Agent Skills: Security Checklist

Security Checklist

Overview

This skill provides comprehensive security guidance for building secure applications. Whether performing a security audit, implementing new features, or hardening existing systems, this framework helps identify and mitigate common vulnerabilities.

When to use this skill:

• Conducting security audits or reviews
• Implementing authentication and authorization
• Validating and sanitizing user input
• Handling sensitive data (PII, credentials, payment info)
• Ensuring compliance (GDPR, HIPAA, SOC2)
• Preparing for security assessments or penetration tests
• Reviewing third-party dependencies for vulnerabilities

Required Tools

This skill requires the following tools to be installed on your system:

For JavaScript/TypeScript Projects

• Node.js 18+ with npm
• Command: npm audit
• Install: Node.js comes with npm pre-installed

For Python Projects

• Python 3.8+ with pip
• pip-audit: Security scanner for Python dependencies
  • Install: pip install pip-audit
  • Command: pip-audit

Optional (Advanced Security Scanning)

• Semgrep: Static analysis tool

  • Install (macOS): brew install semgrep
  • Install (pip): pip install semgrep
  • Command: semgrep --config=auto .

• Bandit: Python security linter

  • Install: pip install bandit
  • Command: bandit -r .

• TruffleHog: Secrets detection

  • Install (macOS): brew install trufflesecurity/trufflehog/trufflehog
  • Install (Go): go install github.com/trufflesecurity/trufflehog/v3@latest
  • Command: trufflehog filesystem .

Installation Verification

# Verify Node.js & npm
node --version
npm --version

# Verify Python & pip
python --version
pip --version

# Verify pip-audit
pip-audit --version

# Verify optional tools
semgrep --version
bandit --version
trufflehog --version

Note: The skill will automatically detect which tools are available and use appropriate commands for your project type.

Security Principles

Defense in Depth

• Multiple layers of security controls
• Assume each layer can fail, design redundancy
• Security at database, application, network, and infrastructure levels

Least Privilege

• Grant minimum permissions necessary
• Separate read/write database accounts
• Service accounts with limited scope

Fail Securely

• Errors don't expose sensitive information
• Authentication failures don't reveal if user exists
• Rate limiting prevents brute force attacks

Don't Trust User Input

• All input is untrusted until validated
• Validate, sanitize, and escape
• Apply principle to query params, headers, cookies, POST data

OWASP Top 10 (2021 Edition)

1. Broken Access Control

Vulnerability: Users can access resources they shouldn't.

Examples:

# ❌ Bad: No authorization check
@app.route('/api/users/<user_id>')
def get_user(user_id):
    return db.query(f"SELECT * FROM users WHERE id = {user_id}")

# ✅ Good: Verify user can access this resource
@app.route('/api/users/<user_id>')
@login_required
def get_user(user_id):
    current_user = get_current_user()
    if current_user.id != user_id and not current_user.is_admin:
        abort(403, "Forbidden")
    return db.query("SELECT * FROM users WHERE id = ?", [user_id])

Mitigations:

• Deny by default (require explicit authorization)
• Enforce ownership checks (users can only access their own data)
• Implement RBAC (Role-Based Access Control)
• Test for IDOR (Insecure Direct Object References)
• Log access control failures

2. Cryptographic Failures

Vulnerability: Sensitive data exposed due to weak or missing encryption.

Examples:

# ❌ Bad: Storing passwords in plaintext
user.password = request.form['password']

# ✅ Good: Hashing passwords with bcrypt
from bcrypt import hashpw, gensalt

hashed = hashpw(password.encode('utf-8'), gensalt())
user.password_hash = hashed

# ❌ Bad: Using weak hashing (MD5, SHA1)
import hashlib
password_hash = hashlib.md5(password.encode()).hexdigest()

# ✅ Good: Using strong hashing (bcrypt, argon2, scrypt)
from argon2 import PasswordHasher

ph = PasswordHasher()
password_hash = ph.hash(password)

Mitigations:

• Use TLS/HTTPS for all traffic (enforce, not optional)
• Hash passwords with bcrypt, argon2, or scrypt
• Encrypt sensitive data at rest (PII, payment info)
• Never store credit card numbers (use tokenization)
• Use strong random number generators (secrets module in Python)
• Rotate encryption keys regularly

3. Injection (SQL, NoSQL, Command, LDAP)

Vulnerability: Untrusted data sent to an interpreter as part of a command.

SQL Injection:

# ❌ Bad: String concatenation (vulnerable to SQL injection)
query = f"SELECT * FROM users WHERE email = '{email}'"
db.execute(query)

# ✅ Good: Parameterized queries
query = "SELECT * FROM users WHERE email = ?"
db.execute(query, [email])

Command Injection:

# ❌ Bad: Shell=True with user input
import subprocess
filename = request.form['filename']
subprocess.run(f"cat {filename}", shell=True)

# ✅ Good: Avoid shell, use list arguments
subprocess.run(["cat", filename], shell=False)

Mitigations:

• Use parameterized queries (prepared statements)
• Use ORMs with proper query builders
• Validate and sanitize all input
• Avoid eval(), exec(), shell=True
• Use allowlists for permitted values
• Escape special characters in dynamic queries

4. Insecure Design

Vulnerability: Design flaws that can't be fixed with implementation.

Examples:

• No rate limiting on login/password reset
• Unlimited file upload sizes
• Sequential or guessable IDs
• Password reset without account verification

Mitigations:

• Threat modeling during design phase
• Rate limiting on all public endpoints
• Use UUIDs instead of sequential IDs for public resources
• Require email verification for password resets
• Implement CAPTCHA for sensitive operations
• Design for secure defaults (opt-in, not opt-out)

5. Security Misconfiguration

Vulnerability: Default configs, incomplete setups, verbose errors.

Examples:

# ❌ Bad: Debug mode in production
app.debug = True

# ✅ Good: Debug mode only in development
app.debug = os.getenv('FLASK_ENV') == 'development'

# ❌ Bad: Verbose error messages
@app.errorhandler(Exception)
def handle_error(e):
    return str(e), 500  # Exposes stack traces

# ✅ Good: Generic error messages
@app.errorhandler(Exception)
def handle_error(e):
    logger.error(f"Error: {e}")
    return {"error": "Internal server error"}, 500

Mitigations:

• Disable debug mode in production
• Remove default credentials
• Close unnecessary ports and services
• Set security headers (CSP, X-Frame-Options, HSTS)
• Keep software updated (dependencies, frameworks, OS)
• Use environment variables for secrets (not hardcoded)

6. Vulnerable and Outdated Components

Vulnerability: Using libraries with known vulnerabilities.

Mitigations:

# Check for vulnerabilities
npm audit
npm audit fix

# Python
pip-audit
safety check

Best Practices:

• Pin dependency versions in lock files
• Scan dependencies regularly (CI/CD integration)
• Subscribe to security advisories (GitHub Dependabot, Snyk)
• Update dependencies regularly (monthly at minimum)
• Remove unused dependencies

7. Identification and Authentication Failures

Vulnerability: Weak authentication, credential stuffing, session hijacking.

Examples:

# ❌ Bad: Weak password requirements
if len(password) < 6:
    return "Password too short"

# ✅ Good: Strong password requirements
import re

def validate_password(password):
    if len(password) < 12:
        return "Password must be at least 12 characters"
    if not re.search(r"[A-Z]", password):
        return "Password must contain uppercase letter"
    if not re.search(r"[a-z]", password):
        return "Password must contain lowercase letter"
    if not re.search(r"[0-9]", password):
        return "Password must contain a number"
    return None  # Valid

Mitigations:

• Require strong passwords (12+ chars, mixed case, numbers, symbols)
• Implement multi-factor authentication (MFA)
• Rate limit login attempts (5 attempts per 15 minutes)
• Use secure session management (HTTPOnly, Secure, SameSite cookies)
• Invalidate sessions after logout or inactivity
• Don't reveal whether email exists during password reset
• Implement account lockout after failed attempts

8. Software and Data Integrity Failures

Vulnerability: Code or infrastructure updates without integrity verification.

Examples:

• Loading libraries from untrusted CDNs
• No verification of CI/CD pipeline artifacts
• Auto-update without signature verification

Mitigations:

• Use Subresource Integrity (SRI) for CDN scripts

<script src="https://cdn.example.com/lib.js"
        integrity="sha384-oqVuAfXRKap7fdgcCY5uykM6+R9GqQ8K/uxy9rx7HNQlGYl1kPzQho1wx4JwY8wC"
        crossorigin="anonymous"></script>

• Verify package signatures (npm, PyPI)
• Review code changes before deployment
• Use signed commits and releases

9. Security Logging and Monitoring Failures

Vulnerability: Insufficient logging prevents detection of breaches.

Examples:

# ❌ Bad: No logging
@app.route('/login', methods=['POST'])
def login():
    user = authenticate(email, password)
    return {"token": create_token(user)}

# ✅ Good: Log security events
import logging

@app.route('/login', methods=['POST'])
def login():
    email = request.form['email']
    user = authenticate(email, password)

    if user:
        logger.info(f"Successful login: {email}")
        return {"token": create_token(user)}
    else:
        logger.warning(f"Failed login attempt: {email}")
        return {"error": "Invalid credentials"}, 401

Mitigations:

• Log all authentication events (login, logout, failed attempts)
• Log authorization failures (403 errors)
• Log input validation failures
• Log security-relevant events (password changes, MFA changes)
• Centralize logs (ELK, Splunk, CloudWatch)
• Set up alerts for suspicious patterns
• Never log passwords, tokens, or sensitive data

10. Server-Side Request Forgery (SSRF)

Vulnerability: Application fetches remote resources without validating URL.

Examples:

# ❌ Bad: Fetching user-provided URL without validation
import requests

@app.route('/fetch')
def fetch():
    url = request.args.get('url')
    response = requests.get(url)  # Can access internal services!
    return response.text

# ✅ Good: Validate URL and use allowlist
from urllib.parse import urlparse

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

@app.route('/fetch')
def fetch():
    url = request.args.get('url')
    parsed = urlparse(url)

    if parsed.hostname not in ALLOWED_DOMAINS:
        abort(400, "Invalid domain")

    response = requests.get(url, timeout=5)
    return response.text

Mitigations:

• Validate and sanitize all URLs
• Use allowlist of permitted domains
• Disable redirects or validate redirect targets
• Block access to internal IPs (127.0.0.1, 10.0.0.0/8, 192.168.0.0/16)
• Network segmentation (separate internal and external services)

Authentication & Authorization

Password Security

# ✅ Secure password hashing
from argon2 import PasswordHasher

ph = PasswordHasher()

# Hashing
password_hash = ph.hash(password)

# Verification
try:
    ph.verify(password_hash, password)
    # Password correct
except:
    # Password incorrect
    pass

Requirements:

• Minimum 12 characters
• Mix of upper/lowercase, numbers, symbols
• No common passwords (use haveibeenpwned)
• Bcrypt, Argon2, or scrypt for hashing
• Salt automatically handled by these algorithms

Session Management

# ✅ Secure session cookies
app.config['SESSION_COOKIE_SECURE'] = True      # HTTPS only
app.config['SESSION_COOKIE_HTTPONLY'] = True    # No JavaScript access
app.config['SESSION_COOKIE_SAMESITE'] = 'Strict'  # CSRF protection
app.config['PERMANENT_SESSION_LIFETIME'] = timedelta(hours=1)

JWT Tokens

import jwt
from datetime import datetime, timedelta

# ✅ Secure JWT generation
def create_token(user_id):
    payload = {
        'user_id': user_id,
        'exp': datetime.utcnow() + timedelta(hours=1),  # Expiration
        'iat': datetime.utcnow(),  # Issued at
    }
    return jwt.encode(payload, SECRET_KEY, algorithm='HS256')

# ✅ Secure JWT verification
def verify_token(token):
    try:
        payload = jwt.decode(token, SECRET_KEY, algorithms=['HS256'])
        return payload['user_id']
    except jwt.ExpiredSignatureError:
        return None  # Token expired
    except jwt.InvalidTokenError:
        return None  # Invalid token

Input Validation & Sanitization

Validation

# ✅ Allowlist validation
def validate_sort_column(column):
    allowed_columns = ['name', 'email', 'created_at']
    if column not in allowed_columns:
        raise ValueError("Invalid sort column")
    return column

# ✅ Type validation
from pydantic import BaseModel, EmailStr, constr

class UserCreate(BaseModel):
    email: EmailStr
    name: constr(min_length=2, max_length=100)
    age: int = Field(ge=0, le=150)

# Usage
try:
    user = UserCreate(**request.json)
except ValidationError as e:
    return {"errors": e.errors()}, 400

Sanitization

# ✅ HTML escaping
from markupsafe import escape

@app.route('/comment', methods=['POST'])
def create_comment():
    content = escape(request.form['content'])
    db.execute("INSERT INTO comments (content) VALUES (?)", [content])
    return {"status": "ok"}

Security Headers

# ✅ Set security headers
@app.after_request
def set_security_headers(response):
    response.headers['Content-Security-Policy'] = "default-src 'self'"
    response.headers['X-Content-Type-Options'] = 'nosniff'
    response.headers['X-Frame-Options'] = 'DENY'
    response.headers['X-XSS-Protection'] = '1; mode=block'
    response.headers['Strict-Transport-Security'] = 'max-age=31536000; includeSubDomains'
    return response

🔍 Automated Security Scanning (v3.5.0)

Overview

Automated security scanning catches vulnerabilities early. This section teaches agents HOW to run security tools and record evidence.

When to auto-scan:

• Before marking code review as complete
• After installing/updating dependencies
• During CI/CD pipeline execution
• When adding new external integrations
• Before production deployments

Scanning Workflow

1. Identify Scan Type (dependencies, code, configuration)
2. Run Appropriate Tool (npm audit, pip-audit, semgrep)
3. Capture Results (exit codes, vulnerability counts)
4. Record Evidence in Context
5. Escalate Critical Findings

1. Dependency Vulnerability Scanning

JavaScript/TypeScript (NPM)

# Run npm audit and capture results
npm audit --json > security-audit.json
EXIT_CODE=$?

# Check exit code
if [ $EXIT_CODE -eq 0 ]; then
  echo "✅ No vulnerabilities found"
else
  echo "⚠️ Vulnerabilities detected (exit code: $EXIT_CODE)"
fi

# Parse for critical/high vulnerabilities
CRITICAL=$(npm audit --json | jq '.metadata.vulnerabilities.critical')
HIGH=$(npm audit --json | jq '.metadata.vulnerabilities.high')

if [ "$CRITICAL" -gt 0 ] || [ "$HIGH" -gt 0 ]; then
  echo "🚨 CRITICAL: $CRITICAL critical, $HIGH high severity vulnerabilities"
fi

Record evidence:

context.quality_evidence = context.quality_evidence || { last_updated: new Date().toISOString() };
context.quality_evidence.security_scan = {
  executed: true,
  tool: 'npm audit',
  critical: 2,
  high: 5,
  moderate: 10,
  low: 3,
  timestamp: new Date().toISOString()
};
context.writeContext();

Python (pip-audit / safety)

# Using pip-audit (official tool)
pip-audit --format=json > security-audit.json
EXIT_CODE=$?

# Alternative: using safety
safety check --json > security-audit.json

# Check for critical vulnerabilities
CRITICAL_COUNT=$(cat security-audit.json | jq '[.vulnerabilities[] | select(.severity == "critical")] | length')

Evidence recording:

# Record in context
context['quality_evidence'] = context.get('quality_evidence', {})
context['quality_evidence']['security_scan'] = {
    'executed': True,
    'tool': 'pip-audit',
    'critical': critical_count,
    'high': high_count,
    'moderate': moderate_count,
    'low': low_count,
    'timestamp': datetime.now().isoformat()
}

2. Static Code Analysis (SAST)

Semgrep (Multi-language)

# Run Semgrep with security rules
semgrep --config=auto --json > semgrep-results.json
EXIT_CODE=$?

# Count findings by severity
CRITICAL=$(cat semgrep-results.json | jq '[.results[] | select(.extra.severity == "ERROR")] | length')
HIGH=$(cat semgrep-results.json | jq '[.results[] | select(.extra.severity == "WARNING")] | length')

Common security patterns detected:

• SQL Injection
• XSS (Cross-Site Scripting)
• Command Injection
• Path Traversal
• Hardcoded secrets
• Insecure cryptography

Bandit (Python)

# Run Bandit for Python security issues
bandit -r . -f json -o bandit-report.json
EXIT_CODE=$?

# Count high/medium severity issues
HIGH=$(cat bandit-report.json | jq '[.results[] | select(.issue_severity == "HIGH")] | length')
MEDIUM=$(cat bandit-report.json | jq '[.results[] | select(.issue_severity == "MEDIUM")] | length')

3. Secret Detection

TruffleHog / Gitleaks

# Scan for secrets in git history
trufflehog git file://. --json > secrets-scan.json

# Check if any secrets found
SECRET_COUNT=$(cat secrets-scan.json | jq '. | length')

if [ "$SECRET_COUNT" -gt 0 ]; then
  echo "🚨 CRITICAL: $SECRET_COUNT secrets detected!"
  # Extract types
  cat secrets-scan.json | jq -r '.[] | .DetectorType' | sort | uniq
fi

Common secrets detected:

• AWS API keys
• GitHub tokens
• Private keys (RSA, SSH)
• Database credentials
• API keys (Stripe, Twilio, etc.)

4. Container Security (Docker)

# Scan Docker images with Trivy
trivy image myapp:latest --format json > trivy-scan.json

# Count vulnerabilities
CRITICAL=$(cat trivy-scan.json | jq '[.Results[].Vulnerabilities[]? | select(.Severity == "CRITICAL")] | length')
HIGH=$(cat trivy-scan.json | jq '[.Results[].Vulnerabilities[]? | select(.Severity == "HIGH")] | length')

5. Evidence Recording Template

After running security scans, record evidence in shared context:

import { ContextManager } from '../lib/context/context-manager.js';

const context = new ContextManager();

// Record security scan evidence
context.recordSecurityScanEvidence({
  executed: true,
  tool: 'npm audit + semgrep',
  critical: 2,
  high: 5,
  moderate: 10,
  low: 3,
  timestamp: new Date().toISOString(),
  scan_details: {
    dependency_scan: {
      tool: 'npm audit',
      critical: 2,
      high: 3,
      vulnerabilities: [
        { id: 'GHSA-xxxx', severity: 'critical', package: 'lodash@4.17.19' }
      ]
    },
    code_scan: {
      tool: 'semgrep',
      critical: 0,
      high: 2,
      patterns: ['sql-injection', 'xss']
    }
  }
});

6. Critical Threshold Escalation

MANDATORY: Escalate if critical/high vulnerabilities found

// After scanning
const securityEvidence = context.getQualityEvidence()?.security_scan;

if (!securityEvidence) {
  console.log('⚠️ WARNING: No security scan performed');
  return;
}

// Check for critical/high vulnerabilities
if (securityEvidence.critical > 0 || securityEvidence.high > 5) {
  console.log('🚨 SECURITY ALERT: Critical vulnerabilities detected');

  // BLOCK deployment
  const blockingReasons = [];

  if (securityEvidence.critical > 0) {
    blockingReasons.push(`${securityEvidence.critical} CRITICAL vulnerabilities`);
  }

  if (securityEvidence.high > 5) {
    blockingReasons.push(`${securityEvidence.high} HIGH vulnerabilities (>5 threshold)`);
  }

  // Escalate to user
  console.log('BLOCKED: ' + blockingReasons.join(', '));
  console.log('Action Required: Fix critical/high vulnerabilities before proceeding');

  return { approved: false, blockingReasons };
}

console.log('✅ Security scan passed');

Escalation Thresholds:

• CRITICAL: Any critical vulnerability → BLOCK
• HIGH: >5 high severity vulnerabilities → BLOCK
• MODERATE: >20 moderate vulnerabilities → WARNING
• LOW: >50 low vulnerabilities → WARNING

7. Auto-Scan Checklist

Use this checklist when performing security reviews:

## Security Scan Checklist

- [ ] **Dependency Scan**: npm audit / pip-audit executed
- [ ] **Exit Code Captured**: 0 = clean, non-zero = vulnerabilities
- [ ] **Severity Counts**: Critical, High, Moderate, Low recorded
- [ ] **Evidence Recorded**: Added to context.quality_evidence.security_scan
- [ ] **Critical Threshold Check**: BLOCK if critical > 0 or high > 5
- [ ] **Scan Results Saved**: JSON output saved for review
- [ ] **False Positives Noted**: Known safe issues documented
- [ ] **Fix Recommendations**: Upgrade paths or mitigations documented

8. Integration with Code Quality Reviewer

When Code Quality Reviewer agent performs review:

1. Run linter/type checker (already implemented)
2. **AUTO-TRIGGER**: Run security scan
   - npm audit (for JS/TS projects)
   - pip-audit (for Python projects)
3. Capture and record evidence
4. Check critical thresholds
5. BLOCK approval if critical vulnerabilities found
6. Include security scan summary in review output

Example output:

## Code Quality Review

### Lint & Type Check: ✅ PASS
- ESLint: 0 errors, 2 warnings
- TypeScript: 0 errors

### Security Scan: ⚠️ WARNING
- Tool: npm audit
- Critical: 0
- High: 3
- Moderate: 8
- Low: 2

**Recommendation**: 3 high severity vulnerabilities detected. Run `npm audit fix` to address:
- lodash@4.17.19 (Prototype Pollution - High)
- minimist@1.2.5 (Prototype Pollution - High)
- axios@0.21.1 (SSRF - High)

### Overall Status: BLOCKED
Security vulnerabilities must be resolved before approval.

9. Tool Installation Guide

JavaScript/TypeScript:

# npm audit (built-in, no install needed)
npm audit

# Semgrep
pip install semgrep

# TruffleHog
docker run --rm trufflesecurity/trufflehog:latest

Python:

# pip-audit (official tool)
pip install pip-audit

# safety
pip install safety

# Bandit
pip install bandit

General:

# Trivy (containers, dependencies, code)
brew install aquasecurity/trivy/trivy

# Gitleaks (secrets)
brew install gitleaks

Compliance

GDPR (General Data Protection Regulation)

• [ ] Data Inventory: Know what personal data you collect
• [ ] Lawful Basis: Have legal basis for processing (consent, contract, etc.)
• [ ] Privacy Policy: Clear, accessible privacy policy
• [ ] User Rights: Implement data access, deletion, portability
• [ ] Data Minimization: Collect only necessary data
• [ ] Breach Notification: Process for reporting breaches within 72 hours

SOC 2 (Service Organization Control)

• [ ] Access Controls: Role-based access, MFA, least privilege
• [ ] Encryption: Data encrypted at rest and in transit
• [ ] Logging & Monitoring: Audit logs for all security events
• [ ] Incident Response: Documented incident response plan
• [ ] Vendor Management: Third-party security assessments

Integration with Agents

Code Quality Reviewer

• Applies security checklist during code reviews
• Identifies SQL injection, XSS, and other vulnerabilities
• Validates authentication and authorization patterns

Backend System Architect

• Designs systems with security in mind (defense in depth)
• Plans for least privilege access
• Implements rate limiting and DDOS protection

Frontend UI Developer

• Implements CSP headers
• Prevents XSS through output encoding
• Validates input on client-side (with server-side validation)

Quick Start Checklist

When securing an application:

• [ ] All input validated and sanitized
• [ ] SQL injection prevented (parameterized queries)
• [ ] XSS prevented (output encoding)
• [ ] CSRF protection enabled
• [ ] HTTPS enforced (no HTTP in production)
• [ ] Security headers set (CSP, X-Frame-Options, HSTS)
• [ ] Passwords hashed with bcrypt/argon2
• [ ] Rate limiting on sensitive endpoints
• [ ] Authentication required for protected routes
• [ ] Authorization checks on all data access
• [ ] Secrets in environment variables (not code)
• [ ] Dependencies scanned for vulnerabilities
• [ ] Error messages don't leak information
• [ ] Logging enabled for security events
• [ ] MFA available for users

Skill Version: 1.0.0 Last Updated: 2025-10-31 Maintained by: AI Agent Hub Team