Deep Research
Security checks across malware telemetry and agentic risk
Overview
This is a plain research-method skill with no executable code or hidden access.
Installers should treat this as a workflow template for research. Review any concrete actions it recommends during use, especially installing tools or writing to memory, but the artifact itself does not request hidden access or run code.
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
64/64 vendors flagged this skill as clean.