Agent Workflow Playbook — AI Multi-Agent Orchestration Framework
Security checks across malware telemetry and agentic risk
Overview
This is a documentation-only playbook for AI agent workflow design, with no executable code, credential use, persistence, or hidden data handling found.
This appears safe to install as a reference guide. Treat it as design and marketing guidance rather than executable automation, and independently review any workflow patterns before applying them to systems that access sensitive data or perform real-world actions.
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
51/51 vendors flagged this skill as clean.