Security audit
Agent Team Orchestration 1
Security checks across malware telemetry and agentic risk
Overview
This is a documentation-only guide for coordinating multiple AI agents, with no executable code or hidden data handling found.
Before installing, verify the publisher because of the metadata mismatch. When using the skill, keep secrets out of shared folders, restrict which agents can edit persistent instructions, cap parallel agents and schedules, and require human approval for high-impact work.
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.