Security audit
Agentic Identity Trust
Security checks across malware telemetry and agentic risk
Overview
This is a documentation-only skill for designing secure agent identity, authorization, and audit systems, with no executable code or hidden access.
Reasonable to install as security architecture guidance. Treat any code generated from it as security-sensitive: review implementations carefully, keep private keys and production credentials out of chat, and use appropriate secret-management or signing infrastructure for real deployments.
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.