// ⛨ Safety, Evaluation & Security
Indirect Prompt Injection
Hides malicious instructions in content the model reads.
间接提示注入 2023 Foundational core
IN PLAIN WORDS
When agents read external data, embedded instructions may be followed as commands, causing leaks or unauthorized actions — the top risk for agents and RAG apps.
Used in: Security of agents, browser and email assistants
TIMELINE · prerequisites → this → what builds on it
01Original paper
Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection
Kai Greshake, Sahar Abdelnabi, Shailesh Mishra, et al. · 2023
arXiv ↗ ✓ Verified · arXiv API
02Where it sits
Learn the left column first; the right column builds on this one.
Then you can learn
- Nothing in the atlas builds on it yet
Related:GCG Universal Jailbreak AttackAutomated Red Teaming
Left: prerequisites · right: builds on this · dashed: related. Full map →
03Learning resources
Official docs, explainers, courses and reference code. All links checked.
Official docs
Explainers
04Where the course teaches it
On learning paths:AI Application Engineer
For AI / data
- Verification: verified, checked 2026-10-21.
- JSON: /api/learn/algorithms.json (id:
prompt-injection)