AI大模型专家｜电商主图

• Outcome: Pass
• Latest audit: Aug 24, 2026 · 2:38 AM UTC
• Version: 1.0.0

Overview

This is a disclosed AI-HIVE e-commerce image-generation skill, with expected API-key, upload, polling, and download behavior, but users should be careful because it can submit paid third-party jobs when invoked.

Install only if you intend to use AI-HIVE for commercial image generation. Confirm which local reference images will be uploaded, watch batch size and routing because jobs may cost money, and protect the AI-HIVE API key stored in ~/.ai-hive/config.json.

SkillSpector

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 86% confidence
• Finding: The skill advertises operational capabilities such as environment access, file read/write, network access, and shell execution, but does not declare any permissions or limits to the user. This reduces transparency and informed consent, and can hide sensitive operations like credential handling, local file writes, or outbound uploads behind a seemingly simple image-generation workflow.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 93% confidence
• Finding: The documented purpose is narrowly framed as e-commerce image generation, but the implementation reportedly includes broader features such as chat, video generation, account/balance access, model catalog enumeration, and generic upload functions. This mismatch is dangerous because users may grant trust, credentials, and local assets for one task while the skill can exercise materially broader access and exfiltration or spend-related behaviors.

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: The file presents itself as a narrowly scoped e-commerce main-image skill, but actually exposes broad generic AI-Hive functionality including chat, model enumeration, uploads, polling, and video generation. This scope mismatch can let users invoke capabilities not disclosed in the skill metadata, undermining least privilege and review assumptions.

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The skill includes a user-info endpoint that retrieves account and wallet information, but this capability is not reflected in the stated purpose of an e-commerce main-image generator. Hidden account-inspection features increase the chance of unnecessary exposure of sensitive billing or account metadata.

Description-Behavior Mismatch

Medium

• Confidence: 94% confidence
• Finding: Video-generation support materially exceeds the advertised image-focused scope of the skill. Undisclosed media-generation paths expand the operational and cost surface, and may allow unintended content production workflows through a skill users expect to be image-only.

Vague Triggers

Medium

• Confidence: 94% confidence
• Finding: The skill enables implicit invocation without any visible trigger narrowing, guard conditions, or exclusions. Because this skill can automatically upload reference images, submit jobs, poll status, and download results through an external service, broad auto-invocation increases the chance of unintended execution, unwanted data transfer, and surprise cost-incurring actions from loosely related user requests.

Static analysis

No suspicious patterns detected.