Security audit
Agent Orchestration Multi Agent Optimize.Skip
Security checks across malware telemetry and agentic risk
Overview
This is an advisory multi-agent optimization guide with no executable payload, hidden behavior, credential request, or destructive instruction.
This skill appears safe to install as a guidance document. Before applying its recommendations in production, define measurable goals, keep rollback plans, limit any caching or memoization of sensitive context, and only profile systems you are authorized to inspect.
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
66/66 vendors flagged this skill as clean.