3 months ago
Scan any GitHub repository for 21 AI/ML vulnerability patterns across Python, Java, Go, C++, and Rust. Detects eval injection, pickle deserialization, SSRF, command injection, SQL injection, unsafe YAML, hardcoded secrets, and more. Severity-weighted risk scoring with line-level findings.
概要
Install
pip install mcp-security-audit
Tools
- audit_repo — Scan a GitHub repository for vulnerabilities
- audit_code — Scan a code snippet directly
- list_patterns — List all 21 detection patterns
21 Patterns Across 5 Languages
Python (9), Java (3), Go (3), C++ (3), Rust (3)
Covers CRITICAL through LOW severity: eval/exec injection, deserialization attacks, command injection, SSRF, SQL injection, buffer overflow, format strings, unsafe blocks, and hardcoded secrets.
Links
サーバー設定
{
"mcpServers": {
"security-audit": {
"command": "python",
"args": [
"-m",
"mcp_security_audit"
],
"env": {}
}
}
}