Security audit

爆款视频原创复刻工作台｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed AI-HIVE video-remake workflow with expected media upload, API-key, and paid-generation behavior, but users should notice its broad activation wording and local credential storage.

Install only if you intend to use AI-HIVE for authorized video production. Review the generated parameters before running commands, avoid uploading media you do not have rights to use, and keep the stored API key protected because generation may cost money.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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

Findings (7)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 92% confidence
Finding: The skill embeds operational code examples that use environment variables, local file paths, shell commands, and outbound network access, but it does not declare permissions or capability boundaries. This creates a mismatch between what the skill appears allowed to do and what a host agent might attempt, increasing the chance of unintended file access, secret exposure, or network actions if invoked automatically.

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The file implements a broad, generic AI client (chat, image, video, model listing, uploads) while the declared skill is narrowly framed as a video-remake workstation. This scope mismatch is dangerous because a caller or orchestrator may expose capabilities to users that were not reviewed, approved, or intended by the manifest, enabling policy bypass and unexpected data flows.

Description-Behavior Mismatch

Medium

Confidence: 86% confidence
Finding: The skill exposes user-info retrieval, including account and wallet data, even though that capability is not described in the manifest. Undeclared access to account metadata increases the risk of privacy leakage and over-privileged skill behavior, especially if an agent can invoke this command without the user's informed intent.

Intent-Code Divergence

Medium

Confidence: 83% confidence
Finding: The module docstring presents the tool as a generic AI capability wrapper, which contradicts the specialized manifest for original-remake video workflows. This discrepancy can mislead reviewers and policy systems, making hidden or extra capabilities easier to smuggle into production under a narrower description.

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The trigger phrase `AI视频` is so broad that it can match many unrelated user requests, causing the skill to activate outside its intended remake workflow. Over-broad activation can route general requests into a capability-heavy skill that discusses uploads, generation, shell commands, and external APIs, raising the risk of inappropriate tool use or user confusion.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The manifest enables implicit invocation for a skill that can generate production-ready workflows and runnable AI-HIVE commands, but it does not define any activation scope or trigger constraints. That increases the chance the agent will invoke this skill in unrelated contexts, causing unintended generation of external API actions, media-processing workflows, or policy-sensitive remake guidance without clear user intent.

Natural-Language Policy Violations

Medium

Confidence: 83% confidence
Finding: The default prompt hard-codes Chinese output regardless of the user's language preference. While not a direct code-execution issue, forced language can obscure safety notices, reduce user understanding of generated commands and checklists, and make consent or authorization boundaries less clear in a workflow involving external media generation and delivery.

Static analysis

No suspicious patterns detected.