TraceGrant: A Contract-Governed Security Framework for the Task-Effect Lifecycle of Networked LLM Agents

Abstract

Networked large language model (LLM) agents retrieve information from email, cloud storage, calendars, transaction platforms, and Web services to complete multistep tasks that produce persistent external effects. The same content needed for legitimate execution may also contain indirect prompt injections that redirect tool use, alter sensitive arguments, or disrupt task completion. Existing defenses mainly constrain untrusted content or individual tool calls, leaving user intent, runtime evidenc

Categories

Framework mappings

Suggested from the entry's categories.

@misc{liao2026tracegrant,
  title = {{TraceGrant: A Contract-Governed Security Framework for the Task-Effect Lifecycle of Networked LLM Agents}},
  author = {Bohao Liao and Jingchao Wang and Qipeng Song and Jin Cao and Jieling Wang and Boyu Deng},
  year = {2026},
  month = aug,
  eprint = {2608.21126},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2608.21126}
}