Security audit

AI大模型专家｜Seedance API 中转渠道替代方案｜AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a planning aid for evaluating a Seedance-to-AI-HIVE video workflow migration and only shows proportionate local file output plus user-directed evidence gathering.

Before installing, treat this as a Chinese-language planning and checklist skill, not an automated migration tool. Use non-production samples, keep API keys in environment variables, verify current AI-HIVE and Seedance terms/pricing yourself, and choose an output path where overwriting a JSON plan file is acceptable.

SkillSpector

Vulnerability Patterns

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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 70% confidence
Finding: Without declared permissions the skill's intent is opaque and cannot be validated.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 93% confidence
Finding: 描述将该技能表述为一个可完成异步视频任务迁移拆解与对比分析的工具，听起来应当产出较完整的方法设计或执行性分析结果；但代码仅根据命令行参数生成一个静态 JSON 模板，写入本地文件，内容是待办检查项和样本占位信息。它没有访问 Seedance、AI-HIVE 或任何外部服务，也没有实现状态机、幂等键、失败恢复、轮询策略、下载校验或平台对比逻辑。因此其实际主要用途与声明存在明显落差，属于描述显著夸大能力的情况。

Intent-Code Divergence

Low

Confidence: 77% confidence
Finding: Lines L57-L67 present an operational guarantee about tool behavior ('只在本地生成 JSON 计划，不访问第三方，也不会提交计费任务') while this file contains only markdown and a sample command, not the implementation of scripts/async-video-plan.py. That creates an intent/documentation claim not verifiable from the provided code artifact and could mislead users about actual side effects.

Intent-Code Divergence

Medium

Confidence: 98% confidence
Finding: The module docstring at L02 explicitly describes the tool as read-only, which implies no persistent side effects. However, the script calls Path(args.output).write_text(...) at L35 to create or overwrite a local file, so the documentation contradicts the actual behavior.

Natural-Language Policy Violations

Low

Confidence: 84% confidence
Finding: The natural-language metadata forces specific languages/locales in the user-facing interface and invocation prompt, but does not indicate that the user can choose their preferred language. Under the policy, locale or language constraints should be opt-in or clearly justified; this file provides neither.

Natural-Language Policy Violations

Medium

Confidence: 95% confidence
Finding: The module docstring, constants, and CLI help text are written in Chinese, which imposes a specific language on users of the skill. There is no indication that the script is region-specific, nor any option for users to choose another language, so this is a natural-language locale policy concern.

Static analysis

No suspicious patterns detected.