Security audit

抖音爆款爬虫

Security checks across malware telemetry and agentic risk

Overview

This skill looks like a Douyin scraper, but its main scripts return invented sample data while presenting the output as scraped results.

Review before installing. Treat this as a demo or mock scraper unless the publisher implements real extraction and clearly labels simulated outputs. Do not rely on its JSON/CSV results for business or research decisions, and avoid saving or sharing creator/content metadata without checking platform rules, privacy obligations, and local laws.

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 (6)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 87% confidence
Finding: The skill instructs the agent to invoke a local Python script with shell arguments and optionally write output files, but it declares no permissions despite requiring shell, environment, and file-write capabilities. This creates a capability/visibility mismatch: operators and policy systems may not realize the skill can execute commands and persist data, increasing the chance of unsafe execution or abuse through crafted inputs.

Description-Behavior Mismatch

Medium

Confidence: 98% confidence
Finding: The skill claims to scrape Douyin search results but generates fabricated placeholder records instead of extracting real page content. This is dangerous because downstream users or agents may treat the output as genuine external data, leading to misinformation, bad decisions, or silent integrity failures in automated workflows.

Description-Behavior Mismatch

Medium

Confidence: 98% confidence
Finding: The hot-list function similarly advertises live trending retrieval but returns invented sample entries. In an agent skill, this creates a data provenance problem: consumers may trust fake rankings as real-time trend intelligence, which can mislead analysis or business actions.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill advertises Douyin scraping but always returns fabricated mock data regardless of whether browser access succeeds. In an agent setting, this can mislead downstream users or automated workflows into trusting invented results as real platform data, causing integrity failures, bad decisions, or silent data poisoning.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The README explicitly documents scraping Douyin content and saving results to local JSON/CSV files, but it does not clearly warn users about collecting potentially personal data, retention, sharing, or compliance obligations. In a scraping skill, this omission increases the risk that users or downstream agents will gather and persist creator metadata without informed handling safeguards, leading to privacy and policy violations.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The trigger description is broad enough to match many ordinary requests about Douyin content, which can cause the skill to fire in contexts the user did not clearly intend. Because the skill launches a scraper script, overbroad triggering can lead to unnecessary external access, command execution, or data collection from ambiguous prompts.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.