← Back to Database
System Prompt Extraction via Tool Logging
🟠 HIGH
model_extraction
proven
AVE-2025-0066
· aka: Prompt Theft
Summary
Agent's system prompt is leaked through verbose tool error messages or debug logging.
Blast Radius
Full system prompt disclosure enables targeted attacks.
Prerequisites
Agent with tools that produce verbose errors.
Environment
- Frameworks: LangGraph, CrewAI
- Models tested: [Available in NAIL SDK]
- Multi-agent: No
- Tools required: Yes
- Memory required: No