Clawdhub

Security checks across malware telemetry and agentic risk

Overview

This skill is for managing ClawHub skills, but it gives an agent examples for broad install/update/publish actions without clear safety checks.

Install only if you intend to let the agent manage local skill folders and potentially publish skills. Require explicit approval before running install, update --all, --force, --no-input, login, or publish commands, and verify the package, target directory, registry, and files being published first.

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: 89% confidence
Finding: The skill documentation explicitly instructs the agent to run install, update, and publish commands that can change local files, upgrade dependencies, or modify remote registry state, but it provides no safety constraints, confirmation requirements, or warnings about those side effects. In an agent context, this increases the risk of unintended package installation, destructive updates, or accidental publication of local content, especially when combined with flags like --all, --force, and --no-input.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal