ClawHub
Back to skill

Security audit

商品精修

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed AI product-photo retouching wrapper that uploads user-selected images to AI Hive and stores an API key locally, with no evidence of hidden execution, exfiltration, or destructive behavior.

Install only if you are comfortable sending selected product images and prompts to AI Hive and storing an AI Hive API key locally. Review prompts before running generation so the tool does not alter legally or commercially important product details, labels, or real damage.

SkillSpector

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
94% confidence
Finding
The skill declares no permissions, yet the documented commands and behavior clearly imply shell execution, file access, environment variable use, network access, and local state/config writes. This is dangerous because users and policy systems cannot accurately assess or constrain what the skill can do, creating hidden attack surface and weakening least-privilege enforcement.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding
The skill is presented as a narrowly scoped product-retouch workflow, but the documented behavior includes broader image generation, standalone upload/task operations, and browser-based API key initialization and persistence. That mismatch is dangerous because it can mislead users and reviewers about the real operational scope, enabling unintended data upload, credential handling, and use beyond the claimed constrained editing context.

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The skill is branded and described as a narrowly scoped product-retouch tool, but the CLI exposes broader capabilities including generic task inspection and standalone media upload. That scope expansion weakens trust boundaries: users or upstream agents may assume only constrained retouch behavior while the code permits more general interaction with the backend API and reuse of uploaded media IDs.

Intent-Code Divergence

Medium
Confidence
89% confidence
Finding
The module docstring claims the tool only exposes fixed-model image generation with bounded reference-image upload, but the implementation forwards arbitrary model parameters and supports generic task querying and upload reuse. This mismatch can mislead reviewers and orchestrators into granting the tool broader trust than intended, creating a deceptive security boundary.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.