List your company

System Prompt Exfiltration on WindSurf Agent Environment

100Pseudo-RecallToolLeak
39.6855.347186.66Sep 6, 2025

Evaluation results

MethodLinks
ToolLeakBackend LLM=sonnet-42025.09100
ToolLeakBackend LLM=sonnet-4.52025.09100
ToolLeakBackend LLM=SWE-1.52025.0991
ToolLeakBackend LLM=gemini-2.5-pro2025.0972
IgnoreBackend LLM=gemini-2.5-pro2025.0963
NaiveBackend LLM=SWE-1.52025.0942
Remember-the-StartBackend LLM=SWE-1.52025.0942
Ignore-RememberBackend LLM=SWE-1.52025.0942
Completion-RememberBackend LLM=SWE-1.52025.0942
Ignore-Completion-RememberBackend LLM=SWE-1.52025.0942