Security audit

AI ASMR商品视频｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed AI-HIVE product-video workflow whose API use, media upload, local video processing, downloads, and API-key storage fit its stated purpose.

Before installing, confirm you are comfortable giving AI-HIVE an API key and uploading only media you are authorized to use. Review generated commands and pricing parameters before running generation, especially because API calls may cost money and outputs may be downloaded locally.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (8)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 86% confidence
Finding: The skill exposes operational capabilities including environment access, filesystem interaction, shell usage, and network calls without declaring permissions or clearly constraining them. This creates a trust and review gap: consumers may believe the skill is planning-only while it can access local resources and invoke external services, increasing the chance of unintended data exposure, secret leakage, or unsafe command execution in an agent runtime.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 94% confidence
Finding: The documented purpose is a narrow ASMR product-video workflow, but the skill appears capable of broader actions such as generic chat/image generation, wallet or account querying, model enumeration, arbitrary media upload, and local ffmpeg processing. That mismatch is dangerous because reviewers and users may authorize the skill under a narrow mental model while it can perform materially broader data access, billing-related, and content-processing operations, enabling misuse or unexpected spend/data handling.

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The file is marketed as a narrowly scoped ASMR product-video skill, but it actually embeds a broad AI-Hive client that supports unrelated capabilities such as chat, image generation, model enumeration, user-info lookup, and generic media upload. This scope mismatch increases attack surface and violates least-privilege expectations: a caller or wrapper that trusts the manifest may unintentionally grant access to operations outside the advertised purpose.

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The skill can query user account information and wallet balance even though its stated function is generating ASMR product videos. Unnecessary access to account metadata expands sensitive data exposure and can enable profiling or misuse of billing-related information if an agent or user invokes the extra command unexpectedly.

Context-Inappropriate Capability

Medium

Confidence: 94% confidence
Finding: The text-chat endpoint is unrelated to the declared video-generation use case and creates a covert general-purpose model access path inside a specialized skill. That broadens what prompts and data can be sent through the skill, making policy bypass and unintended data transmission more likely in environments that trust the skill's narrow manifest.

Context-Inappropriate Capability

Medium

Confidence: 94% confidence
Finding: The image-generation endpoint exceeds the advertised ASMR product-video functionality and introduces extra model actions not disclosed by the skill metadata. In agentic settings, this can be abused to generate unrelated content, transmit prompts or reference media to additional endpoints, and bypass user or platform expectations about what the skill is allowed to do.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: Implicit invocation is enabled with no visible activation constraints, so the skill can be selected in broader contexts than intended. Because this skill can generate runnable AI-HIVE commands and production workflows, over-broad activation increases the chance of unintended execution paths, policy bypass around user intent, or use in sensitive commercial/content-generation scenarios without explicit user confirmation.

Natural-Language Policy Violations

Medium

Confidence: 83% confidence
Finding: The default prompt mandates Chinese output regardless of user preference, which can cause instruction confusion, reduce user visibility into generated commands, and increase the likelihood that a user approves or runs content they do not fully understand. In a skill that produces executable workflow steps and external API commands, forced language selection weakens informed consent and reviewability rather than directly creating code execution risk.

Static analysis

No suspicious patterns detected.