InjecAgent

档案区分来源方法阅读与既有目录信息；已经公开不代表完成独立证据复核或双语编辑复核。

测量构念

工具智能体对外部内容中的恶意指令有多脆弱？

任务与样本

1,054个模拟案例覆盖17种用户工具、62种攻击工具，评价30种智能体，并另报有效输出率。

判分与失效情形

在有效受评输出中计攻击成功，有效率另报。

系统条件

1,054模拟例，基础及一个固定增强提示。

统计含义

来源报告的结果表述；这不等同于已确定运行次数、分母或估计方法： ReAct GPT-4在受评案例24%中受影响，增强条件近乎翻倍。 不可解析等无效输出不进入稳健性评价（§3.1）。

效度与解释

增强设置只用一个固定提示，单轮模拟不能覆盖自适应多步行为。

适用研究问题

模型建议：保留有效率，并区分基础与增强攻击。

• InjecAgent : Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents · AbstractSelected original resource https://arxiv.org/html/2403.02691; sha256:0c09657b3b3f5886ee984f733cec481166307362cdde82378536fc24b254db78; archived source capture; exact capture time not independently established
• InjecAgent : Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents · §2 Benchmark ConstructionSelected original resource https://arxiv.org/html/2403.02691; sha256:0c09657b3b3f5886ee984f733cec481166307362cdde82378536fc24b254db78; archived source capture; exact capture time not independently established
• InjecAgent : Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents · §3.1 Experiment Settings: Output ParsingSelected original resource https://arxiv.org/html/2403.02691; sha256:0c09657b3b3f5886ee984f733cec481166307362cdde82378536fc24b254db78; archived source capture; exact capture time not independently established
• InjecAgent : Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents · §8 LimitationsSelected original resource https://arxiv.org/html/2403.02691; sha256:0c09657b3b3f5886ee984f733cec481166307362cdde82378536fc24b254db78; archived source capture; exact capture time not independently established