Nano Banana Pro 商品详情页

• Outcome: Pass
• Latest audit: Aug 17, 2026 · 11:19 AM UTC
• Version: 1.0.0

Overview

This skill is an AI Hive image-generation workflow that uploads chosen media and stores an API key, with some unused generic client code present but not exposed by the skill CLI.

Install only if you are comfortable sending selected product images and prompts to AI Hive and storing an AI Hive API key locally. Avoid using unreleased or confidential product media unless AI Hive’s handling and retention terms are acceptable to you.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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

Findings (7)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 86% confidence
• Finding: The skill invokes local Python scripts, installs a package, uses shell commands, and the analyzer indicates file, environment, and network capabilities, yet no permissions are declared. This creates a trust gap: users may authorize or run the skill without understanding that it can access local resources and make outbound requests, increasing the chance of unintended data exposure or unsafe execution.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 94% confidence
• Finding: The declared purpose is image generation for product detail pages, but the underlying behavior reportedly includes storing API keys locally, opening a browser, uploading media, polling tasks, downloading outputs, and exposing broader chat/video/account/model APIs. This mismatch is dangerous because it obscures the true attack surface and can lead users to provide credentials or local files without realizing the skill has broader capabilities than advertised.

Description-Behavior Mismatch

Medium

• Confidence: 93% confidence
• Finding: The file is packaged as a narrowly scoped Nano Banana Pro product-detail image skill, but it contains broad generic capabilities for account inspection, model enumeration, text chat, media upload, and video generation. In an agent skill ecosystem, this scope expansion violates least privilege and increases the attack surface because a caller expecting only image generation may unknowingly gain access to unrelated remote operations.

Context-Inappropriate Capability

Low

• Confidence: 88% confidence
• Finding: The user-info capability is unrelated to the declared purpose of generating product-detail images and exposes account metadata and likely wallet/balance information. While not a direct code-execution issue, it creates unnecessary data exposure and can assist reconnaissance against the user's AI Hive account.

Context-Inappropriate Capability

Medium

• Confidence: 90% confidence
• Finding: A generic text chat endpoint is unrelated to the skill's stated Nano Banana Pro image-generation function and enables arbitrary prompt forwarding to remote models. In an agent context, this can be abused to exfiltrate user-provided content, bypass expected safety boundaries of the skill, or incur unintended usage and cost outside the advertised scope.

Context-Inappropriate Capability

Medium

• Confidence: 91% confidence
• Finding: Video generation materially exceeds the declared image-only product-detail-page purpose and allows additional media uploads and remote processing not expected by users. This broadens both privacy risk and billing exposure because the skill can transmit and process more data types than its description implies.

Missing User Warnings

Medium

• Confidence: 86% confidence
• Finding: The skill entry point accepts local file paths and uploads them to a remote service, but it does not present an explicit privacy warning or confirmation at the skill-facing commands. Because product-detail workflows often use unreleased product photos or proprietary assets, silent upload behavior increases the risk of accidental disclosure of sensitive media.

VirusTotal

63/63 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.