Breadcrumbing Search Agents

Abstract

LLM-based search agents are widely used for information-seeking tasks, but their reliance on external tool returns introduces a critical security risk: web content retrieved during execution is untrusted, exposing agents to prompt injection and goal hijacking. Prior work on search-agent safety primarily focuses on static web-content injection, but modern agents issue follow-up queries and cross-check competing sources, so a single injected page is often diluted or rejected. We show that the chan

Categories

Framework mappings

Suggested from the entry's categories.

@misc{li2026breadcrumbing,
  title = {{Breadcrumbing Search Agents}},
  author = {Xuebin Li and Hanqing Zhao and Siyuan Liang and Kejiang Chen and Weiming Zhang and Dacheng Tao and Nenghai Yu},
  year = {2026},
  month = aug,
  eprint = {2608.04565},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2608.04565}
}