Security audit

Seedream 5.0 Lite 营销图片

Security checks across malware telemetry and agentic risk

Overview

This skill is an image-generation helper that uses a user-provided AI Hive API key and user-selected files, with some unused generic client code that is broader than the advertised image-only purpose.

Install only if you are comfortable giving this skill an AI Hive API key, uploading selected reference files to AI Hive or its object storage, and saving generated outputs locally. Avoid passing sensitive files to the upload or image options, and consider reviewing the broader unused helper code if you require strict image-only source hygiene.

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 91% confidence
Finding: The skill exposes shell, filesystem, environment, and network-capable execution via the documented `pip3 install` and `python3 .../imagegen.py` commands, yet declares no permissions. This creates a transparency and sandboxing gap: callers may invoke a seemingly simple image-generation skill without realizing it can read/write files, access secrets from the environment, and make outbound requests.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 95% confidence
Finding: The documented purpose is narrowly limited to Seedream marketing image generation, but the analyzed behavior reportedly includes unrelated capabilities such as text chat, video generation, model enumeration, user/wallet queries, generic media upload, and browser-assisted API key setup. That mismatch is dangerous because it hides materially broader data access and operational behaviors than a user would expect from the skill description.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill is presented as a narrowly scoped Seedream 5 Lite marketing-image generator, but the file actually implements a broad generic AI Hive client with account inspection, model enumeration, media upload, and task management features. This scope mismatch increases attack surface and user surprise, making it easier for a seemingly harmless image skill to be used as a general-purpose API access wrapper.

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The file includes text chat and video generation API methods even though the declared skill is for Seedream 5 Lite marketing-image generation. Unnecessary latent capabilities are risky because they can be repurposed by an agent or operator to perform actions outside the advertised trust boundary, including sending arbitrary prompts and media to other model types.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The module docstring explicitly describes a general multimodal AI tool, contradicting the skill metadata that promises a specific Seedream 5 marketing-image function. This is dangerous because misleading packaging can conceal broader capabilities from reviewers and users, undermining least privilege and informed consent.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.