🛡️ ClawShield

Security scanner for OpenClaw skills — detect malicious patterns, data exfiltration, and supply chain risks before they hit your agent.

Why ClawShield?

OpenClaw skills can execute code, access files, make network requests, and read secrets. A malicious skill can:

• 🔑 Steal API keys and credentials
• 📤 Exfiltrate sensitive data
• 💰 Hijack crypto wallets
• 🚪 Install backdoors and persistence mechanisms
• ⛏️ Run cryptominers

ClawShield catches these threats before you install.

Quick Start

# Install globally
npm install -g clawshield

# Scan a skill
clawshield scan ./my-skill

# Scan before installing
clawshield install @someone/cool-skill

What It Detects

🔴 Critical Threats

• Credential theft patterns (env vars, config files)
• Data exfiltration (network requests with sensitive data)
• Reverse shells and backdoors
• Cryptominer signatures
• Malicious obfuscation

🟠 High Risk

• Shell command injection
• Arbitrary code execution (eval, Function constructor)
• File system abuse (reading sensitive paths)
• Network abuse (suspicious endpoints)

🟡 Medium Risk

• Dependency confusion attacks
• Manifest permission mismatches
• Insecure configurations

CLI Commands

scan - Analyze a skill

clawshield scan <path> [options]

Options:
  -f, --format <type>   Output format: text, json, sarif (default: text)
  -o, --output <file>   Write report to file
  --threshold <level>   Fail if risk >= threshold (low|medium|high|critical)

install - Secure install wrapper

clawshield install <skill> [options]

# Scans the skill first, blocks if threats detected

rules - List detection rules

clawshield rules

# Shows all 30+ detection rules with descriptions

Example Output

🛡️  ClawShield Scan Report
═══════════════════════════════════════════════════

📁 Skill: ./suspicious-skill
📊 Risk Score: 85/100
🚨 Verdict: CRITICAL — DO NOT INSTALL

┌─────────────────────────────────────────────────┐
│ 🔴 CRITICAL: Credential Theft Detected          │
├─────────────────────────────────────────────────┤
│ File: src/index.ts:42                           │
│ Pattern: process.env.OPENAI_API_KEY             │
│ → Skill reads API keys and sends to external    │
│   endpoint: https://evil.com/collect            │
└─────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────┐
│ 🟠 HIGH: Shell Command Injection                │
├─────────────────────────────────────────────────┤
│ File: src/utils.ts:18                           │
│ Pattern: exec(`curl ${userInput}`)              │
│ → User input passed directly to shell command   │
└─────────────────────────────────────────────────┘

Recommendation: DO NOT INSTALL this skill.
Report to: [email protected]

CI/CD Integration

GitHub Actions

# .github/workflows/clawshield.yml
name: Skill Security Scan

on: [push, pull_request]

jobs:
  security:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
      - run: npm install -g clawshield
      - run: clawshield scan . --threshold high

Pricing

| Feature | Free | Pro ($19/mo) | Team ($49/seat/mo) | |---------|------|--------------|---------------------| | Basic scan (30 rules) | ✅ | ✅ | ✅ | | Scans per day | 10 | Unlimited | Unlimited | | CI integration | ❌ | ✅ | ✅ | | API access | ❌ | ✅ (1k/mo) | ✅ (Unlimited) | | Custom rules | ❌ | ❌ | ✅ | | Team dashboard | ❌ | ❌ | ✅ | | Slack/Discord alerts | ❌ | ❌ | ✅ |

Get Pro →

Detection Rules

ClawShield includes 30+ detection rules across categories:

• CRED: Credential access patterns
• EXFIL: Data exfiltration
• EXEC: Code execution risks
• NET: Suspicious network activity
• FS: File system abuse
• PERSIST: Persistence mechanisms
• CRYPTO: Cryptominer detection
• SHELL: Reverse shell patterns
• MANIFEST: Permission mismatches

Run clawshield rules to see the full list.

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Adding Detection Rules

// src/rules/index.ts
export const rules: Rule[] = [
  {
    id: 'CUSTOM001',
    name: 'My Custom Rule',
    severity: 'high',
    pattern: /suspicious\.pattern/,
    description: 'Detects suspicious patterns',
  },
  // ...
];

Security

Found a vulnerability in ClawShield itself? Email [email protected].

License

MIT © ClawShield Team

Built for the OpenClaw community 🦞