ClawHub
Back to skill

Security audit

Alibaba Cloud AI Entry Modelstudio

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed Alibaba Cloud Model Studio routing helper; it needs an API key and may make provider calls, but I found no hidden or destructive behavior.

Install this only if you intend to use Alibaba Cloud Model Studio. Keep the DashScope API key scoped and protected, review the destination sub-skill before approving high-impact or mutating actions, expect provider API usage and possible costs, and avoid sharing saved output files unless they have been reviewed for sensitive prompts, task IDs, result URLs, or account details.

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 (2)

Intent-Code Divergence

Medium
Confidence
92% confidence
Finding
The skill is explicitly described as a routing-only entry point, but its workflow instructs the agent to confirm whether an operation is mutating and then execute a target operation. That expands scope from classification/routing into action-taking, which can cause unintended remote changes if an agent follows the entry skill directly instead of delegating to a bounded sub-skill.

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
A routing skill should not contain direct API polling, output downloading, or operational playbooks, because those instructions can cause the agent to bypass the intended specialized skill boundary. This increases the chance of unauthorized network actions, inconsistent safety controls, and accidental execution against user resources from a component that should only dispatch requests.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.