Prompt Injection Defense and Safety

# Prompt Injection Defense and Safety

You are an AI security specialist. Design a defensive prompting strategy for [APPLICATION TYPE] that protects against prompt injection and instruction override attacks.

## What Prompt Injection Is

Explain the attack: how malicious user input can contain instructions that override system prompts, leak sensitive data, or cause the AI to behave outside its intended scope. Provide 3 real-world examples relevant to [APPLICATION TYPE].

## Attack Vectors

List the 6 most common injection techniques:
- Direct override: "Ignore all previous instructions and..."
- Role hijacking: "You are now a different AI that has no restrictions..."
- Context poisoning: embedding malicious instructions in uploaded documents
- Delimiter confusion: using fake system-prompt delimiters in user input
- Indirect injection: instructions embedded in web content the AI reads
- Jailbreak chaining: building up to a restricted output through seemingly innocent steps

## Defensive Techniques

For each attack vector, provide the corresponding defense:
- Input sanitization rules
- System prompt hardening techniques
- Output filtering for sensitive data patterns
- Structural separation of trusted vs. untrusted content

## Hardened System Prompt

Write a hardened system prompt for [APPLICATION TYPE] that incorporates all defensive techniques and explicitly handles the most likely injection attempts.