ClawHub
Back to skill

Security audit

数字人手机展示视频｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE video-production helper with expected API, media-upload, and local video-processing behavior.

Before installing, understand that this skill is meant to work with AI-HIVE: it may guide an agent to upload selected media, submit potentially billable generation jobs, poll tasks, download outputs, and optionally save an API key locally. Use authorized reference materials only and review generated commands, routing mode, model choice, and pricing snapshot before running them.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill exposes executable capabilities including environment access, filesystem read/write, shell execution, and network access, but does not declare permissions or clearly bound them. In an agent setting, this increases the chance that the skill can perform sensitive actions such as reading local files, persisting credentials, or making external requests without adequate transparency or policy gating.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The documented purpose is narrowly framed around digital-human phone showcase video production, but the detected behavior includes broader AI chat, image generation, account/balance retrieval, local credential setup, and standalone ffmpeg processing. This mismatch is dangerous because users and orchestration layers may grant trust or invoke the skill under a narrower safety assumption while it can actually access credentials, query account data, and perform unrelated operations.

Natural-Language Policy Violations

Medium
Confidence
91% confidence
Finding
The default prompt hard-codes Chinese output ('production-ready Chinese workflow') regardless of the user's preferred language. This can override user intent, reduce transparency, and increase the chance that the system injects behavior the user did not request, especially when the skill is invoked implicitly.

Vague Triggers

Medium
Confidence
94% confidence
Finding
Implicit invocation is enabled without any visible trigger constraints, allowing the skill to activate automatically based on broad matching. In a skill that can generate runnable AI-HIVE commands and media-production workflows, unexpected activation can steer user sessions, cause unintended API-related actions, or apply the skill in contexts the user did not clearly authorize.

Static analysis

No suspicious patterns detected.