Indirect injections
A section of the "Security" topic. The questions check an understanding of practice, not memory of specific library APIs. After each answer comes a review: why the correct option is right, what is wrong with each incorrect one and which chapter to read.
Quiz difficulty Formats: one correct option, several correct options, the mistake in an agent trace, sorting into groups
01
What we check
- A hidden instruction on a web page
- Direct and indirect injection
- An injection in a résumé
- The lethal trifecta
- How to break the trifecta
- Injections through a knowledge base
- Injection through long-term memory
- Page text is data
02
Where to read
- Li, AI Agents in Depthsec. 2.4.7 · sec. 3.3.4 · sec. 4.3 · sec. 5.1.9
- Huyen, AI Engineeringch. 5
- Wilson, The Developer's Playbook for LLM Securitych. 4 · ch. 4, Direct Versus Indirect Prompt Injection · ch. 7