Write a system prompt that's resistant to common prompt injection attacks. Users will try to make the AI ignore its instructions — your job is to make that nearly impossible.

Constraints:

• Must handle 5 common injection patterns
• Maintain helpfulness for legitimate requests
• Log attempted injections