Skip to content
Search
paperSeptember 2026Unreviewed

What's in Your Agent's Context? Context Privilege Escalation Attacks against AI Agent Harness

Zichuan Li, Jian Cui, Ashley Chen, Xiaojing Liao, Luyi Xing

View resourcePDF

Abstract

Real-world, high-profile AI agent harnesses often rely on vendor-proprietary or opaque designs for context assembly, leaving the sources and underlying logic of assembled context poorly understood and the resulting security risks largely unexplored. In this paper, we present the first systematic analysis of context assembly designs in real-world AI agent harnesses. We study and uncover how an agent harness is designed to collect and assemble context from diverse sources, and identify a set of pr

Categories

Threat Modeling

Cite

@misc{li2026whats,
  title = {{What's in Your Agent's Context? Context Privilege Escalation Attacks against AI Agent Harness}},
  author = {Zichuan Li and Jian Cui and Ashley Chen and Xiaojing Liao and Luyi Xing},
  year = {2026},
  month = sep,
  eprint = {2609.01222},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.01222}
}