Recently, OpenClaw (nicknamed 'Lobster' in China) has become incredibly popular.
I won't even mention the Mac minis that sold out because of OpenClaw.
Nor do I need to mention how domestic cloud host providers are jumping on the trend, supporting one-click deployment and installation of OpenClaw.
Major domestic model manufacturers have launched Coding Plans just to let everyone connect to OpenClaw.
Recently, there have even been door-to-door paid installation services, which have now become so competitive that they are being offered for free.
For example, today Tencent arranged for 20 technical staff to install OpenClaw for everyone for free, and the scene was extremely popular.
My intuitive feeling: Recently, every invitation and event is related to OpenClaw. It's insane!
I'm not trying to throw cold water on this.
I believe that the hotter OpenClaw gets, the more ordinary people should settle down and polish their Skills.
Otherwise, installing OpenClaw won't be of much use.
Below, I recommend several Skills written by myself and other netizens that can be called by OpenClaw in the future, just to get the ball rolling.
What are the Common Skills?
I've found that most people create their own set of information scraping and collection Skills.
The reason is simple: no matter how smart AI is, you have to feed it something first.
Below, I will introduce them along three lines: "Scraping & Collection," "Content Creation," and "Efficiency Tools."
I. Scraping & Collection Skills
1. Agent Reach — Giving AI Eyes
Repository:
In one sentence: Zero API cost, allowing AI Agents to access the entire internet.
Web scraping, YouTube subtitle extraction, Twitter/X search, GitHub access, Reddit parsing, Bilibili, Xiaohongshu, Douyin, WeChat Official Accounts, RSS feeds, semantic search... basically every information source you can think of is covered.
Highlights:
- Uses entirely free open-source backends; no need to apply for individual API keys.
- Local credential storage; cookies/tokens are not sent externally.
- Supports Chinese platforms (Xiaohongshu, Douyin, WeChat).
- Built-in diagnostic tool
agent-reach doctorto troubleshoot environment issues with one click.
Installation: Tell the AI Agent:
"Help me install Agent Reach:
https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md "
Requires basic environments like Python, yt-dlp, and gh CLI. Chinese platforms (like Xiaohongshu) require Docker to run the MCP service.
2. Defuddle — Web Content Extraction Tool
In one sentence: Extract clean article content from web pages, removing ads, sidebars, and other clutter.
A command-line tool written by the CEO of Obsidian, which I encapsulated into a Skill.
You tell the AI, "Help me extract the content of this link," and it automatically calls Defuddle, returning clean Markdown text plus metadata like title, author, publish date, and word count.
Installation:
npx skills add joeseesun/defuddle-skill
3. YouTube Search & Download: The First Step to Video Transcription
Github:
In one sentence: YouTube site-wide search + video download + subtitle extraction, all in one place.
Supports searching by date/view count/relevance, channel browsing, and searching within channels.
Multi-quality video downloads (up to 4K), MP3 audio extraction, and subtitle retrieval (SRT with timestamps + TXT plain text).
English titles are also automatically translated into Chinese.
Typical Usage:
- Search for the latest videos on a specific topic.
- Download videos and extract subtitles for subsequent content creation (writing long articles, tweets, etc.).
- Extract only audio for podcast material.
Installation:
npx skills add joeseesun/yt-search-download
Prerequisites: Apply for a free YouTube API Key + yt-dlp (brew install yt-dlp).
Note: Update yt-dlp frequently and use a clean IP.
4. Anything to NotebookLM — Process Everything with NotebookLM
In one sentence: Throw any content (WeChat articles, YouTube videos, PDFs, EPUBs, etc., in 15+ formats) into Google NotebookLM to automatically generate podcasts, PPTs, mind maps, quizzes, etc.
Integrates several open-source projects, such as my friend Tenglin's NotebookLM-py and Microsoft's Markitdown.
This Skill bridges the entire chain from "content acquisition" to "NotebookLM output."
You can say, "Turn this WeChat article into a podcast," and it automatically completes the scraping → conversion → upload → generation.
Supported Input Formats: WeChat Official Accounts, YouTube, PDF, EPUB, web pages, Office documents, images, audio...
Supported Output Formats: Podcasts, PPTs, mind maps, quizzes, reports, videos, infographics.
Installation: Clone the repository and run the installation script—
git clone
https://github.com/joeseesun/anything-to-notebooklm.git cd anything-to-notebooklm ./install.sh
II. Content Creation Skills
5. Teacher Baoyu's Skill Collection — The Content Creation Suite
Repository:
Teacher Baoyu's (@dotey) Skills collection is essentially a "one-person content factory."
It covers the entire chain from graphic creation to social media publishing:
Visual Content Generation:
- Xiaohongshu Infographics: Multiple styles × multiple layout customizations.
- General Infographic Generator: 20 layouts + 17 visual styles.
- Cover Image Tool: 5-dimensional design system (type, color scheme, rendering, texture, typography).
- Slide Creator: 14+ style presets.
- Comic generation, article illustrations.
Social Media Publishing:
- X (Twitter) publishing.
- WeChat Official Account publishing.
- Xiaohongshu auto-publishing.
Content Processing Tools:
- Markdown formatting and conversion.
- Image compression (WebP/PNG).
- DeepL translation.
- URL to Markdown.
Installation:
npx skills add jimliu/baoyu-skills
This collection is particularly suitable for self-media creators, as one Skill package solves all needs from production to distribution.
6. Markdown One-Click X Article Publisher
Repository:
In one sentence: Write in Markdown and publish as an X (Twitter) Articles draft with one click.
Supports full Markdown formatting (titles, bold/italic, lists, quotes, code blocks, links, images), automatically handles image uploads, and features 7-day session persistence.
Installation:
git clone
https://github.com/joeseesun/qiaomu-x-article-publisher.git ~/.claude/skills/qiaomu-x-article-publisher pip install Pillow pyobjc-framework-Cocoa patchright python auth_manager.py setup
7. Knowledge Site Creator — Generate a Learning Site in One Sentence
Repository:
In one sentence: Tell the AI what you want to learn, and it automatically generates a complete learning website and deploys it online.
For example, if you say, "Help me create a website for learning evolutionary psychology."
The AI automatically completes theme analysis → content creation → page design → Vercel deployment, without you writing a single line of code.
Learning Modes: Flashcards, progressive learning, quizzes, index, progress tracking.
Technical Features:
- PWA support, works offline.
- SEO optimized, includes Meta tags and sitemaps.
- Zero frontend dependencies, native HTML/CSS/JS.
- Minimalist yellow theme, clean and fresh.
Installation:
npx skills add joeseesun/knowledge-site-creator
III. Efficiency Tool Skills
8. Spotify Music Player — Listen to Music with Natural Language
In one sentence: Control Spotify with natural language, featuring a database of 5,947 music styles.
Tell the AI, "Play some music suitable for coding" or "Play Bohemian Rhapsody," and it automatically searches, matches, and plays.
Supports search, play, pause, skip, volume adjustment, queue management, and recommendations based on scene/mood.
Highlights:
- 5,947 music styles, hierarchically organized.
- 30+ style quick-play options.
- Natural language mapping to specific styles.
- Zero external dependencies, pure Python standard library.
- Automatic OAuth token refresh.
Installation:
npx skills add joeseesun/qiaomu-music-player-spotify
Requires a Spotify Premium account.
For specific usage and configuration, see:
https://github.com/joeseesun/qiaomu-music-player-spotify
9. Design Advisor — Jobs-style Design Consultant
A UI design Skill generated using my own prompt.
The results often provide unexpected surprises.
In one sentence: A UI/UX design consultant that blends Steve Jobs' product intuition + Dieter Rams' functional purism.
It's not the kind of perfunctory advice like "change the color here."
It digs deep into the real user needs behind surface requirements, scrutinizing every detail (spacing, color temperature, animation timing).
It provides three levels of solutions for every problem (progressive improvement, structural redesign, ideal solution) and transparently displays trade-offs.
Triggers: "redesign", "review UI", "optimize interaction experience".
Installation:
npx skills add joeseesun/qiaomu-design-advisor
IV. Skill Management and Discovery
Once you've written a Skill, how do you publish it?
The best way is to manage it with Git and even publish it to GitHub for sharing (or keep it in a private repo).
I wrote a Skill to help friends who aren't familiar with this—Skill Publisher:
Repository:
It automatically completes: validating SKILL.md metadata → creating a GitHub repository → pushing code → verifying it can be installed via npx skills add.
npx skills add joeseesun/skill-publisher
Requires GitHub CLI (gh) to be installed and authenticated.
Where can you find more Skills?
I recommend three channels:
1. Skills.sh — Vercel's Official Skill Directory
Vercel's open-source Skills directory, featuring over 86,000+ Skills.
Supports 20+ platforms (Claude Code, GitHub Copilot, Cursor, Cline, Gemini, etc.), filterable by popularity and trends.
2. Find Skills — Use a Skill to Find Skills
After installing this Skill, you can search for and install other Skills directly in the terminal.
It can be called a "Meta-Skill."
npx skills add vercel-labs/skills/find-skills
Then you can search using commands like npx skills find react performance.
3. SkillsMP — The Largest Skill Marketplace
A community-driven Skills aggregation platform with over 380,000+ Skills, supporting a Chinese interface.
It automatically scrapes and syncs from public GitHub repositories with basic quality filtering (minimum 2 stars threshold).
Final Thoughts
Skills are the soul of OpenClaw.
OpenClaw without Skills is like a phone without apps.
The hotter OpenClaw gets, the more you should settle down and polish your own Skills.
Instead of chasing the trend to install OpenClaw, think clearly first: what do you want the AI to do for you?
Once you've figured that out, you'll naturally know how to write the Skills.
