Andrej Karpathy Skills
Security checks across malware telemetry and agentic risk
Overview
This is a disclosed coding-guidelines skill that tells agents to work more cautiously and does not contain hidden execution, credential access, or data exfiltration behavior.
Safe to install as a coding-behavior guideline. Before using the optional npm or clone installer, review which agent file it will modify and be cautious with --force or --global because those options can overwrite or persist agent instructions beyond one project.
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.