Security audit
Agent Governance Assistant
Security checks across malware telemetry and agentic risk
Overview
This is a markdown-only enterprise AI governance guide with no executable code, hidden behavior, persistence, or credential handling.
Installers should treat this as a governance and compliance drafting aid, not legal advice. Avoid pasting credentials, raw customer records, confidential logs, or unnecessary enterprise security details unless the agent runtime is approved for that data.
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
VirusTotal findings are pending for this skill version.
Static analysis
No suspicious patterns detected.