Security Best Practices

Security Best Practices Skill

When writing code, follow these security best practices to prevent common vulnerabilities.

Input Validation

Always validate and sanitize user input:

# ❌ Bad - trusting user input
user_id = request.args.get('id')
query = f"SELECT * FROM users WHERE id = {user_id}"

# ✅ Good - parameterized query
user_id = request.args.get('id')
cursor.execute("SELECT * FROM users WHERE id = ?", (user_id,))

Validation Checklist

• Validate type, length, format, and range
• Use allowlists over denylists
• Sanitize for the specific output context
• Never trust client-side validation alone

Authentication

Implement secure authentication:

• Use established libraries (e.g., bcrypt, argon2)
• Never store passwords in plain text
• Implement rate limiting on login attempts
• Use secure session management
• Implement proper logout handling

# ✅ Good - using bcrypt
import bcrypt

def hash_password(password: str) -> bytes:
    return bcrypt.hashpw(password.encode(), bcrypt.gensalt())

def verify_password(password: str, hash: bytes) -> bool:
    return bcrypt.checkpw(password.encode(), hash)

Authorization

Check permissions on every request:

# ✅ Good - checking authorization
def delete_post(post_id: int, user: User):
    post = get_post(post_id)
    if post.author_id != user.id and not user.is_admin:
        raise PermissionError("Not authorized")
    delete_post_from_db(post_id)

Secrets Management

Never hardcode secrets:

# ❌ Bad
API_KEY = "sk-1234567890abcdef"

# ✅ Good
import os
API_KEY = os.environ.get('API_KEY')

Environment Variables

• Use .env files for development (in .gitignore)
• Use secrets managers in production
• Rotate secrets regularly
• Use different secrets per environment

SQL Injection Prevention

Always use parameterized queries:

# ❌ Bad - string concatenation
cursor.execute(f"SELECT * FROM users WHERE name = '{name}'")

# ✅ Good - parameterized
cursor.execute("SELECT * FROM users WHERE name = ?", (name,))

# ✅ Good - ORM
User.query.filter_by(name=name).first()

XSS Prevention

Escape output and use Content Security Policy:

# ✅ Good - escaping in templates (Jinja2)
{{ user_input }}  # Auto-escaped

# ✅ Good - explicit escaping
from markupsafe import escape
safe_output = escape(user_input)

// ❌ Bad
element.innerHTML = userInput;

// ✅ Good
element.textContent = userInput;

CSRF Protection

Implement CSRF tokens:

    
    ...

Secure Headers

Set security headers:

# Flask example
@app.after_request
def set_secure_headers(response):
    response.headers['X-Content-Type-Options'] = 'nosniff'
    response.headers['X-Frame-Options'] = 'DENY'
    response.headers['X-XSS-Protection'] = '1; mode=block'
    response.headers['Content-Security-Policy'] = "default-src 'self'"
    return response

Error Handling

Don't expose sensitive information in errors:

# ❌ Bad - exposing stack trace
except Exception as e:
    return {"error": str(e), "stack": traceback.format_exc()}

# ✅ Good - generic error message
except Exception as e:
    logger.error(f"Error: {e}", exc_info=True)
    return {"error": "An internal error occurred"}

Dependency Security

Keep dependencies updated:

# Check for vulnerabilities
pip-audit
npm audit
snyk test

File Operations

Validate file paths to prevent path traversal:

import os

# ✅ Good - validate path is within allowed directory
def safe_read(user_path: str, base_dir: str) -> str:
    full_path = os.path.realpath(os.path.join(base_dir, user_path))
    if not full_path.startswith(os.path.realpath(base_dir)):
        raise ValueError("Path traversal attempt")
    return open(full_path).read()

Logging

Log security events but not sensitive data:

# ❌ Bad - logging passwords
logger.info(f"Login attempt: user={username}, password={password}")

# ✅ Good - logging without sensitive data
logger.info(f"Login attempt: user={username}, success={success}")

Quick Reference

| Vulnerability | Prevention | |--------------|------------| | SQL Injection | Parameterized queries | | XSS | Output encoding, CSP | | CSRF | CSRF tokens | | Path Traversal | Path validation | | Command Injection | Avoid shell=True | | Hardcoded Secrets | Environment variables | | Weak Passwords | bcrypt/argon2, complexity rules |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: tao12345666333
• Source: tao12345666333/amcp
• License: Apache-2.0
• Homepage: https://x.com/zhangjintao9020/status/1995170132973466018?s=20

Install and usage instructions live in the source repository linked above.