Security audit

guaikei-douyin-discovery

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent Douyin data-collection tool, but it needs review because it can trigger on broad requests and automatically saves collected public data locally.

Install only if you are comfortable sending Douyin search/account/comment requests and GUAIKEI_API_TOKEN to guaikei.com. Use it for explicit Douyin tasks, watch for accidental activation on generic research prompts, and periodically delete or protect the logs directory because it may contain saved competitor, author, and comment data.

SkillSpector

By NVIDIA

Vulnerability Patterns

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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (9)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 84% confidence
Finding: The skill explicitly requires the `GUAIKEI_API_TOKEN` environment variable, but the skill metadata does not declare permissions or clearly scope this environment access. Undeclared env access weakens the trust boundary for agents and reviewers, who may invoke the skill without realizing it depends on secret material.

Description-Behavior Mismatch

Low

Confidence: 86% confidence
Finding: The CLI writes full search results to a local JSON log file after completing what appears to be an on-demand search operation. Even if intended for debugging or auditability, this creates silent data persistence on disk, which can expose searched keywords and returned content to other local users, backups, or later processes without the operator realizing it.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The description says the skill should apply even when the user only says broad requests like '帮我看看同行' without mentioning Douyin. That creates an overly broad trigger surface where generic research or competitor-analysis requests may silently invoke a Douyin data-collection skill, causing unintended external data access and mismatched user expectations.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The '隐式触发' section includes very general requests like competitor research, high-like content search, and hot-topic discovery with no Douyin mention required. In context, this makes accidental activation likely across many normal research workflows, leading the agent to collect platform-specific data the user did not explicitly request.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The natural-language trigger examples are broad enough to match everyday phrases like '搜一下', '看看', or '帮我看看同行', which can cause the skill to activate when the user did not explicitly intend Douyin scraping. Because this skill performs data collection and log export, unintended invocation can lead to unnecessary third-party requests, collection of comment/author data, and privacy-sensitive processing without clear user confirmation.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The intent-recognition priority rules use generic terms such as '热搜', '热点', '搜索', '找', and broad mappings around '视频', creating a high risk of over-triggering on common conversational language. In this context, over-broad routing is dangerous because the skill can query an external service using a configured API token and retrieve/export public but potentially sensitive behavioral or comment data without sufficiently scoped user intent.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The introduction emphasizes convenience and '安全' but does not prominently warn that the skill collects author data, comment content, and writes structured logs for later analysis. This can mislead operators into underestimating privacy, compliance, and retention risks, especially when comments and account-level data are exported and stored locally.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The CLI persists the full fetched results to a local JSON file automatically after successful execution, but there is no explicit user consent, warning, or opt-in control. Because the skill is designed to gather competitor-account content and related data, the output may contain sensitive business intelligence or user-generated data that can remain on disk unexpectedly and be exposed to other local users, backup systems, or later compromise.

Missing User Warnings

Low

Confidence: 89% confidence
Finding: Search results are persisted to a local file without an explicit user-facing warning or consent mechanism, which can surprise users who expect transient console output only. This increases the risk of unintended retention of potentially sensitive operational research, competitor analysis, or query history on shared systems.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.