The evidence behind the guides

Sources

Find a study, check what it supports, and follow it back to a concern.

TopicCollection

Sources for Agent-loop hijack

2 sources

Guide citations are still under editorial review.

Chen and colleagues · 2024 Research study

AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases (original source, new tab)

Demonstrates poisoned retrieval and memory in evaluated agents without requiring model retraining.

Zhan and colleagues · 2024 Research benchmark

InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents (original source, new tab)

Evaluates tool-mediated instruction attacks; benchmark rates are not deployment-wide prevalence estimates.

Sources retain their original publishers’ rights. Inclusion does not imply endorsement. Saved sources stay in this browser on this device. Dates and source descriptions come from the existing editorial records.