Security audit

AI大模型专家｜GPT Image 2 图片生成与编辑

Security checks across malware telemetry and agentic risk

Overview

This skill coherently provides AI-HIVE image generation/editing, with disclosed API-key setup and media upload, and no evidence of hidden exfiltration or destructive behavior.

Install only if you are comfortable storing an AI-HIVE API key locally and sending prompts plus selected media files to AI-HIVE for image generation. Review costs for batch or high-volume work, and avoid providing private or unlicensed media unless you intend it to be uploaded.

SkillSpector

By NVIDIA

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 (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 89% confidence
Finding: The skill documentation describes code paths that read environment variables, write local credential files, access the network, and invoke shell commands, yet it declares no explicit permissions. This creates a trust and review gap: users or platforms may authorize the skill under a narrower mental model than its actual capabilities, increasing the risk of credential exposure, unintended file modification, or uncontrolled external requests.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 94% confidence
Finding: The skill is presented as a narrowly scoped GPT Image 2 image-generation/editing tool, but the detected behavior includes broader model access, video generation, text chat invocation, account and wallet queries, generic media upload, and local credential initialization. That mismatch is dangerous because it can cause users to provide consent, secrets, or media under false assumptions about scope, enabling unexpected data handling, billing actions, or access to unrelated APIs.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The file is packaged as a GPT Image 2 image-generation/editing skill, but it embeds a much broader general-purpose AI-HIVE client with text chat, video generation, model listing, user-info retrieval, and generic media upload. This scope mismatch violates least privilege and increases the attack surface: anyone invoking or repurposing the script can access capabilities unrelated to the declared skill, which is especially risky in an agent-skill context where reviewers and users may trust the advertised narrow function.

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The module docstring explicitly advertises the tool as a general AI capability caller, which contradicts the declared GPT Image 2 image-only purpose of the skill. In security terms, this is a capability-deception/scope-declaration issue: operators may approve or expose the skill assuming narrow image functionality while the implementation is intentionally broader, enabling misuse and weakening policy controls.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The skill enables implicit invocation via `allow_implicit_invocation: true` without any visible activation constraints, scoping rules, or user-consent guardrails. Because this skill can trigger image-generation and editing workflows through an external service, implicit activation increases the chance of unintended third-party API use, unexpected data transfer, or cost-incurring actions from loosely related user requests.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: SKILL.md:74