If OpenClaw is the brain of your AI assistant, then Skills are its hands, its eyes, and its voice. Without skills, your agent is a brilliant thinker trapped in a box—it can reason, but it cannot do. Install the right skills, and suddenly it can browse the web, manage your calendar, control your smart home, write and send emails, and even deploy code to production servers.
But with over 5,700 community-built skills now listed on the official ClawHub marketplace, the ecosystem has grown far beyond what any single person can evaluate. How do you know which skills are essential? Which ones are safe? And which ones might be quietly exfiltrating your data?
This guide answers all of those questions. We will explain exactly what skills are, walk you through the most important official skills you should know about, point you to the official resources, and—critically—give you a security playbook for vetting every skill before it touches your machine.
What Exactly Is a Skill?
At its core, an OpenClaw Skill is a self-contained module that grants your agent a new capability. Think of it like an app on your phone. Your phone's operating system provides the foundation (screen, networking, storage), but it's the apps that let you do useful things like order food, check the weather, or send a message.
Similarly, OpenClaw's core engine provides the AI reasoning, memory management, and messaging infrastructure. Skills plug into this engine to give the agent tools it can use to interact with the outside world.
Every skill follows a standard structure:
my-skill/
manifest.yaml # Declares permissions, triggers, and metadata
actions/
main.js # The code that actually runs
config.yaml # User-configurable settings
README.md # Documentation
The manifest.yaml is the most important file. It tells the AI what the skill can do, what permissions it needs, and when it should be activated. The AI reads this manifest and decides—on its own—when to invoke the skill based on your requests.
ClawHub: The Official Skills Marketplace
ClawHub (clawhub.ai) is the official registry and marketplace for OpenClaw skills. It functions much like npm for Node.js developers or the VS Code extension marketplace—a centralized place to discover, install, rate, and share skills.
As of February 2026, ClawHub hosts over 5,700 community-built skills. A curated subset of approximately 3,002 vetted skills is maintained in the official repository at github.com/openclaw/skills. This curated list filters out spam, cryptocurrency scams, malicious entries, non-English descriptions, and duplicates.
Installing Skills via the CLI
You don't need to visit the website to install skills. OpenClaw's built-in CLI makes it effortless:
# Search for a skill by keyword
openclaw skills search "email automation"
# View detailed information about a specific skill
openclaw skills info agentmail
# Install a skill
openclaw skills install agentmail
# List all installed skills
openclaw skills list
# Uninstall a skill you no longer need
openclaw skills uninstall agentmail
Key ClawHub Features
- Verified Publishers: Skills from verified publishers carry a blue checkmark, indicating they have undergone a basic security review by the OpenClaw team.
- VirusTotal Integration: Every skill uploaded to ClawHub is automatically scanned by VirusTotal for malware and suspicious code patterns. You can view the scan report directly on the skill's page.
- User Reviews & Ratings: Community feedback helps surface the best skills and flag problematic ones.
- Semantic Versioning: Skills follow semver. OpenClaw can auto-update patch versions while requiring your approval for major version changes.
Review of the Major Skills
Not all skills are created equal. Some are essential building blocks that nearly every user should have. Here is our review of the most impactful official and community skills available on ClawHub today.
1. Browser Control
Category: Core | Publisher: OpenClaw (Official)
This is arguably the most transformative skill in the entire ecosystem. Browser Control gives your agent the ability to operate a real Chromium-based web browser. It can navigate to URLs, click buttons, fill out forms, extract text, take screenshots, and even interact with JavaScript-heavy single-page applications.
Use Cases: Web research, price comparison, form filling, automated testing, data extraction from sites without APIs.
Why It Matters: Without this skill, your agent is blind to the web. With it, any website becomes a potential data source or interaction point.
2. Shell Execute
Category: Core | Publisher: OpenClaw (Official)
Shell Execute allows the agent to run terminal commands on your host machine (or inside a Docker sandbox). This is the skill that enables OpenClaw to install software, run scripts, manage files, compile code, and interact with virtually any command-line tool on your system.
Use Cases: Running build scripts, managing Git repositories, processing files with ffmpeg or imagemagick, system administration tasks.
⚠️ Security Note: This is the single most powerful—and most dangerous—skill. Always run it inside a Docker sandbox and enable Human-in-the-Loop (HITL) confirmation. Never give this skill unrestricted access on your host OS.
3. AgentMail
Category: Communication | Publisher: AgentMail (Verified)
AgentMail provides your agent with its own dedicated email address. Unlike a Gmail integration (where the agent uses your inbox), AgentMail creates a sandboxed email identity specifically for the agent. It can send and receive emails, parse attachments, and trigger workflows based on incoming messages.
Use Cases: Automated outreach, monitoring for specific emails (e.g., shipping notifications), receiving data feeds, and agent-to-agent communication.
Why It Matters: Giving the agent its own email address prevents it from accidentally sending messages as you and provides a clean separation of concerns.
4. GitHub Integration
Category: Developer Tools | Publisher: OpenClaw (Official)
This skill connects your agent to GitHub's API, enabling it to create and manage issues, review pull requests, browse repository contents, monitor CI/CD pipeline status, and even commit code changes.
Use Cases: Automated code review triage, issue management, release monitoring, and development workflow automation.
5. Playwright Scraper
Category: Data & Research | Publisher: Community (Verified)
While Browser Control is great for interactive browsing, Playwright Scraper is optimized for structured data extraction. It uses Microsoft's Playwright library to render JavaScript-heavy pages and extract data into clean JSON or CSV formats.
Use Cases: Monitoring competitor pricing, scraping job listings, extracting product reviews, building datasets for analysis.
6. Obsidian Integration
Category: Knowledge Management | Publisher: Community (Verified)
For users of the Obsidian note-taking app, this skill is a game-changer. It allows OpenClaw to read, create, search, and link notes within your Obsidian vault. The agent can query your personal knowledge base to find relevant information, create new notes from research, and even maintain a "daily briefing" document.
Use Cases: Personal knowledge management, research synthesis, automated note-taking from meetings or articles.
7. Linear Integration
Category: Project Management | Publisher: Linear (Verified)
Linear is a popular project management tool for engineering teams. This skill allows your agent to create issues, update statuses, assign tasks, and query project boards—all through natural language commands.
Use Cases: "Create a bug report for the login crash," "What's blocking the v2.1 release?" "Assign the API refactor to Sarah."
8. Home Assistant
Category: Smart Home | Publisher: Community (Verified)
This skill bridges OpenClaw to your Home Assistant installation, giving your agent control over every smart device in your home. Lights, thermostats, locks, cameras, and media players all become controllable through natural language.
Use Cases: "Turn off all the lights downstairs," "Set the thermostat to 22°C," "Lock the front door and arm the alarm."
9. Calendar & Scheduling
Category: Productivity | Publisher: OpenClaw (Official)
Connects to Google Calendar or Outlook to read events, create appointments, send reminders, and suggest optimal meeting times based on your availability patterns.
Use Cases: "What's on my schedule tomorrow?" "Book a 30-minute meeting with the design team on Thursday afternoon." "Remind me about the dentist appointment an hour before."
10. Spotify Controller
Category: Entertainment | Publisher: Community (Verified)
Control your Spotify playback, search for music, create playlists, and set up automation routines (like playing a focus playlist when you start a work session).
Use Cases: "Play some lo-fi beats," "Add this song to my Road Trip playlist," "What was the last song I listened to?"
11. RSS Reader
Category: Data & Research | Publisher: OpenClaw (Official)
RSS Reader lets your agent subscribe to and fetch content from any RSS or Atom feed. It can pull the latest articles from news sites, blogs, podcasts, and release notes—then parse them into clean, structured data that other skills and pipelines can act on. It supports filtering by keyword, date range, and max items per feed.
Use Cases: Monitoring industry news, tracking competitor blog posts, aggregating release notes from open-source projects, and feeding articles into automated summarisation pipelines.
Why It Matters: RSS is the backbone of content monitoring. This skill turns your agent into a tireless reader that never misses an update across hundreds of sources.
12. Scheduler
Category: Core | Publisher: OpenClaw (Official)
Scheduler is the clock that makes your agent proactive instead of reactive. It allows you to define cron-based schedules, one-off delayed tasks, and recurring intervals that trigger actions automatically. Under the hood, it integrates with OpenClaw's pipeline orchestration engine, making it the go-to skill for any workflow that needs to run on a timer.
Use Cases: Daily news briefings at 6 AM, hourly price checks on competitor websites, weekly report generation, and reminder notifications at specific times.
Why It Matters: Without Scheduler, your agent only acts when you ask. With it, entire workflows run autonomously on your timetable—turning OpenClaw from an assistant into an employee.
Security: The Skills Vetting Playbook
The rapid growth of ClawHub has been overwhelmingly positive, but it has also attracted bad actors. Security researchers have identified hundreds of malicious skills that were designed to steal passwords, exfiltrate session data, or execute unauthorized commands on users' machines.
Here is your playbook for staying safe.
Rule 1: Only Install Verified Skills
Look for the blue checkmark on ClawHub. Verified publishers have been reviewed by the OpenClaw team. While this is not a guarantee of safety, it significantly reduces risk.
Rule 2: Read the Manifest Before Installing
Before you run openclaw skills install, always check what permissions the skill requests:
openclaw skills info suspicious-skill
Ask yourself: Does this skill need these permissions? A weather skill should not need file_write or shell_execute. A calendar skill should not need browser_control. If the permissions don't match the stated purpose, walk away.
Rule 3: Check the VirusTotal Report
Every skill on ClawHub now has an associated VirusTotal scan. Click the "Security" tab on any skill's page to view the report. If any engine flags the skill, investigate further before installing.
Rule 4: Review the Source Code
For critical skills (anything with shell_execute or network_access permissions), take five minutes to read the actual source code. Look for:
- Obfuscated code (base64-encoded strings,
eval()calls). - Unexpected network requests (data being sent to unknown servers).
- File access patterns (reading
~/.ssh,~/.aws, or browser cookie stores).
Rule 5: Pin Your Versions
Never use latest when installing skills. Pin to a specific, audited version:
openclaw skills install agentmail@2.1.3
This prevents a compromised update from silently replacing a trusted skill with a malicious version.
Rule 6: Use the Curated List
When in doubt, stick to the curated awesome list at github.com/openclaw/skills. These 3,002 skills have been filtered for quality and safety by the community maintainers.
Rule 7: Sandbox Everything
Enable Docker sandboxing for all skills, especially those with shell_execute or network_access permissions. This ensures that even if a skill is compromised, the blast radius is contained to a disposable container.
The Golden Rule
The OpenClaw skills ecosystem is one of the most exciting developments in the agentic AI space. It transforms a single AI assistant into an infinitely extensible platform that can adapt to virtually any workflow. But with great power comes great responsibility.
Install thoughtfully. Vet aggressively. Sandbox everything.
Your agent is only as trustworthy as the skills you give it. Choose wisely, and you will have the most capable digital assistant on the planet. Choose carelessly, and you might just hand the keys to your digital life to a stranger.
Happy skilling! 🐾
