Security audit

家居商品空间植入效果图｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed AI-HIVE furniture image workflow that uses an API key, uploads user-selected media, and may incur generation costs, with scope notes but no evidence of hidden or malicious behavior.

Install only if you are comfortable using AI-HIVE, storing or supplying an AI-HIVE API key, uploading selected reference images, and potentially spending credits for generation. Review prompts, routing, batch size, and uploaded files before running generate commands, and keep product claims, rights, and measurements independently verified.

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 (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 88% confidence
Finding: The skill advertises runnable scripts that use environment variables, local file I/O, shell execution, and network access, yet no permissions are declared. This creates a transparency and policy-enforcement gap: operators and users cannot accurately assess what the skill can do, and a downstream agent may execute broader capabilities than expected, including writing files or making external API calls.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 93% confidence
Finding: The documented purpose is narrowly framed around furniture placement image mockups, but the described behavior extends to general chat, video generation, account/wallet queries, model enumeration, credential setup, and generic media upload. This mismatch is dangerous because it obscures materially broader functionality, increasing the risk of unauthorized data exposure, unexpected billing, credential handling, and abuse of capabilities outside the user’s intended task.

Context-Inappropriate Capability

Medium

Confidence: 90% confidence
Finding: The file contains a user-info capability that retrieves account and wallet details, which is outside the stated furniture room-placement image workflow. Even though the final skill entrypoint does not expose this subcommand directly, bundling unrelated account-inspection logic increases the blast radius if the script is reused, invoked differently, or copied into a broader agent context.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The skill enables implicit invocation without any visible activation constraints, so the agent may auto-select it in conversations that only loosely match home design or image-generation topics. Because this skill can steer users into a production workflow with external API usage and generated deliverables, accidental activation can cause unintended tool use, unwanted data handling, or unauthorized image-generation actions.

Natural-Language Policy Violations

Medium

Confidence: 88% confidence
Finding: The default prompt forces Chinese output regardless of the user's language preference, which can override user intent and reduce transparency around what the skill will do. In a workflow that produces runnable commands and operational steps, language forcing can also increase the chance that users misunderstand external API actions, pricing, or consent-sensitive steps.

Static analysis

No suspicious patterns detected.