The paper introduces SARA, a framework that separates action induction from runtime authorization in tool‑augmented LLM agents. By treating these as distinct roles, SARA uses an Action Probe to record action provenance and only authorizes tool calls that align with the user objective and past successful executions. Experiments on AgentDojo and AgentDyn show that SARA reduces action‑to‑side‑effect risk to below 0.63% while preserving task performance.
By Xiaokun Guo, Zhen Xu, Dongdong Huo, Yanqiu Zhang, Wei Wang, Qinfu Yang, Dongjin Yu, Yu Wang
arXiv:2609.37196v1 Announce Type: cross
Abstract: Tool-using LLM agents remain vulnerable to indirect prompt injection because trusted instructions and untrusted observations share one context, allow...
By Yanjie Li, Xiangyu He, Xuelong Dai, Bin Xiao
The paper introduces PACE, a Provenance-Aware Capability Enforcement system designed to secure tool-using large language model agents by mediating every tool call before execution. PACE employs path confinement to limit influence paths and verifies effects against authenticated authority, distinguishing certified execution contracts from evaluated configurations. Experiments on eight agent‑security benchmarks show that the evaluated configuration reduces attack success in most cases while maintaining near‑native utility.
By Fengpeng Li, Qizhou Wang, Yuke Hu, Kemou Li, Jun Liu, Haiwei Wu, Jiantao Zhou, Di Wang
arXiv:2608. 10509v1 Announce Type: new Abstract: Shared memory helps language-model agents reuse information across long workflows, yet relevant evidence may not be admissible for a particular agent or action.
By Yiqi Wang, Zihao Yan, Jiaqi Zhang, Zhangkai Wu, Mingkai Zheng, Zequn Sun, Yanming Zhu, Taotao Cai
arXiv:2609.14987v1 Announce Type: cross
Abstract: Large language model (LLM) agents interact with external environments through tool invocation, but tool outputs can also expose them to indirect prom...
By Bingzheng Wang, Xiaoyan Gu, Wentao Wang, Xingyou Yang, Hongcheng Li, Rong Yin
arXiv:2607. 01236v1 Announce Type: cross Abstract: As LLM agents gain increasing access to powerful tools, ensuring that their actions are aligned with the user's intent becomes critical.
By Yining She, Yiliang Liang, Eunsuk Kang
arXiv:2606. 04990v1 Announce Type: cross Abstract: Large language model (LLM)-based agents increasingly solve complex tasks by interacting with external tools, retrieval systems, memory modules, environments, and other agents.
By Yiqi Wang, Jiaqi Zhang, Taotao Cai, Zirui Liu, Qingqiang Sun, Zequn Sun, Zhangkai Wu, Mingkai Zhang, Yanming Zhu
Shared memory helps language-model agents reuse information across long workflows, yet relevant evidence may not be admissible for a particular agent or action. Because restrictions propagate through derivations, summaries can conceal private, poisoned, untrusted, or revoked sources, enabling unauthorized reads or unsafe actions.
arXiv:2607. 20827v1 Announce Type: new Abstract: LLM agents choose tools and arguments from context that mixes user requests, tool outputs, retrieved records, memory, and untrusted text.
By Junchi Liao
ActGov is a runtime enforcement framework that validates each action proposed by a large language model (LLM) agent before it interacts with external tools, ensuring that actions stay within task‑scoped authorization boundaries and comply with dynamically constructed policies. It builds policies from tool specifications, benign tasks, and failure traces, verifying updates via SMT‑based counterexample checking. In evaluations on AgentDojo and AgentDyn benchmarks, ActGov consistently reduces indirect prompt‑injection attack success while maintaining task utility, outperforming existing defenses.
By Kaiyuan Zhang, Yuke Peng, Ke Jiang, Yinqian Zhang
The paper "Chaining Skills to Hijack LLM Agents" introduces APEX, a method that constructs adversarial skill chains to manipulate large language model agents into performing attacker-selected actions. By exploiting the sequential invocation of open-source skills, APEX creates a record of genuine task progress that falsely signals user approval, enabling downstream skills to carry out malicious actions. Experiments on SkillsBench and GPT‑5.4 show high success rates, and a prompting defense reduces but does not eliminate the threat.
By Tian Dong, Zixuan Ma, Haodong Zhao, Huaien Zhang, Shaofeng Li, Hao Chen
arXiv:2608. 12789v1 Announce Type: cross Abstract: Tool-using agents consume external data from sources with different levels of trust, yet tool responses rarely identify who produced each component or what it should convey.
By Sanjay Kariyappa, Severin Klingler, G. Edward Suh