The findings reveal a confidentiality risk in LLM agents: protecting explicit artifacts alone is insufficient, as observable execution behavior can leak the procedural knowledge required to reconstruct proprietary task-solving capabilities in low-capability and attacker-controlled agents.
Abstract
Large language model (LLM) agents increasingly achieve long-horizon tasks by combining foundation models with explicit skills and implicit procedural knowledge acquired through execution. The resulting task-solving capabilities have become valuable proprietary assets, raising a new security question: can a substantially weaker attacker-controlled agent acquire the capabilities of a stronger proprietary agent through limited black-box interaction? Existing skill-stealing attacks recover explicit skill artifacts, yet we show that artifact leakage does not necessarily transfer capability: a weaker agent may possess the same skills but still fail because it lacks procedural behaviors implicitly realized by the stronger agent. Our key insight is that the skill execution gap itself forms a leakage surface, where missing behaviors are exposed through observable differences between successful victim executions and failed attacker executions. Based on this, we present AgentLeak, a black-box capability-cloning attack that identifies capability-critical behaviors from these execution differences and incorporates them into attacker-side skills, while keeping the attacker's model, harness, and tools unchanged. Across 20 task scenarios comprising 600 instances, diverse agent systems, and multiple backbone models, AgentLeak improves task pass rates by over 40% compared with direct skill reuse and recovers more than 80% of the victim--attacker capability gap. Our findings reveal a confidentiality risk in LLM agents: protecting explicit artifacts alone is insufficient, as observable execution behavior can leak the procedural knowledge required to reconstruct proprietary task-solving capabilities in low-capability and attacker-controlled agents.
Agent skills extend the capabilities of AI agents by bundling instructions, code, and resources into reusable packages. During skill execution, agents may produce or modify artifacts, such as files, configurations, and cached data, that are reused by other skills in later sessions. Since these artifacts can persist acr...
Dian-Shu Liao, Shi-Dong Pan, Guang-Dong Bai et al.· 2 citations
It is found that directly composing benign skills can already induce malicious behaviors, even when every individual skill passes security vetting, and some target malicious behaviors remain difficult to realize through direct composition, even when the selected skills collectively provide the required capabilities.
Zun-Long Zhou, Zi-Yuan Yang, Meng-Yu Sun et al.· 0 citations
Verifying the provenance of autonomous LLM agents is a critical security challenge. Existing semantic attribution is easily obfuscated by system prompts, while active watermarking conflicts with schema-constrained tool generation. We propose AgentTrace, an active framework that identifies the foundation model behind a...
K KnotScryer is developed to detect insecure skill compositions by treating skills as typed capability flows, and is designed to design a search policy that tracks protected or untrusted data along handoff edges to locate paths reaching insecure operations.
Yi-Hao Li, Chuan-Pu Fu, Yi Zhao et al.· 0 citations
This work presents AgentXploit, a two-role auditing system that separates repository-level attack-path discovery from runtime exploitation and introduces AgentXploit-Bench, containing 72 reproducible vulnerabilities across 12 open-source AI-agent systems and frameworks.
Wei-Da Liang, Shi Qiu, Zhun Wang et al.· arXiv.org· 1 citation
This paper introduces skill cascading attacks, a threat paradigm in which a malicious objective is distributed across multiple skills so that each modification looks benign in isolation, yet their combined execution is harmful.
Zi-Hao Zhu, Siwei Lyu, A. Bibi et al.· arXiv.org· 0 citations