The paper introduces AgentLeak, a black‑box attack that clones the task‑solving capabilities of a strong LLM agent onto a weaker one by exploiting differences between successful and failed executions. Unlike prior skill‑stealing methods that only recover explicit skill artifacts, AgentLeak identifies and incorporates missing procedural behaviors, boosting task pass rates by over 40% and closing more than 80% of the capability gap across 20 scenarios. The study demonstrates that observable execution behavior can leak proprietary procedural knowledge, posing a confidentiality risk for LLM agents.
By Xiaoting Lyu, Yuhong Wu, Yufei Han, Shichang Liu, Liang Zhang, Bin Wang, Bin Wang, Xiaobo Ma, Wei Wang
arXiv:2610.01564v1 Announce Type: cross
Abstract: LLM agents use skills to improve performance on specialized tasks. To complete a user request, an agent may invoke several skills in sequence, allowi...
By Tian Dong, Zixuan Ma, Haodong Zhao, Huaien Zhang, Shaofeng Li, Hao Chen
arXiv:2608. 15108v1 Announce Type: cross Abstract: Large language model agents are increasingly connected to high-value resources such as computing infrastructure, credentials, usage budgets, identities, private knowledge, communication channels, and organizational workflows.
By Puyu Zeng, Qibing Ren
arXiv:2603. 19423v2 Announce Type: replace-cross Abstract: Large language model (LLM) agents increasingly rely on external tools (file operations, API calls, database transactions) to autonomously complete complex multi-step tasks.
By Shawn Li, Yue Zhao
arXiv:2609.36570v1 Announce Type: cross
Abstract: Indirect prompt injection makes an LLM agent treat untrusted retrieved text as instructions. We present CounterSteer, an inference-time defense that...
By Mark Russinovich
The paper introduces a new skill poisoning technique for large language model agents that decouples the pretext (rationale) from the actuation (operation). By separating these two risk‑realization factors, the authors create coordinated pretext‑actuation skill pairs that allow malicious actions to remain hidden within legitimate agent behavior. An automated framework is presented to discover execution dependencies, synthesize these skill pairs, and refine them through closed‑loop feedback, achieving high attack success in both single‑session and persistent scenarios.
By Wenxin Wu, Lingyong Yan, Lei Sha, Shuaiqiang Wang, Jiashu Zhao
arXiv:2608.30207v1 Announce Type: cross
Abstract: Computer use agents (CUAs) are vision-language models that perceive a screen and act on a real operating system through mouse, keyboard, and terminal...
By Chen Xiong, Zhiyuan He, Pin-Yu Chen, Stjepan Picek, Tsung-Yi Ho
The paper "SkillBloat: Token Amplification Attacks via Skill Injection in LLM Coding Agents" investigates how agent skills—task‑specific instructions, scripts, and resources—can be exploited to create a trusted instruction channel that enables token amplification attacks. It introduces a two‑phase framework, SkillBloat, which first screens a library of attack‑type conditions across multiple amplification mechanisms and then refines the strongest candidate through LLM‑guided full‑document skill rewriting. Evaluated on a real‑world skill benchmark, SkillBloat achieves an average best amplification of 5.4184×–10.1455× across multiple coding‑agent target configurations, and an ablation study shows that the second‑stage refinement consistently improves performance over the initial screening alone.
By Yuanjin Zheng, Jingbang Chen
ClawSentry is an open‑source, framework‑agnostic security supervision gateway designed to protect autonomous large language model (LLM) agents from progressive risks that can arise at four points in the agent control loop: skill admission, invocation‑time intent, execution‑time effect, and post‑action consequence. It introduces a multi‑tier decision engine—deterministic L1, rule‑anchored L2, and read‑only L3—alongside a First‑Use Skill Package Review (FSPR) and an Agent Harness Protocol (AHP) that applies a single policy across multiple agent runtimes without modifying their internals. Evaluation on SkillInject and the SkillsSafety benchmark shows that ClawSentry significantly reduces contextual adversarial skill risk (ASR) while maintaining high task success rates (TSR).
By Kai Wang, Zeming Wei, BiaoJie Zeng, Chang Jin, An Wang, Xiaokun Luan, Zhixiao Lin, Jingjing Qu, Xia Hu, Xingcheng Xu
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
Computer use agents (CUAs) are vision-language models that perceive a screen and act on a real operating system through mouse, keyboard, and terminal, and they are increasingly deployed to automate ev...
Guardrail models, which screen malicious prompts in LLM services, often use lightweight Transformers with short context windows and bucketed positional encodings. The study identifies a new failure mode called Overflip, where repeating a prompt causes the guardrail’s prediction to flip from malicious to benign as the sequence length increases. Experiments on nine popular guardrails show that 5 models exhibit MAL→BEN flips on 100 prompts, with flip rates ranging from 8% to 92% and first flips occurring between 2.6k and 9.4k tokens, highlighting a gradual attention dispersion distinct from traditional attention‑dilution attacks.
By Xu He, Chih-Hsuan Lin, Hung-Mao Chen, Junjie Xiong, Yan Zhai, Kun Sun