AI Training Specialist
Security checks across malware telemetry and agentic risk
Overview
This is a non-executable guidance skill for AI training and annotation work, with no evidence of hidden code, credential use, persistence, or unsafe system access.
Review the skill as public guidance and coaching material for AI training work. Do not share platform passwords, private qualification materials, or confidential client data unless you intentionally provide them for a specific task.
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
63/63 vendors flagged this skill as clean.