Vulnerability Pattern Matcher Skill for Cursor

Vulnerability Pattern Matcher

Overview

This skill detects security vulnerabilities by matching code against a comprehensive database of known vulnerability patterns, insecure coding idioms, and historical CVE-style patterns. It identifies risky code, explains why each pattern is dangerous, assesses exploitability conditions, and provides severity ratings with confidence levels.

Detection Workflow

Follow these steps to detect and assess vulnerabilities:

1. Analyze Code Structure

Identify key security-relevant elements:

• User input sources (request parameters, form data, file uploads)
• Data sinks (database queries, file operations, command execution)
• Authentication and authorization mechanisms
• Cryptographic operations
• Network communications
• File and resource handling

Map data flow:

• Trace user input through the application
• Identify where untrusted data reaches sensitive operations
• Note any validation or sanitization applied

2. Match Vulnerability Patterns

Scan for known patterns:

• Injection vulnerabilities (SQL, NoSQL, LDAP, XSS, command injection)
• Authentication issues (hardcoded credentials, weak session management)
• Cryptographic failures (weak algorithms, insecure random numbers)
• Access control problems (IDOR, missing authorization)
• Deserialization vulnerabilities
• Path traversal and file inclusion
• SSRF and XXE
• Buffer overflows and memory safety
• Race conditions

For each match:

• Identify the specific pattern
• Note the exact code location
• Determine the vulnerability category

3. Assess Exploitability

For each detected vulnerability, evaluate:

Attack Complexity:

• Low: Simple exploit, no special conditions required
• Medium: Requires specific conditions, timing, or configuration
• High: Complex exploit chain, rare conditions, or significant obstacles

Privileges Required:

• None: Unauthenticated attacker can exploit
• Low: Basic user account needed
• High: Administrative or privileged access required

User Interaction:

• None: Fully automated exploit
• Required: Victim must take action (click link, open file, etc.)

Exploitation Conditions:

• What must be true for successful exploitation?
• Are there mitigating controls present?
• What is the attack surface?

4. Determine Severity

Critical:

• Remote code execution
• Authentication bypass
• Complete system compromise
• Mass data breach

High:

• Privilege escalation
• Sensitive data exposure
• SQL injection with data access
• SSRF to internal services

Medium:

• XSS with session theft
• IDOR to user data
• Information disclosure
• Weak cryptography

Low:

• Minor information leakage
• Low-impact XSS
• Configuration issues
• Verbose error messages

5. Generate Report

For each vulnerability, provide:

• Pattern name and category
• Code location (file:line)
• Why the pattern is risky
• Exploitation conditions
• Severity level
• Detection confidence (High/Medium/Low)
• Recommended fix

Detection Examples

Example 1: SQL Injection

Code (Python):

def get_user(username):
    query = "SELECT * FROM users WHERE username = '" + username + "'"
    cursor.execute(query)
    return cursor.fetchone()

Detection:

• Pattern: SQL Injection (String Concatenation)
• Location: Line 2
• Category: Injection Vulnerability

Why Risky:

User input is directly concatenated into SQL query without parameterization. Attacker can inject SQL code to bypass authentication, extract data, or modify the database.

Exploitation Conditions:

• User controls username parameter
• No input validation or sanitization
• Database errors may be exposed to attacker

Severity: High

Confidence: High (95%)

Recommended Fix:

def get_user(username):
    query = "SELECT * FROM users WHERE username = ?"
    cursor.execute(query, (username,))
    return cursor.fetchone()

Example 2: Hardcoded Credentials

Code (JavaScript):

const config = {
    apiKey: "sk-1234567890abcdef",
    dbPassword: "admin123",
    jwtSecret: "my-secret-key"
};

Detection:

• Pattern: Hardcoded Credentials
• Location: Lines 2-4
• Category: Authentication & Session Management

Why Risky:

Credentials are embedded in source code and likely committed to version control. Anyone with repository access can authenticate as the application. Credentials are difficult to rotate if compromised.

Exploitation Conditions:

• Code repository is public or leaked
• Credentials are still valid
• No additional authentication factors

Severity: Critical

Confidence: High (100%)

Recommended Fix:

const config = {
    apiKey: process.env.API_KEY,
    dbPassword: process.env.DB_PASSWORD,
    jwtSecret: process.env.JWT_SECRET
};

Example 3: XSS via innerHTML

Code (JavaScript):

function displayWelcome(username) {
    document.getElementById('welcome').innerHTML = "Hello " + username;
}

Detection:

• Pattern: DOM-based XSS
• Location: Line 2
• Category: Cross-Site Scripting

Why Risky:

User-controlled data is directly inserted into DOM via innerHTML without encoding. Attacker can inject JavaScript to steal cookies, session tokens, or perform actions as the victim.

Exploitation Conditions:

• username contains user input
• No Content-Security-Policy
• Sensitive data in cookies or localStorage

Severity: Medium

Confidence: High (90%)

Recommended Fix:

function displayWelcome(username) {
    document.getElementById('welcome').textContent = "Hello " + username;
}

Example 4: Path Traversal

Code (Node.js):

app.get('/download', (req, res) => {
    const filename = req.query.file;
    res.sendFile('/uploads/' + filename);
});

Detection:

• Pattern: Directory Traversal
• Location: Line 3
• Category: Path Traversal

Why Risky:

User controls file path without validation. Attacker can use ../ sequences to access files outside the intended directory, potentially reading sensitive configuration files, source code, or credentials.

Exploitation Conditions:

• User controls file parameter
• No path canonicalization or validation
• Sensitive files accessible on filesystem

Severity: High

Confidence: High (95%)

Recommended Fix:

const path = require('path');
app.get('/download', (req, res) => {
    const filename = path.basename(req.query.file);
    const filepath = path.join('/uploads/', filename);
    if (!filepath.startsWith('/uploads/')) {
        return res.status(400).send('Invalid file');
    }
    res.sendFile(filepath);
});

Example 5: Insecure Deserialization

Code (Python):

import pickle

@app.route('/load', methods=['POST'])
def load_data():
    user_data = pickle.loads(request.data)
    return process(user_data)

Detection:

• Pattern: Unsafe Deserialization
• Location: Line 5
• Category: Insecure Deserialization

Why Risky:

Deserializing untrusted data with pickle can lead to arbitrary code execution. Attacker can craft malicious serialized objects that execute code during deserialization.

Exploitation Conditions:

• Application deserializes user input
• Python gadget chains available
• No integrity checks on serialized data

Severity: Critical

Confidence: High (100%)

Recommended Fix:

import json

@app.route('/load', methods=['POST'])
def load_data():
    user_data = json.loads(request.data)
    return process(user_data)

Example 6: SSRF via URL Parameter

Code (Python):

@app.route('/fetch')
def fetch_url():
    url = request.args.get('url')
    response = requests.get(url)
    return response.content

Detection:

• Pattern: Server-Side Request Forgery
• Location: Line 4
• Category: SSRF

Why Risky:

Application makes HTTP requests to user-controlled URLs. Attacker can access internal services, cloud metadata endpoints (169.254.169.254), or perform port scanning of internal network.

Exploitation Conditions:

• User controls URL parameter
• Internal network accessible from server
• Cloud metadata endpoint available
• No URL allowlist or validation

Severity: High

Confidence: High (95%)

Recommended Fix:

from urllib.parse import urlparse

ALLOWED_HOSTS = ['api.example.com', 'cdn.example.com']

@app.route('/fetch')
def fetch_url():
    url = request.args.get('url')
    parsed = urlparse(url)
    if parsed.hostname not in ALLOWED_HOSTS:
        return "Invalid URL", 400
    response = requests.get(url, timeout=5)
    return response.content

Confidence Levels

High Confidence (90-100%):

• Exact pattern match with known vulnerability
• Clear exploit path identified
• No mitigating controls present
• Well-documented vulnerability type

Medium Confidence (60-89%):

• Pattern resembles known vulnerability
• Exploitation requires specific conditions
• Some mitigating controls may be present
• Context suggests vulnerability but not certain

Low Confidence (30-59%):

• Potential vulnerability identified
• Unclear or complex exploit path
• Multiple mitigating factors present
• Requires manual verification
• May be false positive

Constraints

MUST:

• Identify specific vulnerability patterns in code
• Explain why each pattern is risky
• Assess exploitability conditions
• Provide severity and confidence ratings
• Reference specific code locations (file:line)
• Suggest concrete fixes

MUST NOT:

• Report theoretical vulnerabilities without code evidence
• Provide exploits or attack code
• Assume vulnerabilities without pattern match
• Report false positives without noting low confidence
• Ignore mitigating controls when assessing severity

Report Format

Vulnerability: [Pattern Name]
Category: [Vulnerability Category]
Location: [file:line]
Severity: [Critical/High/Medium/Low]
Confidence: [High/Medium/Low] ([percentage]%)

Description:
[Why this pattern is risky]

Exploitation Conditions:
- [Condition 1]
- [Condition 2]
- [Condition 3]

Vulnerable Code:
[Code snippet]

Recommended Fix:
[Safe code example]

Resources

references/vulnerability_patterns.md

Comprehensive catalog of security vulnerability patterns including:

• Injection vulnerabilities (SQL, NoSQL, LDAP, XSS, command injection)
• Authentication and session management issues
• Cryptographic failures
• Access control problems
• Deserialization vulnerabilities
• Path traversal and file inclusion
• SSRF and XXE
• Buffer overflows
• Race conditions
• CVE-style pattern examples
• Exploitability assessment framework