ClawHub
Back to skill

Security audit

AI珠宝首饰试戴效果图｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE jewelry try-on workflow that uses user-provided images and an API key to generate previews, with no hidden destructive or exfiltration behavior found.

Before installing, confirm you are comfortable giving this skill access to AI-HIVE, user-selected reference images, and an AI-HIVE API key. Use the confirmation step before any generation, review pricing/routing, and avoid using unlicensed product, person, logo, or brand assets.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding
The skill embeds runnable commands and documents capabilities that use environment variables, local files, network access, and shell execution, but it declares no permissions. This creates a transparency and policy-enforcement gap: a host may invoke or approve the skill without realizing it can read/write files, access secrets such as API keys, and make outbound requests that may incur cost or exfiltrate data.

Tp4

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding
The stated purpose is a narrowly scoped jewelry try-on workflow, but the documented behavior extends into generic text chat, video generation, account and wallet queries, broad model enumeration, and unrelated blueprint generation. This scope mismatch is dangerous because users and orchestration systems may trust or auto-invoke the skill for a limited task while it can access broader APIs and account metadata than expected, increasing the chance of data exposure, unintended charges, or misuse of privileged capabilities.

Vague Triggers

Medium
Confidence
72% confidence
Finding
The trigger phrases include broad everyday terms such as AI model, jewelry product image, and virtual shooting, which can cause the skill to be invoked in situations outside the user's intent. Overbroad invocation is a security concern because this skill can lead to network actions and potentially billable operations, so accidental activation can expose data or trigger external requests without clear necessity.

Static analysis

No suspicious patterns detected.