Minimax H3 视频生成与编辑

• Outcome: Pass
• Latest audit: Aug 16, 2026 · 2:52 PM UTC
• Version: 1.0.0

Overview

This skill is a disclosed AI Hive video-generation helper that stores an API key locally and uploads user-selected media, with no hidden destructive or exfiltration behavior found.

Before installing, understand that this skill sends your prompts and any selected media to AI Hive, may incur API charges, stores an API key in a local config file, and downloads generated media to your chosen output folder.

SkillSpector

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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 (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 89% confidence
• Finding: The skill documents capabilities that read environment/config data, write files, invoke shell commands, and access the network, yet it declares no permissions. That creates a transparency and consent gap: users and the platform may authorize a seemingly narrow video skill while it can also open browsers, persist secrets, and make arbitrary outbound requests.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 93% confidence
• Finding: The documented purpose is a focused Minimax H3 video workflow, but the finding indicates broader behaviors such as chat, image generation, user-info/wallet access, generic model enumeration, and interactive API-key initialization with browser opening. This mismatch is dangerous because it defeats least surprise: users may provide credentials or invoke the skill expecting a constrained video tool, while it functions as a more general API client with access to additional sensitive account and billing context.

Description-Behavior Mismatch

Medium

• Confidence: 88% confidence
• Finding: The file exposes a broad AI Hive client with chat, image generation, model listing, uploads, and task polling even though the skill is marketed as Minimax H3 video generation/editing. This scope expansion violates least privilege and increases the attack surface, enabling users or downstream tooling to invoke unrelated capabilities and access more data/functions than the skill description suggests.

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The skill includes a task path that exposes user account information and wallet balance despite being scoped to video generation/editing. Revealing account metadata is unnecessary for the stated purpose and can leak sensitive operational or billing details to users or automations that should only perform media generation.

Intent-Code Divergence

Medium

• Confidence: 84% confidence
• Finding: The module docstring explicitly presents the file as a general-purpose AI capability tool, contradicting the narrower skill description. That mismatch can mislead reviewers and orchestration systems about the true capability set, making risky overbroad exposure easier to miss.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.