文献结构逻辑要点提炼
Security checks across malware telemetry and agentic risk
Overview
This is a narrow academic-paper logic analysis skill with no executable code or hidden behavior found.
This skill appears safe to install for analyzing academic papers. Only provide confidential or unpublished papers if you are comfortable having your agent process that content.
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
65/65 vendors flagged this skill as clean.