paperApril 2026Unreviewed

SafeHarness: Lifecycle-Integrated Security Architecture for LLM-based Agent Deployment

Xixun Lin, Yang Liu, Yancheng Chen, Yongxuan Wu, Yucheng Ning, Yilong Liu, Nan Sun, Shunhong Zhang, Bin Chong, Chuan Zhou, Yanan Cao, Li Guo

arXiv.org

View resource

Abstract

The performance of large language model (LLM) agents depends critically on the execution harness, the system layer that orchestrates tool use, context management, and state persistence. Yet this same architectural centrality makes the harness a high-value attack surface: a single compromise at the harness level can cascade through the entire execution pipeline. We observe that existing security approaches suffer from structural mismatch, leaving them blind to harness-internal state and unable to

Framework mappings

OWASP Top 10 for Agentic Applications

ASI02Tool Misuse & Exploitation

MITRE ATLAS

AML.T0053AI Agent Tool Invocation

Suggested from the entry's categories.

Cite

@misc{lin2026safeharness,
  title = {{SafeHarness: Lifecycle-Integrated Security Architecture for LLM-based Agent Deployment}},
  author = {Xixun Lin and Yang Liu and Yancheng Chen and Yongxuan Wu and Yucheng Ning and Yilong Liu and Nan Sun and Shunhong Zhang and Bin Chong and Chuan Zhou and Yanan Cao and Li Guo},
  year = {2026},
  month = apr,
  eprint = {2604.13630},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2604.13630},
  url = {https://www.semanticscholar.org/paper/dc80f1113a0087e81f47cc746590b802c8ef6013}
}