agentic-eval
Security checks across malware telemetry and agentic risk
Overview
This is a markdown-only reference skill about evaluating and improving AI agent outputs, with no executable installer, credential access, or hidden behavior found.
Reasonable to install as a reference skill. When applying its patterns, cap iteration loops, sandbox any generated code or tests, and avoid retaining sensitive prompts or outputs in evaluation logs.
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.