抖音搜索爬虫

• Outcome: Pass
• Latest audit: May 31, 2026 · 6:42 AM UTC
• Version: 1.2.0

Overview

This is a disclosed Douyin scraping helper that installs browser automation tools, browses Douyin, and can save results locally, with no evidence of hidden credential access, exfiltration, or destructive behavior.

Install only if you are comfortable with a tool that installs Playwright/Chromium, opens automated browser sessions to Douyin, and may save scraped metadata to local files. Keep limits and delays conservative, avoid logging into Douyin in the automated browser, choose output paths carefully, and review applicable laws, platform terms, copyright, privacy, and retention obligations before collecting or sharing results.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 91% confidence
• Finding: The skill declares no permissions while clearly directing shell execution, browser/network activity via Playwright, environment use, and file output. This creates a capability transparency gap: an agent or reviewer may treat the skill as low-risk even though it can execute commands and write data, increasing the chance of unintended execution or abuse.

Missing User Warnings

Medium

• Confidence: 83% confidence
• Finding: The README explicitly promotes scraping Douyin videos, metadata, and captions and saving them to local JSON/CSV files, but only gives generic compliance language and does not meaningfully warn about privacy, copyright, retention, consent, or secure handling of collected third-party data. In a scraping skill, this omission increases the likelihood that users will collect, store, and reuse personal or platform-governed data in unsafe or non-compliant ways.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The natural-language matching rules are broad enough that common phrases like '看看热榜' or '搜一下' could trigger scraping actions without clear user intent boundaries. In a skill that launches browser automation and may save files, overbroad triggers increase the risk of unintended network activity, data collection, and side effects from ambiguous requests.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The skill explicitly tells the agent to execute scraping commands directly without confirmation, even though the commands initiate browser automation, external network access, and optional file writes. This removes an important safety checkpoint and can cause unintended scraping, policy violations, or writing results to disk based on a loosely matched natural-language request.

Vague Triggers

Medium

• Confidence: 83% confidence
• Finding: The examples encourage the agent to accept broad natural-language prompts and automatically translate them into scraper actions without documenting scope limits, consent checks, or restricted targets. In a scraping skill, this can lead to unintended collection requests, overbroad automation, or misuse because users and downstream agents are not given explicit boundaries on what should or should not be executed.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.