Security audit

品牌数字代言人内容日历｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed AI-HIVE workflow helper for brand-avatar content planning and media generation, with expected API-key, upload, cost, and local-file handling risks.

Install only if you intend to use AI-HIVE for brand-avatar marketing content. Treat media uploads and generation as potentially billable, review prompts and routing before running commands, use only assets you are authorized to upload, and decide whether storing the API key in ~/.ai-hive/config.json is acceptable for your environment.

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 (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 88% confidence
Finding: The skill documents executable capabilities including environment-variable access, file read/write, network calls, and shell usage, but does not declare permissions or clearly scope them. This weakens least-privilege controls and transparency, increasing the risk that an agent could invoke local file access, external network operations, or shell-based tooling unexpectedly in response to user prompts.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 92% confidence
Finding: The documented purpose is narrower than the behaviors implied by the skill, which include generic chat, account/model querying, standalone uploads, and local ffmpeg processing. Description-behavior mismatch is dangerous because it can mislead operators and downstream policy systems about what the skill can actually do, enabling unreviewed data exfiltration, unintended media manipulation, or broader-than-expected external interactions.

Description-Behavior Mismatch

Medium

Confidence: 92% confidence
Finding: The file implements a broad AI Hive CLI for chat, image, video, model listing, uploads, and task polling, which materially exceeds the declared skill purpose of producing a brand-avatar content calendar workflow. This scope expansion increases attack surface and enables repurposing of the skill for unrelated generation tasks, weakening policy and operational controls tied to the manifest.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill-specific generate path accepts arbitrary prompts and media, selects a video mode, and forwards directly into a generic video-generation function without enforcing delivery of a brand-avatar content calendar artifact. That mismatch makes the skill easy to use as a general video generator, bypassing the narrower purpose and any contextual safeguards implied by the manifest.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The manifest allows implicit invocation with no activation boundaries, so the skill may be triggered in contexts the user did not explicitly intend. Because this skill can generate production-ready workflows and runnable AI-HIVE commands for content generation, unintended invocation could cause unauthorized external API usage, content generation, or execution guidance in sensitive marketing workflows.

Natural-Language Policy Violations

Medium

Confidence: 84% confidence
Finding: The default prompt hard-codes Chinese output without user choice, which can override user expectations and reduce transparency about how the skill will behave. On its own this is not a classic security flaw, but it can contribute to unsafe or misleading operation if users misunderstand generated instructions, compliance notes, or external-tool actions because the skill silently forces a language preference.

Static analysis

No suspicious patterns detected.