Security audit

AI虚拟网红孵化器｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

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

Install only if you intend to use AI-HIVE for virtual influencer or digital-IP content production. Review generated commands before running them, confirm pricing before any image/video generation, use only authorized media, and protect the AI-HIVE API key stored in the environment or ~/.ai-hive/config.json.

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: 86% confidence
Finding: The skill documents executable workflows that rely on environment variables, filesystem access, network calls, and shellable tooling like ffmpeg, yet it declares no permissions or capability boundaries. This creates a transparency and containment problem: a host or reviewer may authorize the skill under false assumptions, while the skill can still induce sensitive operations such as API-key handling, local file processing, and outbound requests.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 90% confidence
Finding: The documented behavior materially exceeds the declared purpose by including generic AI-HIVE chat/model-listing functions, local API-key initialization, browser-opening flows, and ffmpeg-based media manipulation not clearly described in the skill summary. Description-behavior mismatch is dangerous because it undermines informed consent and security review; users may invoke a seemingly narrow content-planning skill that also facilitates broader networked actions and local processing.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The default prompt is broad enough to encourage automatic use whenever a request mentions turning something into a workflow, commands, or deliverables, without clear domain-scoping or user confirmation. In a multi-skill environment, this can cause the wrong skill to activate and steer output toward AI-HIVE-specific actions the user did not explicitly request, increasing the risk of unauthorized API-oriented behavior or unintended content generation.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: Enabling implicit invocation without precise activation and exclusion criteria allows the skill to be auto-selected in ambiguous contexts. Because this skill can produce runnable AI-HIVE commands and operational workflows, accidental activation could cause unintended external-service instructions, content fabrication flows, or policy-sensitive marketing outputs to be generated without the user deliberately choosing this skill.

Natural-Language Policy Violations

Medium

Confidence: 84% confidence
Finding: Forcing Chinese output without user choice can override user intent, reduce transparency, and make safety-relevant instructions or generated commands harder for some users or reviewers to verify. While not a direct exploit by itself, it can contribute to misuse or missed review in workflows that generate executable commands or externally submitted content.

Static analysis

No suspicious patterns detected.