Skynet: Workflow-Level Anomaly Detection for Agentic AI via Semantic and Structural Modeling

Abstract

Agentic AI systems execute complex tasks through long-horizon workflows of planning, tool use, and multi-agent coordination. Task failures in these systems often originate from a single step, such as an injected prompt or a flawed plan, and are then amplified through downstream dependencies as the corrupted step propagates across many subsequent agents and tool calls. Existing defenses either target a specific class of attacks or failures, or inspect individual prompts and steps in isolation. Bo

Categories

Framework mappings

Suggested from the entry's categories.

@misc{zhang2026skynet,
  title = {{Skynet: Workflow-Level Anomaly Detection for Agentic AI via Semantic and Structural Modeling}},
  author = {Chaoyu Zhang and Hexuan Yu and Heng Jin and Shanghao Shi and Ning Zhang and Yi Shi and Yulia R. Gel and Y. Thomas Hou and Wenjing Lou},
  year = {2026},
  month = sep,
  eprint = {2609.06835},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.06835}
}