Security audit

guaikei-douyin-comments-for-sentiment-insight

Security checks across malware telemetry and agentic risk

Overview

The skill mostly does what it says, but it has broad auto-activation rules and contradictory runtime token/contact behavior that users should review before installing.

Install only if you are comfortable sending Douyin keywords, target URLs, request parameters, and your GUAIKEI_API_TOKEN to guaikei.com, and with successful search/post/comment results being saved under the local logs directory. For ambiguous requests, explicitly confirm Douyin before using this skill.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 84% confidence
Finding: The skill claims to perform data retrieval and analysis, but also states that results are automatically saved under `logs/` as local JSON files. That persistence behavior expands data handling beyond transient processing and can expose scraped comments, URLs, keywords, and possibly sensitive research context to other local users, backups, or downstream tools if not clearly disclosed and controlled.

Vague Triggers

Medium

Confidence: 77% confidence
Finding: The trigger guidance says the skill should activate for generic short-video research even when the user does not mention Douyin. That broad scope can cause unintended invocation in unrelated contexts, leading the agent to collect or process data from the wrong platform or surprise the user with external data access they did not explicitly request.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The README defines invocation mappings using broad conversational phrases like '搜索 / 搜一下 / 找 AI 相关视频', which can cause the skill to activate from ordinary language without sufficiently explicit user intent to use an external Douyin scraping capability. In an agent environment, this raises the risk of unintended data transmission and unintended execution against a third-party API, especially because the skill performs external queries based on user text.

Vague Triggers

Low

Confidence: 88% confidence
Finding: The hotlist examples include generic phrases like '今天有什么热点 / 热搜榜 / 热点榜单', which are common across many domains and may spuriously route unrelated requests into this Douyin-specific skill. While lower severity than the broader search mappings, it still creates an unintended invocation risk that can lead to unnecessary external API calls and confusing behavior.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The documentation tells users to obtain and configure an API token, making clear that the skill relies on an external service, but it does not clearly warn that user-supplied keywords, URLs, and related request data will be transmitted to that external provider. This creates a transparency and consent gap: users may disclose sensitive search terms, account URLs, or investigative targets without understanding they are being sent off-platform.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.