ClawHub
Back to skill

Security audit

GPT Image 2 营销图片

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent AI Hive image-generation helper for marketing assets, with expected API key setup, user-selected media upload, task polling, and result downloads.

Before installing, understand that this skill sends prompts and any files you explicitly pass with --image or --file to AI Hive, stores an AI Hive API key locally if you run init, and downloads generated outputs to ~/Downloads/AiHive unless you choose another output directory. Only pass media you are comfortable uploading to that service.

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 (2)

Lp3

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding
The skill advertises image-generation workflows in documentation but exposes effective capabilities associated with shell execution, filesystem access, environment access, and network use without declaring permissions. This increases the attack surface because a user or downstream system may trust the skill as low-risk while it can read local data, write files, and make outbound requests during execution.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The documented purpose is narrowly scoped to GPT Image 2 marketing asset generation, but the underlying behavior reportedly includes text chat, video generation, account and wallet lookups, model enumeration, generic uploads, browser-based API key setup, and local config management. That mismatch is dangerous because it can mislead users and reviewers about what the skill actually does, enabling unnecessary access to credentials, account metadata, local state, and broader remote actions than expected.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.