Seedream 5.0 — AI Image Generation & Editing by ByteDance

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:17 PM UTC
• Version:
  1.0.2

Overview

This skill is a disclosed paid Atlas Cloud image-generation helper that sends user-provided prompts and selected images to the service, with no hidden persistence or destructive behavior found.

Install only if you are comfortable sending prompts, image URLs, and deliberately uploaded local images to Atlas Cloud and paying per generated image. Avoid sensitive images or confidential prompts, monitor Atlas Cloud usage, and confirm the selected model and batch size before running larger jobs.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (3)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 89% confidence
• Finding: The skill requires an API key and clearly performs outbound network requests, but it does not declare corresponding permissions. That creates a transparency and governance gap: users or hosting platforms may authorize the skill without understanding that it can access environment secrets and transmit prompts or image references to a third-party service.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 95% confidence
• Finding: The documented purpose narrows the skill to ByteDance Seedream image generation, but the behavior reportedly supports broader Atlas Cloud model enumeration, arbitrary model selection, and local file upload. This mismatch is security-relevant because users may consent to a limited image-generation workflow while the implementation can route data to other models or upload local content in ways not clearly disclosed.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The trigger text is extremely broad, covering nearly any request related to visuals, graphics, illustrations, or AI art. Over-broad invocation increases the chance the skill activates in contexts where users did not intend third-party image generation or external data transfer, which can lead to unnecessary disclosure of prompts, images, or paid API usage.

VirusTotal

66/66 vendors flagged this skill as clean.