ClawHub
Back to skill

Security audit

AI Expert Team

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed expert-team orchestration guide with one bounded bootstrap step to clone reference repositories for inspection.

Before installing, be aware that if you ask the skill to inspect or initialize its source assets, it may create an assets directory and clone two external GitHub repositories. That means network access and local disk changes; review or pin those repos first if your environment requires reproducible or offline-only behavior.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Description-Behavior Mismatch

Low
Confidence
84% confidence
Finding
The skill includes operational instructions to clone external GitHub repositories and inspect repository state. In an agent context, this expands the trust boundary to unpinned remote content and can trigger unintended network access or local workspace changes if the user asks to initialize assets, which is risky for a skill that is otherwise framed as analysis/orchestration.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.