🛡️ AI Code Guard

Detect security vulnerabilities in AI-generated code before they reach production

AI coding assistants (GitHub Copilot, Claude, ChatGPT, Cursor) are revolutionizing development — but they can introduce security vulnerabilities that slip past code review. AI Code Guard scans your codebase for security issues commonly found in AI-generated code.

🎯 What It Detects

Category	Examples
Prompt Injection Risks	User input in system prompts, unsafe template rendering
Hardcoded Secrets	API keys, passwords, tokens in AI-suggested code
Insecure Code Patterns	SQL injection, command injection, path traversal
Data Exfiltration Risks	Suspicious outbound requests, data leakage patterns
Dependency Confusion	Typosquatting packages, suspicious imports

🚀 Quick Start

# Install
pip install ai-code-guard

# Scan a directory
ai-code-guard scan ./src

# Scan a single file
ai-code-guard scan ./src/api/chat.py

# Output as JSON
ai-code-guard scan ./src --format json

📊 Example Output

$ ai-code-guard scan ./my-project

🔍 AI Code Guard v0.1.0
Scanning 47 files...

┌─────────────────────────────────────────────────────────────────────┐
│ CRITICAL: SQL Injection Vulnerability                               │
├─────────────────────────────────────────────────────────────────────┤
│ File: src/db/queries.py, Line 42                                    │
│ Code: query = f"SELECT * FROM users WHERE id = {user_id}"          │
│                                                                     │
│ AI-generated code often uses f-strings for SQL queries.            │
│ Use parameterized queries instead.                                  │
│                                                                     │
│ ✅ Fix: cursor.execute("SELECT * FROM users WHERE id = ?", (id,))  │
└─────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────┐
│ HIGH: Prompt Injection Risk                                         │
├─────────────────────────────────────────────────────────────────────┤
│ File: src/api/chat.py, Line 23                                      │
│ Code: prompt = f"You are a helper. User says: {user_input}"        │
│                                                                     │
│ User input directly concatenated into LLM prompt.                   │
│ Attacker can inject malicious instructions.                         │
│                                                                     │
│ ✅ Fix: Sanitize input and use structured prompt templates          │
└─────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────┐
│ HIGH: Hardcoded API Key                                             │
├─────────────────────────────────────────────────────────────────────┤
│ File: src/config.py, Line 15                                        │
│ Code: api_key = "sk-proj-abc123..."                                 │
│                                                                     │
│ AI assistants often generate code with placeholder secrets          │
│ that developers forget to remove.                                   │
│                                                                     │
│ ✅ Fix: Use environment variables: os.environ.get("API_KEY")        │
└─────────────────────────────────────────────────────────────────────┘

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 SUMMARY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Files scanned:  47
Issues found:   3

🔴 CRITICAL:    1
🟠 HIGH:        2
🟡 MEDIUM:      0
🔵 LOW:         0
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🔧 Configuration

Create .ai-code-guard.yaml in your project root:

# Severity threshold (ignore issues below this level)
min_severity: medium

# Patterns to ignore
ignore:
- "tests/*"
- "*.test.py"
- "examples/*"

# Specific rules to disable
disable_rules:
- "SEC001"  # Hardcoded secrets (if using .env.example)

# Custom secret patterns to detect
custom_secrets:
- pattern: "my-company-api-.*"
name: "Company API Key"

📋 Rule Reference

Rule ID	Category	Description
SEC001	Secrets	Hardcoded API keys, passwords, tokens
SEC002	Secrets	AWS/GCP/Azure credentials in code
INJ001	Injection	SQL injection via string formatting
INJ002	Injection	Command injection via os.system/subprocess
INJ003	Injection	Path traversal vulnerabilities
PRI001	Prompt Injection	User input in LLM system prompts
PRI002	Prompt Injection	Unsafe prompt template rendering
PRI003	Prompt Injection	Missing input sanitization for LLM
DEP001	Dependencies	Known typosquatting packages
DEP002	Dependencies	Suspicious import patterns
EXF001	Data Exfiltration	Outbound requests with sensitive data
EXF002	Data Exfiltration	Base64 encoding of sensitive variables

🔌 CI/CD Integration

GitHub Actions

name: Security Scan

on: [push, pull_request]

jobs:
ai-code-guard:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- uses: actions/setup-python@v4
with:
python-version: '3.10'
- run: pip install ai-code-guard
- run: ai-code-guard scan ./src --format sarif > results.sarif
- uses: github/codeql-action/upload-sarif@v2
with:
sarif_file: results.sarif

Pre-commit Hook

# .pre-commit-config.yaml
repos:
- repo: https://ThorneShadowbane/ai-code-guard
rev: v0.1.0
hooks:
- id: ai-code-guard

🧠 Why AI-Generated Code Needs Special Attention

AI coding assistants are trained on vast amounts of code — including insecure patterns. Common issues include:

1. Outdated Security Practices: Training data includes old, insecure code
2. Placeholder Secrets: AI generates realistic-looking API keys as examples
3. Prompt Injection Blindspots: Most training data predates LLM security concerns
4. Context-Free Suggestions: AI doesn’t understand your security requirements

This tool specifically targets patterns commonly introduced by AI assistants.

🤝 Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

Adding New Detection Patterns

# ai_code_guard/patterns/my_pattern.py
from ai_code_guard.patterns.base import BasePattern, Finding, Severity

class MyCustomPattern(BasePattern):
"""Detect my custom security issue."""

rule_id = "CUS001"
name = "Custom Security Issue"
severity = Severity.HIGH

def scan(self, content: str, filepath: str) -> list[Finding]:
findings = []
# Your detection logic here
return findings

📚 Research Background

This tool implements patterns identified in research on AI coding assistant security vulnerabilities. Key references:

• AI Security Vulnerability Assessment Framework — Research on prompt injection and data exfiltration risks in AI coding assistants

📄 License

MIT License — see LICENSE for details.

🙏 Acknowledgments

• Security patterns informed by OWASP guidelines
• Prompt injection research from the AI security community
• Inspired by tools like Semgrep, Bandit, and GitLeaks

---

Built with 🛡️ by security engineers who use AI coding assistants daily