reinvent4
Security checks across malware telemetry and agentic risk
Overview
The submitted skill artifacts are coherent internal engineering and moderation guidance, with sensitive actions disclosed and gated by user intent or confirmation.
Install only in a trusted ClawHub maintainer or Convex development workspace. Review commands before running workflows that use staff moderation, publish proof artifacts, run full-access review helpers, install packages, or touch authenticated deployments.
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.