AI 书稿图片重设计

• Outcome: Review
• Latest audit: May 29, 2026 at 9:36 PM UTC
• Version:
  1.0.100

Overview

This is a usable static image-redesign tool, but it stores API keys in the browser and can send keys and uploaded manuscript images to user-entered remote endpoints without strong warnings.

Install only if you are comfortable with a browser-based tool handling API credentials and uploaded images. Use a trusted endpoint, prefer short-lived or low-privilege keys, avoid confidential manuscript images unless the provider is approved, and clear browser storage or modify the page to avoid saving the API key.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (3)

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill explicitly supports browser-side entry of API Base URL and API Key, which normalizes direct secret handling in a static frontend. Even though the document later notes that production use should prefer a backend proxy, the warning is not prominent up front, and users may expose long-lived credentials to browser storage, page scripts, shoulder-surfing, or accidental sharing of the generated static page.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The code stores the API key in localStorage, which is long-lived and readable by any script executing in the same origin, including malicious third-party scripts or future XSS payloads. This increases credential exposure risk well beyond the current session, especially because the page also loads external resources and is designed for repeated use.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The application sends uploaded images and user prompt content to a user-configurable remote API but does not clearly warn users that potentially sensitive content will leave the browser and be transmitted to third-party infrastructure. In this skill context, users may upload manuscript graphics, tables, or proprietary publishing material, so silent transmission creates meaningful confidentiality and compliance risk.

VirusTotal

64/64 vendors flagged this skill as clean.