List your company

Tool-invocation hijacking to RCE on Agent Security Scenarios Old LLM Lineup

100RCE Success RateRCE-2
-4235077Sep 6, 2025

Evaluation results

MethodLinks
RCE-2Agent=Cline, Agent Version=Old2025.09100
RCE-2Agent=Cline, Agent Version=New2025.09100
RCE-2Agent=Cursor, Agent Version=Old2025.09100
RCE-2Agent=Copilot, Agent Version=Old2025.09100
RCE-1Agent=WindSurf, Agent Version=Old2025.0990
RCE-2Agent=Trae, Agent Version=Old2025.0980
RCE-1Agent=WindSurf, Agent Version=New2025.0970
RCE-2Agent=WindSurf, Agent Version=Old2025.0960
RCE-2Agent=Claude Code, Agent Version=Old2025.0960
RCE-1Agent=Cline, Agent Version=Old2025.0960
RCE-1Agent=Cline, Agent Version=New2025.0960
RCE-2Agent=WindSurf, Agent Version=New2025.0950
RCE-1Agent=Copilot, Agent Version=Old2025.0930
RCE-1Agent=Cursor, Agent Version=Old2025.0920
RCE-1Agent=Trae, Agent Version=Old2025.0920
RCE-2Agent=Cursor, Agent Version=New2025.0910
RCE-1Agent=Cursor, Agent Version=New2025.0910
AgentDojoAgent=WindSurf, Agent Version=New2025.0910
AgentDojoAgent=Cline, Agent Version=New2025.090
AgentDojoAgent=Cline, Agent Version=Old2025.090