Bullet-proof your custom GPT system prompt security with KEVLAR, the ultimate prompt protector against rules extraction, prompt injections, and leaks of AI agent secret instructions.