ClawKit Logo
ClawKitReliability Toolkit
Back to Registry
Official Verified

Mcp Security Audit

Skill by aptratcn

skill-install — Terminal

Install via CLI (Recommended)

clawhub install openclaw/skills/skills/aptratcn/mcp-security-audit
Or

name: skill-mcp-security-audit description: Security audit for MCP (Model Context Protocol) servers. Detect data exfiltration risks, command injection, permission escalation, and supply chain vulnerabilities before adding MCP servers to your agent. Trigger on: 'audit MCP', 'MCP security', 'check MCP server', 'scan MCP'. metadata: openclaw: requires: {}

MCP Security Audit šŸ”’

Don't blindly trust MCP servers. Audit them first.

The Problem

MCP (Model Context Protocol) servers give AI agents powerful capabilities - file access, API calls, code execution. But they can also:

  • Exfiltrate data to external servers
  • Execute arbitrary commands on your machine
  • Access files beyond intended scope
  • Chain vulnerabilities for privilege escalation

Real incident: CVE-2026-23744 exposed MCP injection vulnerabilities. Supply chain attacks via compromised MCP packages are a growing threat.

Quick Audit Checklist

1. Source Verification āœ…

ā–” Is this an official/verified package?
ā–” Check npm/PyPI download counts and maintainer history
ā–” Review recent commits for suspicious changes
ā–” Verify package signature if available

2. Network Audit 🌐

ā–” List all external URLs/domains the MCP connects to
ā–” Check for hardcoded API endpoints
ā–” Verify TLS certificate validation is enabled
ā–” Flag any data sent to unknown domains

3. File Access Audit šŸ“

ā–” What directories can the MCP read/write?
ā–” Is access scoped to project directory only?
ā–” Check for path traversal vulnerabilities
ā–” Flag any access to ~/.ssh, ~/.config, env files

4. Command Execution Audit ⚔

ā–” Does the MCP execute shell commands?
ā–” Are commands user-controlled or hardcoded?
ā–” Check for command injection vectors
ā–” Verify sandboxing/isolation if present

5. Permission Scope Audit šŸ”‘

ā–” What permissions does the MCP request?
ā–” Are permissions minimal (principle of least privilege)?
ā–” Check for excessive scope requests
ā–” Verify user consent for sensitive operations

6. Dependency Audit šŸ“¦

ā–” Run npm audit / pip-audit / cargo audit
ā–” Check for known CVEs in dependencies
ā–” Flag outdated packages with security fixes
ā–” Review transitive dependencies

Audit Commands

For npm-based MCP servers:

# Check package.json for suspicious scripts
cat package.json | jq '.scripts'

# Audit dependencies
npm audit

# Check for post-install scripts
cat package.json | jq '.scripts.postinstall, .scripts.preinstall'

# List network calls (requires grep)
grep -r "fetch\|axios\|http\|https\|ws://" src/ --include="*.js" --include="*.ts"

For Python MCP servers:

# Check requirements.txt for suspicious packages
cat requirements.txt

# Audit dependencies
pip-audit

# Check for network calls
grep -r "requests\|urllib\|httpx\|aiohttp" src/ --include="*.py"

# Check for subprocess calls
grep -r "subprocess\|os.system\|exec\|eval" src/ --include="*.py"

Risk Scoring

Metadata

Author@aptratcn
Stars4473
Views0
Updated2026-05-01
View Author Profile
AI Skill Finder

Not sure this is the right skill?

Describe what you want to build — we'll match you to the best skill from 16,000+ options.

Find the right skill
Add to Configuration

Paste this into your clawhub.json to enable this plugin.

{
  "plugins": {
    "official-aptratcn-mcp-security-audit": {
      "enabled": true,
      "auto_update": true
    }
  }
}
Safety NoteClawKit audits metadata but not runtime behavior. Use with caution.