Skip to main content

The ClawHub Malware Incident

In early 2026, OpenClaw faced its first major security crisis — and it's a cautionary tale every user should know.

🚨

Real incident: Malicious skills were published to ClawHub, OpenClaw's official skill marketplace. Users who installed them unknowingly gave attackers full access to their machines — including file systems, shell execution, and messaging apps.

🕸️

How it happened

Attackers uploaded skills that appeared legitimate — with good descriptions and fake reviews. Once installed, they used OpenClaw's own shell execution capabilities to run malicious code.

💥

Why it was dangerous

OpenClaw runs with local machine access by design. A malicious skill has the same privileges as you — it can read files, send emails, and execute commands without asking.

🛠️

What was fixed

The OpenClaw team added skill signature verification and permission sandboxing. But user vigilance remains the most important line of defence.

Common Threats & Their Risk Level

Understanding what can go wrong is the first step to staying safe.

Threat Risk Description
Malicious Skills High Third-party skills with hidden malware that exploit OpenClaw's local access to compromise your machine.
API Key Theft High Your AI model API key stored in the .env file can be stolen and used to rack up huge bills on your account.
Message Injection Medium Malicious content in emails or web pages tricks OpenClaw into executing unintended commands.
Overprivileged Skills Medium Skills requesting more permissions than they need — a red flag that can lead to unintended data access.
Exposed Messaging Tokens Medium WhatsApp, Telegram, or Slack tokens stored insecurely can give attackers access to your messaging accounts.
Unencrypted Config Files Low Storing your .env file in a shared or synced folder (e.g. iCloud, Dropbox) can expose your credentials.

Step-by-Step Security Setup

Follow these steps when setting up or reviewing your OpenClaw installation.

1

Run OpenClaw in a Virtual Machine

The single most effective protection. Install OpenClaw inside a VM (like VirtualBox or UTM on Mac). Even if a malicious skill runs, it's isolated from your real files and system.

2

Only Install Skills You Trust

Treat skills like apps from an unknown developer. Before installing any skill from ClawHub, check: Who published it? How many verified installs does it have? What permissions does it request? If anything looks off, skip it.

3

Protect Your .env File

Your API keys and tokens live here. Make sure it's never committed to Git, never stored in a synced cloud folder, and has restricted file permissions.

chmod 600 .env
4

Use Scoped API Keys

When generating API keys for OpenClaw, give them the minimum permissions needed. If possible, set usage limits and expiry dates so a stolen key can't do much damage.

5

Review Skill Permissions Before Installing

Every OpenClaw skill declares what it can access. A recipe skill asking for shell access is a major red flag. Only approve permissions that make sense for what the skill does.

6

Keep OpenClaw Updated

Security patches are released regularly. Stay on the latest version to benefit from signature verification, sandboxing improvements, and other fixes.

git pull origin main

Security Checklist

Run through this list before and after setting up OpenClaw.

  • Running in a VM or container OpenClaw is isolated from your main system files and OS.
  • .env file has restricted permissions Only your user account can read the file (chmod 600 .env).
  • .env is in .gitignore Your secrets will never be accidentally committed to a public repo.
  • Only trusted skills installed Each skill has been reviewed for permissions and source credibility.
  • API keys are scoped and have expiry dates No master keys — each key only does what it needs to.
  • OpenClaw is on the latest version All known security patches have been applied.
  • Messaging tokens stored securely WhatsApp, Telegram, Slack tokens are not shared or stored in plain text outside .env.
🟢

You're in good shape if you can check every box above. Security isn't a one-time task — revisit this list whenever you add new skills or update your setup.