bszczyglo/
Menu

Topic

Agent Security

Prompt injection, fail-closed guardrails, untrusted content and keeping secrets out of a model's reach.