Security audit
Structured Multi-Agent Deliberation
Security checks across malware telemetry and agentic risk
Overview
This is a coherent deliberation workflow skill with optional local helper scripts, and I found no hidden network access, credential handling, destructive behavior, or background persistence.
Before installing, understand that using the optional Python scripts means running local code that creates or validates deliberation files in your chosen state and artifact directories. Review any LLM API integration you add to round-controller.py, since the provided version is only a stub and does not itself handle credentials or make network calls.
SkillSpector
By NVIDIA
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
VirusTotal
66/66 vendors flagged this skill as clean.
Static analysis
No suspicious patterns detected.