Tool-invocation hijacking to RCE on Agent Security Scenarios Old LLM Lineup
100RCE Success RateRCE-2
Evaluation Results
| Method | Links | |
|---|---|---|
| RCE-2Agent=Cline, Agent Version=Old2025.09 | 100 | |
| RCE-2Agent=Cline, Agent Version=New2025.09 | 100 | |
| RCE-2Agent=Cursor, Agent Version=Old2025.09 | 100 | |
| RCE-2Agent=Copilot, Agent Version=Old2025.09 | 100 | |
| RCE-1Agent=WindSurf, Agent Version=Old2025.09 | 90 | |
| RCE-2Agent=Trae, Agent Version=Old2025.09 | 80 | |
| RCE-1Agent=WindSurf, Agent Version=New2025.09 | 70 | |
| RCE-2Agent=WindSurf, Agent Version=Old2025.09 | 60 | |
| RCE-2Agent=Claude Code, Agent Version=Old2025.09 | 60 | |
| RCE-1Agent=Cline, Agent Version=Old2025.09 | 60 | |
| RCE-1Agent=Cline, Agent Version=New2025.09 | 60 | |
| RCE-2Agent=WindSurf, Agent Version=New2025.09 | 50 | |
| RCE-1Agent=Copilot, Agent Version=Old2025.09 | 30 | |
| RCE-1Agent=Cursor, Agent Version=Old2025.09 | 20 | |
| RCE-1Agent=Trae, Agent Version=Old2025.09 | 20 | |
| RCE-2Agent=Cursor, Agent Version=New2025.09 | 10 | |
| RCE-1Agent=Cursor, Agent Version=New2025.09 | 10 | |
| AgentDojoAgent=WindSurf, Agent Version=New2025.09 | 10 | |
| AgentDojoAgent=Cline, Agent Version=New2025.09 | 0 | |
| AgentDojoAgent=Cline, Agent Version=Old2025.09 | 0 |