Week 13 — Trust & Security · Retrospective

Material	Where
Lecture notes	published before class
Lab	published before class
Papers	Indirect Prompt Injection (Greshake et al., 2023) (PDF) · InjecAgent (Zhan et al., 2024) (PDF)

This week

Agents create a new attack surface: attacks on the system, not the model. Direct vs. indirect prompt injection, the lethal trifecta, and structural vs. detection defenses. Closing retrospective: the semester map (brain / perception / action / operations) and open research problems.

Lab plan

Weekly LangChain and LangGraph practice follows LangChain Academy and official RAG tutorials.