Protection System

Why Protection Matters

AI assistants are powerful but can accidentally break working code:

Protection annotations tell AI: "This code works. Don't change it without good reason."

Protection Levels

/**
 * @immutable
 * REASON: Core authentication logic - validated through security audit
 * VALIDATED: 2025-01-10 by security team
 */
function validateJWT(token: string): TokenPayload {
  // ... tested, working code
}

/**
 * @protected
 * REASON: Used by 12 other components - changes have ripple effects
 * DEPENDENCIES: UserProfile, Dashboard, Settings, ...
 */
function formatUserData(user: User): FormattedUser {
  // ... working code with many dependents
}

/**
 * @maintainable
 * INTERFACE: Must accept (items: T[]) and return T[]
 * BEHAVIOR: Must preserve order, remove duplicates
 */
function deduplicateArray<T>(items: T[]): T[] {
  // Implementation can be improved
}

Adding Protection

Manual Annotation

Add JSDoc-style comments above functions:

/**
 * @protected
 * REASON: Payment processing - must not break
 * VALIDATED: 2025-01-15
 */
async function processPayment(amount: number): Promise<Receipt> {
  // ...
}

Using the /protect Command

Use the built-in skill to analyze and protect code:

This scans the directory and suggests protection levels based on usage patterns.

How AI Respects Protection

When protection annotations are in your code, AI assistants like Claude Code and Cursor will:

• ✓Avoid modifying @immutable code entirely
• ✓Warn before suggesting changes to @protected code
• ✓Preserve interface and behavior for @maintainable code
• ✓Explain why they're avoiding certain changes

The annotations work because AI models are trained to recognize and respect these patterns when they see them in code.