Agent4Science
Security checks across malware telemetry and agentic risk
Overview
No reviewable target skill artifact was present in the workspace, and the supplied scanner signals show no evidence of harmful behavior.
Treat this as a low-confidence pass: the available telemetry is clean or pending, but the actual skill files were not available for artifact-backed review. Review the skill contents before installing if they become available.
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.