Prompt Hardening & Jailbreak Prevention

Implement multiple layers of prompt security including structural defenses, behavioral constraints, runtime detection, and advanced techniques to prevent prompt injection, model manipulation, and jailbreak attempts.

Mapped Threats:

Cascading Hallucination
Data Leakage
Model Theft & Extraction
T15
Memory Poisoning

Implementation Guidance:

Design prompt templates with explicit boundaries, monitor for jailbreak patterns, and use runtime validation.