流体网络求解与 Agent Skill 封装
Security checks across malware telemetry and agentic risk
Overview
This is an instruction-only fluid-network analysis skill with no code execution, credentials, persistence, or hidden data access evident in the artifacts.
This appears safe to install from an agentic-security perspective. Treat it as a prototype engineering aid: independently validate any pressure, flow, reliability, or design conclusions before using them for real-world industrial decisions.
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.