Security audit

红果短剧生成工作室｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI-HIVE short-video production helper with paid API and local media handling that fit its stated purpose, though users should be aware of implicit invocation and API-key storage.

Install only if you intend to use AI-HIVE for Chinese vertical short-drama image/video workflows. Expect local media paths to be read when you pass them, outputs to be written to the configured output directory, and an API key to be stored locally if you run init. Review prompts and pricing before submitting generation tasks, especially batch jobs.

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 (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 91% confidence
Finding: The skill advertises executable scripts that use environment variables, filesystem access, networking, and shell-adjacent tooling like ffmpeg, but no permissions are declared. This creates a capability-transparency gap: operators or platforms may invoke the skill without understanding it can access local files, make external API calls, and process user-supplied paths, increasing the chance of over-privileged or unsafe deployment.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 94% confidence
Finding: The documented purpose is a constrained short-drama production assistant, but the described/code-detected behavior includes broader AI-HIVE text chat, account/wallet retrieval, generic model enumeration, and standalone ffmpeg operations. This mismatch is dangerous because users, reviewers, or orchestrators may trust the skill with inputs appropriate for planning only, while it can access account metadata, perform broader remote operations, or manipulate local media beyond the stated scope.

Vague Triggers

Medium

Confidence: 85% confidence
Finding: The default prompt is broad enough to activate on generic requests and automatically steer the conversation into this skill, even when the user did not explicitly ask for it. Because the skill is specialized for AI-HIVE workflow generation and runnable commands, overbroad triggering can cause unintended tool-routing, irrelevant automation, or prompt-context takeover in normal conversations.

Vague Triggers

High

Confidence: 96% confidence
Finding: Enabling implicit invocation without tightly scoped activation constraints allows the skill to be invoked automatically based on weak semantic matches rather than explicit consent. In this skill, that is more dangerous because it can inject production-oriented workflow steps and runnable AI-HIVE commands into unrelated sessions, increasing the chance of unauthorized actions, confusion, or abuse of downstream generation capabilities.

Natural-Language Policy Violations

Medium

Confidence: 82% confidence
Finding: The manifest instructs the system to produce a 'production-ready Chinese workflow,' which forces a language choice without checking the user's preference. This can override user intent, reduce transparency, and increase the risk of misunderstandings in workflow or code-generation contexts where precision matters.

Static analysis

No suspicious patterns detected.