paperJune 2026Unreviewed

Understanding and Evaluating Claw-like Agent Security Through a Computer-Systems Lens

Peizhi Niu, Wenjie Qu, Shangding Gu, Tianneng Shi, Yuankai Li, Ahmad Tawaha, Hend Alzahrani, Vincent Siu, Boyi Li, Chenguang Wang, Jiaheng Zhang, Basel Alomair, Ming Jin, Muhao Chen, Chi Wang, Costas Spanos, Dawn Song

View resource PDF

Abstract

Claw-like AI agents (e.g., OpenClaw) are always-on processes with persistent access to credentials, files, tools, and external services. They take on system-level responsibilities -- installing packages, maintaining state, scheduling subtasks, and mediating I/O -- making security failures far more severe than in other agents. Yet existing benchmarks focus on model responses and tool calls, leaving cross-component failure modes largely unmeasured. We adopt a computer-system analogy: treating a Cl

Cite

@misc{niu2026understanding,
  title = {{Understanding and Evaluating Claw-like Agent Security Through a Computer-Systems Lens}},
  author = {Peizhi Niu and Wenjie Qu and Shangding Gu and Tianneng Shi and Yuankai Li and Ahmad Tawaha and Hend Alzahrani and Vincent Siu and Boyi Li and Chenguang Wang and Jiaheng Zhang and Basel Alomair and Ming Jin and Muhao Chen and Chi Wang and Costas Spanos and Dawn Song},
  year = {2026},
  month = jun,
  eprint = {2606.30755},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.30755}
}