---

name: flaw0 description: Security and vulnerability scanner for OpenClaw code, plugins, skills, and Node.js dependencies. Powered by OpenClaw AI models. version: 1.0.0 author: Tom homepage: https://github.com/yourusername/flaw0 license: MIT metadata: openclaw: emoji: "🔍" category: "security" tags:

• security
• vulnerability-scanner
• code-analysis
• dependency-checker
• openclaw

---

flaw0 - Zero Flaws Security Scanner

Security and vulnerability scanner for OpenClaw ecosystems. Analyzes source code, plugins, skills, and Node.js dependencies to detect potential security flaws.

Goal: Achieve flaw 0 (zero flaws detected) 🎯

Installation

Install this skill via ClawHub:

npx clawhub@latest install flaw0

Or install globally via npm:

npm install -g flaw0

When to Use This Skill

Use flaw0 to ensure your OpenClaw code and dependencies are secure:

Before Installing Skills

# Check a skill before installing
flaw0 scan ~/.openclaw/skills/new-skill

During Development

# Scan your code as you develop
flaw0 scan src/

# Check dependencies
flaw0 deps

Before Committing

# Full security audit
flaw0 audit

Auditing OpenClaw Installation

# Scan all OpenClaw components
flaw0 scan --target all

# Check specific components
flaw0 scan --target skills
flaw0 scan --target plugins
flaw0 scan --target core

Usage

Basic Commands

Scan Code

# Scan current directory
flaw0 scan

# Scan specific directory
flaw0 scan /path/to/code

# Use specific AI model
flaw0 scan --model claude-opus-4-6

Check Dependencies

# Quick dependency scan
flaw0 deps

# Deep scan (entire dependency tree)
flaw0 deps --deep

Full Security Audit

# Comprehensive scan (code + dependencies)
flaw0 audit

# Save report to file
flaw0 audit --output report.json

# JSON output for CI/CD
flaw0 audit --json

Scan OpenClaw Components

# Scan OpenClaw core
flaw0 scan --target core

# Scan all plugins
flaw0 scan --target plugins

# Scan all skills
flaw0 scan --target skills

# Scan everything
flaw0 scan --target all

What flaw0 Detects

Code Vulnerabilities (12+ Types)

1. Command Injection

   • exec() with unsanitized input
   • Shell command construction with user input

2. Code Injection

   • eval() usage
   • Function() constructor with strings

3. SQL Injection

   • String concatenation in SQL queries
   • Unparameterized queries

4. Cross-Site Scripting (XSS)

   • innerHTML assignments
   • dangerouslySetInnerHTML usage

5. Path Traversal

   • Unvalidated file path operations
   • readFile() with user input

6. Hardcoded Secrets

   • API keys in source code
   • Passwords and tokens
   • AWS credentials

7. Weak Cryptography

   • MD5 and SHA1 usage
   • Weak hashing algorithms

8. Insecure Randomness

   • Math.random() for security operations
   • Predictable token generation

9. Unsafe Deserialization

   • JSON.parse() without validation
   • Unvalidated input parsing

10. Missing Authentication

    • API endpoints without auth middleware
    • Unprotected routes

Dependency Issues

1. Known CVEs - Vulnerabilities from CVE database
2. Outdated Packages - Packages with security updates available
3. Malicious Packages - Known malware or suspicious packages
4. Duplicate Dependencies - Bloated dependency trees

Understanding Results

Flaw Score

Results are reported with a flaw score - lower is better:

• flaw 0 🎯 - Perfect! No issues detected
• flaw 1-3 🟡 - Minor issues
• flaw 4-10 🟠 - Needs attention
• flaw 10+ 🔴 - Critical issues

Score Calculation

Each issue is weighted by severity:

• Critical: 3 points
• High: 2 points
• Medium: 1 point
• Low: 0.5 points

Total flaw score = sum of all weighted issues (rounded)

Example Output

Clean Code (flaw 0)

🔍 flaw0 Security Scan Results
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📊 Result: flaw 0
✅ Status: SECURE

✓ No security issues detected!
✓ All checks passed

Great job! 🎉

Issues Found (flaw 12)

🔍 flaw0 Security Scan Results
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📊 Result: flaw 12
⚠️  Status: ISSUES FOUND

Code Flaws: 5
├─ 🔴 Critical: 2
├─ 🟠 High: 1
├─ 🟡 Medium: 2
└─ ⚪ Low: 0

Dependency Flaws: 7
├─ 🔴 Critical CVEs: 3
├─ 🟠 High CVEs: 2
├─ 🟡 Medium: 2
└─ ⚪ Low: 0

Detailed Report:
─────────────────────────────────

1. [CRITICAL] Command Injection
   Location: src/executor.js:78
   Code: `exec(\`ls ${userInput}\`)`
   Description: Unsanitized exec() call
   → Fix: Use execFile() or validate input
   🤖 AI Confidence: high
   💡 AI Suggestion: Replace exec() with execFile()
      and validate input against whitelist

2. [HIGH] Hardcoded API Key
   Location: config/api.js:5
   Code: `const API_KEY = "sk-1234..."`
   Description: API key exposed in source code
   → Fix: Use process.env.API_KEY

3. [CRITICAL] CVE-2024-12345 in lodash@4.17.19
   Package: lodash@4.17.19
   Description: Prototype pollution vulnerability
   → Fix: npm install lodash@4.17.21

...

AI-Powered Analysis

flaw0 uses OpenClaw's AI models for intelligent code review:

Available Models

claude-sonnet-4-5 (default)

• Balanced speed and accuracy
• Best for most use cases
• Good false positive reduction

flaw0 scan --model claude-sonnet-4-5

claude-opus-4-6

• Most thorough analysis
• Deepest context understanding
• Slower but most accurate

flaw0 scan --model claude-opus-4-6

claude-haiku-4-5

• Fastest scanning
• Good for quick checks
• Use in CI/CD for speed

flaw0 scan --model claude-haiku-4-5

AI Features

• Context-aware analysis - Understands code flow and context
• False positive reduction - Filters out non-issues
• Confidence scoring - Rates detection confidence
• Fix suggestions - Provides specific remediation steps

Configuration

Create Config File

flaw0 init

This creates .flaw0rc.json:

{
  "severity": {
    "failOn": "high",
    "ignore": ["low"]
  },
  "targets": {
    "code": true,
    "dependencies": true,
    "devDependencies": false
  },
  "exclude": [
    "node_modules/**",
    "test/**",
    "*.test.js"
  ],
  "model": "claude-sonnet-4-5",
  "maxFlawScore": 0
}

Configuration Options

• severity.failOn - Exit with error on this severity level or higher
• severity.ignore - Skip these severity levels
• targets - What to scan (code, dependencies)
• exclude - File patterns to ignore
• model - AI model to use
• maxFlawScore - Maximum acceptable flaw score

CI/CD Integration

GitHub Actions

name: Security Scan

on: [push, pull_request]

jobs:
  flaw0:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - uses: actions/setup-node@v3

      - name: Install flaw0
        run: npm install -g flaw0

      - name: Run security scan
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
        run: flaw0 audit

      - name: Check flaw score
        run: |
          SCORE=$(flaw0 audit --json | jq '.flawScore')
          if [ "$SCORE" -gt 0 ]; then
            echo "❌ Flaws detected: flaw $SCORE"
            exit 1
          fi
          echo "✅ No flaws: flaw 0"

Pre-commit Hook

#!/bin/bash
echo "🔍 Running flaw0 scan..."
flaw0 scan

if [ $? -ne 0 ]; then
  echo "❌ Flaws detected! Commit blocked."
  exit 1
fi

Examples

Scan Before Installing a Skill

# Download a skill to review
git clone https://github.com/user/some-skill.git /tmp/some-skill

# Scan it
flaw0 scan /tmp/some-skill

# If flaw 0, safe to install
# If flaw > 0, review issues first

Audit Your OpenClaw Skills

# Scan all installed skills
flaw0 scan --target skills

# Example output:
# ✓ clawdex - flaw 0
# ✓ database-helper - flaw 0
# ⚠ crypto-bot - flaw 3
# ✓ git-assistant - flaw 0
# Overall: flaw 3

Check Dependencies After Install

# After installing new packages
npm install some-package

# Check for vulnerabilities
flaw0 deps

Full Project Audit

# Comprehensive security check
flaw0 audit --output security-report.json

# Review the report
cat security-report.json | jq '.flawScore'

API Usage

Use flaw0 programmatically in your own tools:

const Flaw0 = require('flaw0');

const scanner = new Flaw0({
  target: './src',
  model: 'claude-sonnet-4-5'
});

// Run full scan
const results = await scanner.scan();

console.log(`Flaw Score: ${results.flawScore}`);

if (results.flawScore === 0) {
  console.log('✅ No flaws detected!');
} else {
  results.codeFlaws.forEach(flaw => {
    console.log(`[${flaw.severity}] ${flaw.name}`);
    console.log(`  Location: ${flaw.file}:${flaw.line}`);
    console.log(`  Fix: ${flaw.fix}`);
  });
}

How It Works

1. Pattern Matching - Fast regex-based detection of common vulnerabilities
2. AI Analysis - Claude AI reviews each issue in context
3. False Positive Filtering - AI identifies and removes non-issues
4. Dependency Checking - Integrates with npm audit and CVE databases
5. Scoring - Calculates weighted flaw score
6. Reporting - Generates detailed, actionable reports

Tips for Achieving flaw 0

1. Fix Critical issues first - Biggest security impact
2. Update dependencies - Resolve known CVEs quickly
3. Use parameterized queries - Prevent SQL injection
4. Validate all inputs - Stop injection attacks
5. Use environment variables - No hardcoded secrets
6. Apply security headers - Use helmet.js
7. Implement authentication - Protect all endpoints
8. Use strong crypto - SHA-256 or better
9. Sanitize output - Prevent XSS
10. Review AI suggestions - Learn from recommendations

Comparison with Other Tools

Feature	flaw0	npm audit	Snyk	ESLint Security
Dependency CVEs	✅	✅	✅	❌
AI Code Analysis	✅	❌	❌	❌
OpenClaw-specific	✅	❌	❌	❌
Context-aware	✅	❌	⚠️	⚠️
False positive reduction	✅	❌	⚠️	❌
Fix suggestions	✅	⚠️	✅	⚠️

Requirements

• Node.js: 14+
• API Key: Anthropic API key for AI analysis
• npm: For dependency checking

Setup API Key

export ANTHROPIC_API_KEY='your-api-key-here'

Get your API key from: https://console.anthropic.com/

Troubleshooting

"No API key found"

export ANTHROPIC_API_KEY='sk-...'
# Or add to ~/.bashrc or ~/.zshrc

"npm audit failed"

Ensure you have a valid package.json:

npm init -y
npm install

Rate Limit Exceeded

If you hit API rate limits:

1. Use haiku model: --model haiku
2. Scan smaller portions
3. Wait and retry

Support

• Documentation: See USAGE.md for detailed guide
• Examples: Check examples/ directory
• Issues: Report at GitHub repository
• Demo: Run ./demo.sh for interactive demo

About

flaw0 helps the OpenClaw community achieve secure, vulnerability-free code.

• Built with OpenClaw/Claude AI
• Uses industry-standard security patterns
• Continuously updated with new vulnerabilities
• Open source under MIT license

Contributing

Contributions welcome! Areas for contribution:

• New vulnerability patterns
• Additional AI models
• Python/Go support
• Web dashboard
• Custom rule engine

License

MIT License - see LICENSE file

---

Goal: flaw 0 for everyone! 🎯

Remember: Security is not a one-time check. Run flaw0 regularly to maintain flaw 0 status!