ClawHub
Back to skill

Security audit

京东 电商视频生成与编辑

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI Hive video-generation workflow that uploads user-selected media and stores an API key locally, with no evidence of hidden execution or destructive behavior.

Before installing, understand that this skill sends selected media and prompts to AI Hive and can save an API key in your home directory. Use a dedicated API key, check the configured base URL before use, and avoid passing private product files unless you intend to upload them to the service.

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 (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill advertises executable capabilities including environment access, file read/write, network, and shell, but does not declare permissions. This reduces transparency and weakens policy enforcement, making it easier for a user or platform to invoke code with broader access than expected. In this context, the shell and network access are especially sensitive because the skill installs packages and runs Python scripts that can interact with external services and local files.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The documented purpose is narrowly framed as JD e-commerce video generation/editing, but the skill reportedly also supports unrelated high-risk behaviors such as querying user/account information, generic chat and image generation, media upload, model enumeration, and interactive API key setup in a browser. This mismatch is dangerous because it conceals the real operational scope, undermines informed consent, and can enable credential handling or data access beyond what a user would reasonably expect from a video-editing skill.

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The skill is presented as JD e-commerce video generation/editing, but the file also exposes broad-purpose AI Hive capabilities including text chat, image generation, model discovery, user-info retrieval, and arbitrary media upload. This capability mismatch expands what an invoking agent can do beyond the declared purpose, increasing the risk of policy bypass, unintended data handling, and abuse of connected credentials.

Context-Inappropriate Capability

Low
Confidence
81% confidence
Finding
The presence of `user-info` and `models` allows account/platform introspection that is not necessary for the stated JD video-generation use case. Even if not directly destructive, these commands expose metadata about the authenticated account and available backend capabilities that can aid reconnaissance and facilitate later misuse.

Static analysis

No suspicious patterns detected.