
You can add specific requirements to your instructions:
Summarize this English article in English with just 3 core points, each no more than 20 words:
https://example.com/article
Summarize this PDF, focusing on extracting numbers and data: ~/report.pdf
Turn this YouTube video into a short social-media-style post:
https://www.youtube.com/watch?v=xxxxx
FAQ
Can OpenClaw summarize PDFs? How?
Yes. After installing the Summarize Skill, just send the PDF's local path (like ~/Downloads/file.pdf) or the URL of an online PDF to AI with "summarize this PDF." No extra API Key needed — it runs on your already-configured model.
What formats does the Summarize Skill support?
Web pages (any URL), PDFs (local and online), YouTube videos (auto-extracts captions), images (PNG/JPG/GIF), and audio (mp3/wav/m4a). For videos without captions, you need the Whisper Skill to transcribe first.
Can I control the summary length?
Yes — specify it in your instructions, like "summarize in 3 bullet points," "under 200 words," or "a version I can read in 5 minutes." The Skill will honor your requirements.
Is there a word or page count limit for summarization?
The limit comes from your configured model's context window. MiniMax M2.1 supports a 200K token context, which covers most web articles and PDF reports. For extremely long content (like an entire book), we recommend summarizing chapter by chapter.
Will it automatically translate non-English content?
No automatic translation, but you can request it: Summarize this French article in English. AI will both summarize and translate.
Can it summarize a PDF that's not on the internet?
Yes — just send the local file path like ~/Desktop/file.pdf. No uploading to any cloud service; all processing happens locally.
Example Workflows
Daily information digest (with Lesson 06 scheduled tasks):
# Auto-summarize 3 subscribed articles every morning at 8 and push to Telegram
/summarize https://news.ycombinator.com
/summarize https://github.com/trendingAutomatic meeting minutes:
Convert this meeting recording into meeting minutes with: decisions, action items, owners:
~/Recordings/weekly-sync.mp3
Quick paper screening:
Tell me if this paper is worth reading in depth, focusing on methodology and experimental conclusions:
https://arxiv.org/pdf/2501.12345