CARE: Pre-Execution Command Verification for Shell-Executing LLM Agents

Wenxiao Zhang, Yu Liu, Zhiwei Yang +7

Large Language Model (LLM) agents are increasingly used for coding and terminal automation, making shell-command dispatch a high-stakes runtime control point. We study command-leve…