Defend Against Prompt Injection

Separate trusted instructions from untrusted user and retrieved content, then test that policy under attack.

In this module

• Instructions, Data, and Trust Boundaries
• Layered Injection Defenses
• Prove the Boundary Holds