Security audit
Model Behavior Layer (Ares MBL)
Security checks for vulnerabilities and agentic risk
Overview
This is a markdown-only prompt guide for changing agent behavior, with no executable install steps or hidden data access.
Review the prompt before adding it to a system prompt because it changes how an agent responds and encourages more autonomous low-risk action. The artifact does not install code or access data, but its metadata tags appear overbroad for what the markdown actually does.
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
- Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Static analysis
No suspicious patterns detected.