Security audit

guaikei-douyin-comments-extract

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed, read-only Douyin public-data retrieval tool, with the main caution that it saves fetched results locally by default and sends requests through a third-party API.

Install only if you are comfortable sending Douyin keywords, links, IDs, requested limits, and your GUAIKEI_API_TOKEN to guaikei.com. Treat the token as a secret, avoid committing it to scripts or dotfiles, and review or delete the generated logs directory when collected comments or account/content metadata should not be retained locally.

SkillSpector

By NVIDIA

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 (10)

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The CLI persists the full fetched results to a local JSON file after completing the request, which creates an additional data sink beyond immediate console output. Even if intended for debugging or audit purposes, silently storing potentially sensitive scraped content on disk can expose data to other local users, backups, or later exfiltration, especially because the behavior is not surfaced to the user in this file.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The top-level description lists many broad business and analysis scenarios, which can cause the agent to invoke this skill for loosely related requests without strong user confirmation. Because the skill sends user-supplied keywords/links and an API token to a third-party service, over-broad triggering increases the chance of unintended data disclosure and unnecessary external calls.

Vague Triggers

Low

Confidence: 88% confidence
Finding: The natural-language examples include vague phrases like '帮我做抖音竞品分析' that do not clearly distinguish between needing external data retrieval versus general strategic advice. This can over-activate the skill and cause unnecessary transmission of user prompts, URLs, or identifiers to the third-party API when a clarifying question should have been asked first.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The README prominently advertises automatic JSON log export of collected Douyin data to a local logs directory, but it does not clearly warn users up front that executing the skill writes potentially sensitive collected data to disk. This can lead to unintended local retention, later disclosure through backups/sharing, or mishandling of scraped comments and account data.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The quick-start instructs users to obtain and set GUAIKEI_API_TOKEN before running commands, but does not provide an upfront security warning about credential handling. Users may paste tokens into insecure shells, commit them to dotfiles/scripts, or expose them in shared environments, risking unauthorized API use.

Vague Triggers

Medium

Confidence: 79% confidence
Finding: The natural-language trigger examples use broad everyday phrases such as '搜索' and '搜一下', which can cause overbroad or unintended activation by an agent. In an automated assistant context, ambiguous triggers increase the chance of collecting external platform data without sufficiently explicit user intent.

Vague Triggers

Medium

Confidence: 81% confidence
Finding: The intent-recognition rule maps generic phrases like '热搜', '热点', '榜单', and '今天什么火' directly to a Douyin hot-list command with minimal contextual constraints. In a multi-skill or conversational environment, this can misroute general queries into external data collection and create unintended third-party requests.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The CLI writes full comment extraction results to a local JSON file derived from the target URL without clearly notifying the user in this file. Because comments may contain personal data, sensitive opinions, usernames, or monitoring targets, this creates an unannounced local data retention channel that can expose scraped data to other local users, backups, or later compromise. In the context of a Douyin comment scraping and analysis skill, silent persistence is more dangerous because the tool is specifically designed to collect potentially sensitive user-generated content at scale.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: This code writes fetched Douyin results to a local JSON file without warning or consent in the CLI flow shown here. Undisclosed persistence increases privacy and compliance risk because users may believe the tool only retrieves/analyzes data transiently, while the skill actually leaves durable copies on disk.

Missing User Warnings

Low

Confidence: 92% confidence
Finding: The CLI persists the full search output to a local JSON file using a filename derived from user input, but the user is not clearly warned that results will be stored on disk. In the context of a data-extraction skill, persisted logs can expose potentially sensitive search queries and returned content to other local users, backup systems, or later unintended processing.

VirusTotal

59/59 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.