guaikei-douyin-find-trending-videos

• Outcome: Review
• Latest audit: Aug 20, 2026 · 3:19 AM UTC
• Version: 1.0.0

Overview

The skill is mostly coherent for Douyin public-data lookup, but it needs Review because it can trigger too broadly and automatically stores scraped results locally without a user-controlled retention option.

Install only if you intend to query public Douyin data through Guaikei's API and are comfortable with automatic local JSON logs. Review or clear the logs directory after use, avoid collecting more comments or profile-linked data than needed, and confirm ambiguous requests before letting an agent invoke this skill for generic trend research.

SkillSpector

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (14)

Description-Behavior Mismatch

Medium

• Confidence: 91% confidence
• Finding: The schema explicitly describes `video_url` as both playback and download addresses, which conflicts with the skill’s stated restriction that it is not for downloading videos. Even though this file is only a schema, it documents and normalizes download-oriented output that downstream agents or users could use to enable prohibited content retrieval, increasing the chance of policy bypass or misuse.

Description-Behavior Mismatch

Medium

• Confidence: 89% confidence
• Finding: The CLI persists fetched comment data to a local JSON file after displaying results, which goes beyond a purely read/query behavior and creates a new local data-retention surface. Comment content can contain personal or sensitive information, and silently writing it to disk increases exposure through unintended collection, later reuse, or access by other local users/processes.

Description-Behavior Mismatch

Medium

• Confidence: 94% confidence
• Finding: The CLI persists fetched Douyin results to a local JSON file automatically, even though the stated skill purpose is read-only lookup of public data. Silent retention creates an unnecessary data-at-rest footprint that may expose scraped content, account identifiers, or query history to other local users, later processes, or accidental publication.

Description-Behavior Mismatch

Medium

• Confidence: 88% confidence
• Finding: The CLI writes full search outputs, including queried keywords and returned public-data results, to a local JSON log file without any visible consent or retention controls. Even if the data is public, local persistence can expose a user's research interests, monitoring targets, or collected datasets to other local users, backups, or downstream tooling, which exceeds a simple read-only/search expectation.

Vague Triggers

Medium

• Confidence: 86% confidence
• Finding: The description states the skill applies even when the user does not mention Douyin, which broadens routing beyond a clearly bounded platform-specific intent. That can cause the agent to invoke this data-collection skill for generic research requests, leading to unintended third-party data access, unnecessary token use, and confusion about user intent or platform scope.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The '隐式触发' examples include broad phrases such as '帮我做短视频竞品调研' and '最近什么热门' without a platform qualifier, making accidental invocation likely. In context, this skill performs external data retrieval against Douyin, so over-broad triggers can misroute ordinary analysis requests into a platform-specific scraping workflow the user did not ask for.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The package description describes broad 'public data intelligent analysis' use cases and marketing-oriented capabilities without clearly constraining invocation conditions, data boundaries, or prohibited actions beyond a few examples. In an agentic environment, overly broad descriptions can cause the skill to be selected for ambiguous requests, increasing the chance of unnecessary third-party data access, scope creep, or use outside the user's actual intent.

Natural-Language Policy Violations

Medium

• Confidence: 79% confidence
• Finding: The description is entirely in Chinese and does not state whether Chinese is required for input, output, or target content, nor whether users may choose another language. This can lead to incorrect routing or a poor user-consent boundary in multilingual environments, where a user may not expect language-specific processing or Chinese-only outputs.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The natural-language trigger examples are broad enough that ordinary phrases like '看看这个视频' or '今天什么火' could invoke this scraping skill even when a user did not clearly intend to use Douyin-specific data collection. In an agent setting, overbroad routing can cause unintended collection of public profile, video, or comment data and unexpected external API usage.

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The intent-recognition rules rely on generic keywords such as '热点', '榜单', '搜索', and '博主', which are common across many domains and can easily misroute unrelated user requests into this skill. That increases the chance of unintended scraping, incorrect tool execution, and disclosure or retention of scraped content when the user did not specifically request it.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The README advertises automatic JSON log export for search results, posts, and comments but does not clearly warn about retention, sensitivity, or downstream handling of scraped content. Stored comment text and account-linked data may contain personal data or sensitive opinions, so silent persistence increases privacy, compliance, and secondary exposure risk.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The code writes the full comment results to a local JSON file without prior warning or explicit user consent. Because comments may include usernames, text, timestamps, and other public-but-still-sensitive data, silent persistence increases privacy risk and may violate the expected behavior of a query-only skill.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The CLI writes fetched results to a local file without warning, confirmation, or an explicit output option. This can surprise users and create unintended storage of potentially sensitive browsing targets, scraped content, or operational metadata, increasing the chance of local disclosure or downstream misuse.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The program silently stores search results to a local file after completing the request, without explicit user warning or confirmation. This creates a privacy and data-governance risk because users may expect ephemeral query execution, while the tool leaves behind a searchable artifact containing their keywords and collected results.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.