ClawHub
Back to skill

Security audit

Cloud Infra Automation

Security checks across malware telemetry and agentic risk

Overview

This skill is not malicious, but it documents powerful production cloud actions without enough scoping or safeguards.

Review before installing or following the commands. Do not run ./cloud.sh apply prod or ./cloud.sh destroy prod unless you have verified the script source, selected the exact cloud account and environment, reviewed a plan, and have approval, backups, and rollback steps in place.

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
Findings (1)

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The documentation explicitly includes a destructive command (`./cloud.sh destroy prod`) without any warning, confirmation guidance, or explanation of the consequences. In a cloud automation skill, this is operationally dangerous because users may copy-paste commands and unintentionally delete production infrastructure, causing outages and data loss.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.