珠宝微距闪耀细节视频｜AI-HIVE

• Outcome: Pass
• Latest audit: Aug 25, 2026 · 8:11 AM UTC
• Version: 1.0.0

Overview

This is a disclosed AI-HIVE jewelry-video workflow that uses an API key and uploads user-selected media, with no hidden destructive or exfiltrating behavior found.

Install only if you are comfortable using an AI-HIVE API key, potentially storing it locally, and uploading selected media to AI-HIVE for generation. Confirm parameters before billable tasks, avoid sensitive or unlicensed assets, and override the default prompt if you do not want Chinese output.

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• 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
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (9)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 94% confidence
• Finding: The skill exposes operational capabilities including environment access, file read/write, network calls, and shell usage without declaring permissions, which weakens reviewability and informed consent. In this context the documented workflow includes local scripts, API access, and ffmpeg execution, so undeclared capabilities could lead to unexpected data access, local file modification, or external transmission of user assets and secrets.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 97% confidence
• Finding: The skill is presented as a narrowly scoped jewelry video workflow, but the described behavior includes broader generic model browsing, chat, media upload, account and wallet queries, local ffmpeg manipulation, and API-key handling. This mismatch is dangerous because users and reviewers may grant trust or provide assets under a limited-use assumption while the skill can perform materially broader actions, increasing the risk of data exposure, unintended charges, and misuse of local or remote capabilities.

Description-Behavior Mismatch

Medium

• Confidence: 89% confidence
• Finding: The file is presented as a specialized jewelry micro-detail video skill, but it embeds a broader general-purpose AI client with chat, image generation, model listing, upload, and account inspection features. This violates least privilege for skills and increases the reachable capability surface, making misuse easier if the skill is invoked in a constrained environment that assumes only jewelry-video behavior.

Intent-Code Divergence

Medium

• Confidence: 83% confidence
• Finding: The module docstring explicitly describes a universal AI capability tool rather than a narrowly scoped jewelry-video skill. That mismatch can conceal the true breadth of the code from reviewers and operators, increasing the chance that overly broad capabilities are approved and exposed unintentionally.

Context-Inappropriate Capability

Low

• Confidence: 72% confidence
• Finding: The skill exposes user-info functionality that can reveal account metadata and wallet or billing-related information unrelated to jewelry video generation. In a skill context, unnecessary account inspection expands access to sensitive operational data without a legitimate need tied to the advertised purpose.

Context-Inappropriate Capability

Medium

• Confidence: 87% confidence
• Finding: Generic text chat is unrelated to the declared jewelry-detail video workflow and enables arbitrary prompting against the provider's text models. In a skill ecosystem, this broadens the skill into a multi-purpose agent primitive that can be repurposed beyond its reviewed scope.

Context-Inappropriate Capability

Medium

• Confidence: 86% confidence
• Finding: The generic image-generation pathway extends beyond the stated purpose of producing jewelry detail videos and allows users to generate unrelated images via arbitrary prompts and parameters. This unnecessary capability increases abuse potential and breaks the expectation that the skill is narrowly specialized.

Context-Inappropriate Capability

Low

• Confidence: 69% confidence
• Finding: Model enumeration lets callers discover the provider's broader available model inventory, which exceeds the skill's stated specialized function. While not severe by itself, it expands reconnaissance and makes it easier to pivot into unsupported capabilities using the embedded generic client.

Natural-Language Policy Violations

Medium

• Confidence: 87% confidence
• Finding: The default prompt hardcodes Chinese output ('production-ready Chinese workflow') regardless of the user's language preference. This is not a direct security exploit, but it can override user intent, reduce transparency, and cause downstream misunderstandings in operational or commercial contexts where language accuracy matters.

Static analysis

No suspicious patterns detected.