ClawHub
Back to skill

Security audit

Clawhub Guard

Security checks across malware telemetry and agentic risk

Overview

This skill is not clearly malicious, but it is a safety tool that can install other skills while overstating the strength of its checks.

Treat this as an experimental marketplace helper, not a trusted security gate. Use scan/search only for rough discovery, manually inspect any skill before installing, and be cautious with the install command because it can persistently change your local skill environment.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill advertises shell-driven functionality through CLI commands and workflow descriptions, but the metadata shown does not declare corresponding permissions. Undeclared shell capability is dangerous because it obscures the true execution surface from users and reviewers, especially for a skill that can browse, audit, and install other skills via a local CLI.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding
The documented purpose frames the skill as a safety-review assistant, but the behavior includes local environment inspection and automatic installation of skills. That mismatch is risky because users may trust it as a passive reviewer while it can perform privileged, state-changing actions on the local ClawHub environment, increasing the chance of unintended installs or abuse if the vetting logic is weak or bypassed.

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The skill presents itself as a marketplace safety guard that browses, searches, and vets skills, but it also performs actual installation of a selected skill. That is dangerous because users may trust the safety framing and allow execution of untrusted third-party code after only a superficial keyword-based check, creating a supply-chain risk.

Intent-Code Divergence

Medium
Confidence
89% confidence
Finding
The audit command claims to check installed skills for updates, but it actually runs inspect on each skill instead of performing an update check. This mismatch can mislead users into believing they are assessing patch status when they are not, which may leave vulnerable or outdated skills installed.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.