Security audit

短剧场景连续性工作室｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE workflow helper for short-drama continuity that uses user-directed media generation, uploads, downloads, and local video processing.

Before installing, understand that this skill is for AI-HIVE media production: it may upload media you choose, use an AI-HIVE API key, create paid generation tasks after confirmation, save downloads locally, and process videos with ffmpeg. Do not provide private or unlicensed media, and prefer environment or CLI API keys unless you are comfortable with the local config file.

SkillSpector

Vulnerability Patterns

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
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: 93% confidence
Finding: The skill exposes executable examples and operational guidance that rely on environment variables, local file I/O, shell execution, and network access, yet no permissions are declared. That mismatch weakens sandboxing and user consent boundaries because a host may invoke the skill without understanding that it can access local files, call external APIs, and run local tooling like ffmpeg.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 95% confidence
Finding: The documented purpose is a continuity/planning studio, but the behavior includes broader capabilities such as generic chat, arbitrary media upload, local credential setup/storage, and general ffmpeg editing operations. This description-behavior gap is dangerous because users and platform policy may authorize a narrow planning tool while actually enabling a much more general toolchain that can move data externally, persist secrets, and manipulate local media beyond the stated use case.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The manifest enables implicit invocation with no visible narrowing conditions, allowing the skill to be auto-selected in situations broader than its intended scope. Because this skill can generate runnable AI-HIVE commands and workflow guidance, over-broad triggering increases the chance of unintended execution paths, inappropriate use for sensitive marketing/media tasks, or user-surprising behavior.

Natural-Language Policy Violations

Medium

Confidence: 88% confidence
Finding: The default prompt hard-codes Chinese output regardless of the user's language preference, which can cause misunderstandings about generated workflows, commands, or safety constraints. In a skill that produces production-ready instructions and runnable commands, language mismatch can lead to operator error, incorrect execution, or reduced ability to notice problematic outputs.

Static analysis

No suspicious patterns detected.