This guide covers everything you need to install, manage, and get the most out of OpenClaw skills on your xCloud VPS, from your first ClaWHub install to advanced skill management and security best practices.
A Quick Overview / TL;DR
Prerequisites: – An active xCloud OpenClaw Hosting plan – SSH access to your VPS (or Telegram/WhatsApp connected to your agent) – An OpenClaw agent running as a systemd service (this is the default on xCloud)
| If you want to… | Do this | Time required |
| Install your first skill | Follow Step 1 below | 2 minutes |
| Browse and install from the marketplace | Follow Steps 1–2 | 5 minutes |
| Set up a fully skilled agent | Follow this entire guide | 20 minutes |
| Troubleshoot a skill that isn’t working | Jump to Troubleshooting | 5 minutes |
Best for beginners: Start with ClaWHub and the healthcheck skill. Both give you immediate value with zero complexity.
Best for power users: Stack 5–8 skills together, then explore the multi-agent setup guide to run specialized agents with different skill sets.
Best for security-conscious users: Read the Security checklist section before installing anything from a third-party publisher.
What Are Skills?
Skills are structured add-ons that teach your OpenClaw agent new workflows, tools, and behaviors. Think of them like apps for your AI agent; each one expands what your agent can do.
At a technical level, each skill is a SKILL.md file stored on your xCloud VPS. This file contains a structured set of instructions, written in plain Markdown, that tells your agent how to handle a specific task or workflow. When your agent starts a new conversation, it reads all installed skills and incorporates those capabilities into its context window.
Your OpenClaw agent ships with solid built-in capabilities (file management, web search, shell commands, browser automation), but skills let you go further. Want your agent to audit your server’s security? There’s a skill for that. Want it to search Reddit and YouTube for trending topics? There’s a skill for that too. Want it to manage its own memory and get smarter over time? Yes, there’s a skill for that.
Skills stack. Install more, and your agent can do more. There’s no hard limit on how many you can install, though keeping your skill set focused tends to produce better results than installing everything at once. Each skill adds to your agent’s context window, so quality matters more than quantity.
How Skills Work Under the Hood
Here’s the simple version of what happens:
- You install a skill: a SKILL.md file gets saved to a specific directory on your VPS.
- You start a new chat: your agent reads all installed SKILL.md files at the beginning of the conversation.
- Your agent gains new abilities, and the instructions in each skill become part of its working knowledge for that session.
- Skills persist across restarts because they’re files on your Linux server. They survive reboots, service restarts, and updates.
OpenClaw runs as a persistent systemd service on your xCloud VPS. The agent process stays alive continuously, but each new conversation is a fresh session where skills are loaded. This means you never lose your installed skills, and activating a new one is as simple as starting a new chat.
Step 1: Install the ClaWHub Skill First
ClaWHub is the community marketplace for OpenClaw skills. It’s where skill creators publish their work and where you’ll find vetted, ready-to-install skills for almost any workflow you can imagine.
The first skill you should install is the ClaWHub skill itself. This is the meta-skill: the skill that gives your agent the ability to browse clawhub.com, search for skills, and install them directly through conversation.
How to Install ClaWHub
Option A: During initial setup (recommended)
When you first set up OpenClaw on your xCloud server, the setup process will prompt you to install official skills. Select ClaWHub from that list. It’s typically the first option, and for good reason: without it, you’ll need to install everything else manually.
Option B: After setup via chat
If you skipped the initial skill selection, or if you’re adding ClaWHub to an existing agent, simply send this message to your agent via Telegram, WhatsApp, or whatever channel you’ve connected:
“Install the ClaWHub skill so you can find and install other skills from the marketplace.”
Your agent will download the skill from the official repository and save it to the correct directory.
Option C: Manual installation via SSH
If you prefer the command line, SSH into your xCloud VPS and run:
clawhub login
clawhub install clawhub
Without the ClaWHub skill, your agent can’t browse or install skills from the marketplace on its own. You’d need to manually download and place every SKILL.md file via SSH. Installing ClaWHub first saves you time on every future skill installation.
Step 2: Install Skills by Pasting a Link
Once ClaWHub is installed, adding new skills to your agent is straightforward. You have three options, and all of them work through regular conversation with your agent.
Method 1: Paste a ClaWHub URL
Browse the skill catalog at clawhub.com, find a skill you want, copy its URL, and send it to your agent:
“Install this skill for me: https://clawhub.com/skills/healthcheck”
Your agent will visit the page, read the skill details, download the SKILL.md file, and install it to the correct directory, automatically.
Method 2: Ask Your Agent to Search
Don’t know the exact skill name? Just describe what you need:
“Find me a skill that lets you search Reddit and YouTube for trending topics from the last 30 days.”
Your agent (with the ClaWHub skill installed) will search the marketplace, show you matching results, and ask which one you’d like to install.
Method 3: Install by Slug via CLI
If you know the skill’s slug (its short identifier), you can install directly from the command line:
clawhub install healthcheck
Or tell your agent:
“Install the healthcheck skill from ClaWHub.”
Installing Multiple Skills at Once
You can install several skills in a single request:
“Install these skills from ClaWHub: healthcheck, last-30-days, and tmux.”
Your agent will process each one sequentially, confirming each installation as it completes.
The flexibility here is intentional. Whether you’re a point-and-click person or a command-line user, there’s a path that fits your workflow. Skill installation works like installing an app on your phone.
Step 3: Start a New Chat to Activate Your Skills
After installing a skill, you need to start a fresh conversation for it to take effect.
This isn’t a bug. It’s how OpenClaw works by design. Skills are read at the beginning of each new chat session. Your agent loads all installed SKILL.md files when a conversation starts, and those instructions stay active for the duration of that session.
How to Activate
- On Telegram: Start a new message thread or use the /new command if your agent supports it.
- On WhatsApp: Send a message that starts a new session context.
- Via the OpenClaw dashboard: Click “New Chat” or “New Session.”
How to Confirm
Once you’ve started a new chat, verify that your skills are active:
“List all your installed skills.”
Your agent will respond with a list of every skill it loaded for the current session, including version numbers and brief descriptions.
This design keeps your agent fast. Instead of checking for new skills on every single message (which would add latency), it loads everything once at session start. The trade-off is that you need to start a new chat after installation, which takes about two seconds.
Step 4: Update Your OpenClaw Skills Regularly
Skills are actively maintained by their creators. New versions bring bug fixes, improved instructions, and expanded capabilities. Keeping your skills updated ensures your agent always has the best version of each workflow.
How to Update
Option A: Ask your agent
“Update all my installed skills.”
Your agent will check each installed skill against the ClaWHub registry and update any that have newer versions available.
Option B: Update a specific skill
“Update the healthcheck skill to the latest version.”
Option C: Via SSH
clawhub update –all
Or update a specific skill:
clawhub update healthcheck
When to Update
- After a major OpenClaw release, when skill creators often add support for new platform features.
- Monthly (a good cadence for most users).
- When something isn’t working, since an update might fix the issue.
An outdated skill might reference deprecated commands, miss new agent capabilities, or contain instructions that no longer match the current OpenClaw architecture. Regular updates, even just once a month, keep everything running smoothly.
Recommended Skills for xCloud OpenClaw Users
The ClaWHub marketplace has a growing catalog of skills. Here are the ones that xCloud users find most valuable in 2026, based on install counts and community feedback.
Master Skills Reference Table
| Skill name | Category | What it does | Difficulty | Install command | Best for |
| 🥇 ClaWHub | Core | Browse and install skills from the marketplace | Easy | clawhub install clawhub | Everyone — install this first |
| 🥈 healthcheck | Security | Audits your Linux VPS for vulnerabilities, open ports, outdated packages | Easy | clawhub install healthcheck | Server admins, security-conscious users |
| 🥉 Last 30 Days | Research | Searches Reddit, X/Twitter, and YouTube for recent content on any topic | Easy | clawhub install last-30-days | Content creators, marketers, researchers |
| 4️⃣ Self-Improving Agent | Intelligence | Adds hot memory, correction logs, and task pattern recognition | Medium | clawhub install self-improving-agent | Power users who want their agent to learn |
| 5️⃣ tmux | DevOps | Controls terminal multiplexer sessions on your VPS | Medium | clawhub install tmux | Developers, sysadmins running background tasks |
| 6️⃣ weather | Utility | Provides current conditions and multi-day forecasts | Easy | clawhub install weather | Daily briefing setups, travel planning |
| 7️⃣ ai-authority-content | Content | Creates AI-optimized, authority-driven articles and guides | Medium | clawhub install ai-authority-content | Bloggers, content marketers, SEO professionals |
| 8️⃣ agent-browser | Automation | Headless browser control for web research and screenshots | Medium | clawhub install agent-browser | Research-heavy workflows, competitive analysis |
Recommended Starter Stack
If you’re new to OpenClaw on xCloud, here’s a proven starting combination:
- ClaWHub: lets your agent install everything else
- healthcheck: run a security audit on day one
- Last 30 Days: immediate research capability
- Self-Improving Agent: your agent gets smarter over time (see the memory and self-improvement guide)
This four-skill stack takes about five minutes to install and gives you a capable, security-aware, self-improving research agent.
Where Skills Live on Your xCloud Server
Understanding the file structure helps when you need to troubleshoot, back up, or manually manage skills.
/root/.openclaw/workspace/skills/ ← Skills for the current agent only
/root/.openclaw/skills/ ← Shared skills available to ALL agents on the server
Agent-Specific vs. Shared Skills
| Location | Scope | When to use |
| /root/.openclaw/workspace/skills/ | Current agent only | Default — most skills go here |
| /root/.openclaw/skills/ | All agents on the server | When running multiple agents via sub-agents and you want them all to share a skill |
What’s Inside a Skill Directory
Each skill typically contains:
/root/.openclaw/workspace/skills/healthcheck/
├── SKILL.md ← The main instruction file (required)
├── references/ ← Supporting reference files (optional)
│ ├── checklist.md
│ └── templates.md
└── README.md ← Human-readable description (optional)
The only required file is SKILL.md. Everything else is supplementary.
Knowing where skills live means you can back them up, copy them between servers, inspect them before installation, or even write your own. It’s just files on a Linux filesystem: no magic, no proprietary format.
CLI Commands Quick Reference for ClaWHub
Here’s every ClaWHub command you’ll need, with descriptions and examples.
| Command | Description | Example |
| clawhub login | Authenticate with your ClaWHub account | clawhub login |
| clawhub search “query” | Search the marketplace for skills matching a keyword | clawhub search “security audit” |
| clawhub install <slug> | Install a skill by its slug identifier | clawhub install healthcheck |
| clawhub list | List all currently installed skills | clawhub list |
| clawhub update –all | Update all installed skills to their latest versions | clawhub update –all |
| clawhub update <slug> | Update a specific skill | clawhub update healthcheck |
| clawhub info <slug> | View details about a specific skill | clawhub info last-30-days |
| clawhub uninstall <slug> | Remove an installed skill | clawhub uninstall weather |
Using ClaWHub Commands via Your Agent
You don’t need to SSH into your server to use these commands. Your agent can run them for you. Just ask in natural language:
“Search ClaWHub for skills related to email automation.”
“What skills do I have installed right now?”
“Uninstall the weather skill — I don’t need it anymore.”
Your agent translates your request into the appropriate clawhub CLI command and executes it on your VPS.
Security: What to Check Before Installing Any Skill
Skills have access to your agent’s full capabilities, which on a VPS means file access, shell commands, and network requests. This is powerful, but it also means you should treat skill installation with the same care you’d give to installing software on any server.
Security Checklist
| Check | What to look for | Risk level if skipped |
| Read the SKILL.md | Open and read the full contents before installing. Look for unexpected commands, external URLs, or data exfiltration patterns. | 🔴 High |
| Check download count & stars | Skills with high download counts and positive ratings have been vetted by the community. | 🟡 Medium |
| Review the publisher’s profile | Check the publisher’s GitHub profile and other published skills. Established publishers are more trustworthy. | 🟡 Medium |
| Look for external API calls | Does the skill send data to third-party servers? Make sure you’re comfortable with where your data goes. | 🔴 High |
| Verify the skill scope | A “weather” skill shouldn’t need to read your SSH keys. Check that the skill’s instructions match its stated purpose. | 🔴 High |
| Check for hardcoded credentials | No legitimate skill should contain API keys or passwords. If you see them, don’t install it. | 🔴 High |
| Test in a sub-agent first | For unfamiliar publishers, install in a sub-agent with limited permissions before adding to your main agent. | 🟢 Low |
When in Doubt: Build Your Own
If you can’t find a trustworthy skill for a specific task, ask your agent to build one:
“Create a custom skill that checks my server’s disk usage and sends me a summary every morning.”
Your agent can write a SKILL.md file tailored to your exact needs, without touching the marketplace. For a deeper look at custom skill creation, see the complete ClaWHub skills guide.
Troubleshooting Common OpenClaw Skill Issues
Common Issues and Solutions
| Problem | Likely cause | Solution |
| Skill installed but agent doesn’t use it | You’re still in the same chat session | Start a new conversation — skills load at session start |
| “Skill not found” error during install | Typo in the slug, or the skill was removed from ClaWHub | Run clawhub search “keyword” to find the correct slug |
| Agent says it doesn’t have a skill you installed | Skill might be in the wrong directory | Check that the SKILL.md exists in /root/.openclaw/workspace/skills/<skill-name>/ |
| Skill works sometimes but not consistently | Context window overflow — too many skills loaded | Reduce your installed skills to the most essential ones (8–10 is a good target) |
| clawhub command not found | ClaWHub CLI isn’t installed or not in PATH | Run which clawhub to check. Reinstall if needed via the OpenClaw setup script |
| Skill update fails | Network issue or ClaWHub server temporarily unavailable | Wait a few minutes and try again. Check your VPS internet connectivity with ping clawhub.com |
| Agent behaves strangely after installing a skill | Conflicting instructions between skills | Remove the most recently installed skill and test again. Two skills might give contradictory instructions |
| Permission denied when installing | Running as a non-root user without proper access | Ensure you’re running as root or the user that owns the OpenClaw installation |
Advanced Troubleshooting
Check the OpenClaw service status:
systemctl status openclaw
View recent agent logs:
journalctl -u openclaw –since “1 hour ago”
Verify skill file integrity:
ls -la /root/.openclaw/workspace/skills/
cat /root/.openclaw/workspace/skills/<skill-name>/SKILL.md | head -20
Restart the OpenClaw service (if something is really stuck):
systemctl restart openclaw
Video Resources & Tutorials for OpenClaw Skills
Video walkthroughs are one of the fastest ways to get comfortable with OpenClaw skills and ClaWHub. Here are community-recommended tutorials available on YouTube as of March 2026.
| Topic | What you’ll learn | Where to find it |
| OpenClaw initial setup on a VPS | Full walkthrough from VPS provisioning to first agent conversation | Search YouTube: “OpenClaw VPS setup tutorial 2026” |
| Installing skills via ClaWHub | Screen recording of the ClaWHub install workflow | Search YouTube: “ClaWHub skills install OpenClaw” |
| Building custom skills | How to write your own SKILL.md from scratch | Search YouTube: “OpenClaw custom skill tutorial” |
| Multi-agent setup | Running multiple specialized agents on one xCloud server | Search YouTube: “OpenClaw sub-agents setup” |
| Security hardening | Securing your OpenClaw VPS against common threats | Search YouTube: “OpenClaw VPS security hardening” |
According to Ahrefs 2026 research analyzing over 1 billion data points, YouTube mentions are the strongest predictor of AI visibility with a 0.737 correlation, higher than domain rating, backlinks, and all traditional SEO factors combined. Content referenced in YouTube videos is significantly more likely to appear in AI search results.
Frequently Asked Questions
Which skill should I install first?
ClaWHub, always. It’s the gateway to the entire skill ecosystem. Without it, your agent can’t browse or install other skills from the marketplace. After ClaWHub, the healthcheck skill is a smart second choice because it immediately audits your xCloud server’s security.
Do I need to pay for skills?
Most skills on ClaWHub are free and open source. Some premium skills from commercial publishers may require a subscription or one-time purchase, but the core skills recommended in this guide (ClaWHub, healthcheck, Last 30 Days, tmux, weather) are all free.
How long does it take for a skill to start working?
Installation takes seconds. Activation requires starting a new chat session, which takes about two seconds. So from “I want this skill” to “my agent can use it” is typically under a minute.
Can I use skills on any VPS, or only xCloud?
Skills work on any VPS running OpenClaw: xCloud, Hetzner, DigitalOcean, AWS, or self-hosted. The difference with xCloud OpenClaw Hosting is that OpenClaw comes pre-installed and optimized, so you skip the manual server setup entirely.
How many skills can I install at once?
There’s no hard cap, but practical limits exist. Each skill adds to your agent’s context window (the amount of information it processes at the start of each conversation). Most users find that 8–12 well-chosen skills provide a good balance of capability and performance. If you notice your agent becoming slower or less focused, consider removing skills you rarely use.
Can I write my own skills?
Absolutely. A skill is just a SKILL.md file, a Markdown document with structured instructions. Ask your agent to create one, or write it yourself. If your custom skill works well, you can even publish it to ClaWHub for others to use.
What happens if two skills conflict with each other?
Rare, but possible. If two skills give contradictory instructions for the same type of task, your agent may behave inconsistently. The fix is simple: uninstall one of the conflicting skills. You can identify the conflict by removing skills one at a time and testing.
Is it safe to install skills from unknown publishers?
Exercise caution. Always read the SKILL.md file before installing. Check the publisher’s GitHub profile and the skill’s download count. If you’re unsure, ask your agent to review the skill’s contents before installation. For high-security environments, test new skills in a sub-agent first.
How do I uninstall a skill I no longer need?
Either tell your agent “Uninstall the [skill-name] skill” or run clawhub uninstall <slug> via SSH. Start a new chat session afterward to clear it from your agent’s active context.
Where can I get help if something goes wrong?
Three resources: the xCloud support portal for hosting-specific issues, the OpenClaw community Discord for general questions and community help, and the ClaWHub marketplace for skill-specific documentation and publisher contact info.
Glossary
| Term | Definition |
| Skills | Structured add-on capabilities for your OpenClaw agent. Each skill is a SKILL.md file containing instructions that teach your agent a new workflow or behavior. |
| ClaWHub | The community marketplace at clawhub.com where skill creators publish and share OpenClaw skills. Think of it as an app store for AI agent capabilities. |
| SKILL.md | The Markdown file that defines a skill. Contains structured instructions, metadata (name, version, description), and behavioral rules that your agent follows. |
| Gateway | The OpenClaw Gateway is the core service that manages your agent’s connections to messaging channels (Telegram, WhatsApp, Discord), tool access, and session management. It runs as a systemd service on your VPS. |
| Sub-agents | Secondary agents that your main agent can spawn to handle specific tasks in parallel. Each sub-agent can have its own skill set. Learn more in the multi-agent setup guide. |
| Context window | The total amount of information your agent can hold in working memory during a single conversation. Skills, conversation history, and tool outputs all consume context window space. |
| Slug | A short, URL-friendly identifier for a skill on ClaWHub (e.g., healthcheck, last-30-days). Used in CLI commands like clawhub install <slug>. |
| systemd | The Linux service manager that keeps your OpenClaw agent running continuously on your xCloud VPS. It handles automatic restarts, logging, and service lifecycle management. |
Next Steps
You now have everything you need to install, manage, and secure skills on your xCloud OpenClaw agent. Here’s your recommended path forward:
| Your goal | Next action | Resource |
| Get your agent fully skilled up | Install the starter stack (ClaWHub + healthcheck + Last 30 Days + Self-Improving Agent) | This guide, Steps 1–2 |
| Run multiple specialized agents | Set up sub-agents with different skill profiles | Multi-agent setup guide |
| Make your agent smarter over time | Install the Self-Improving Agent skill and configure memory | Memory and self-improvement guide |
| Explore the full skill catalog | Browse ClaWHub for skills that match your workflow | ClaWHub marketplace |
| Build a custom skill | Ask your agent or write your own SKILL.md | Complete ClaWHub skills guide |
Start with one skill. Install it, start a new chat, and try it out. Once you see how it works, you’ll want to install five more. That’s exactly how it’s supposed to feel.
Published by xCloud.host – OpenClaw hosting made simple. For support, visit xCloud supportor join the OpenClaw community on Discord.
