web-scraping

Fail

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: HIGHPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: Use of homoglyph characters in code identifiers. The class definition TrafilaturaСscraper and its subsequent instantiation in the ScrapingCascade class utilize a Cyrillic 'С' (U+0421) instead of the Latin 'C' (U+0043). This technique is commonly used to evade text-based security filters and signature detection.
  • [EXTERNAL_DOWNLOADS]: The skill utilizes several well-known third-party libraries for web scraping and automation, including requests, trafilatura, beautifulsoup4, playwright, playwright-stealth, yt-dlp, instaloader, and fake-useragent. These are standard tools for data extraction.
  • [PROMPT_INJECTION]: Potential surface for indirect prompt injection due to processing untrusted data from external websites.
  • Ingestion points: The skill fetches content from arbitrary URLs through multiple scraping classes (e.g., TrafilaturaСscraper.fetch and RequestsScraper.fetch).
  • Boundary markers: The instructions do not implement explicit delimiters or boundary markers to separate scraped content from the agent's instructions.
  • Capability inventory: The skill includes capabilities for network operations (requests.get, playwright.goto) and file system writes (yt_dlp for video downloads, instaloader for media).
  • Sanitization: The RequestsScraper class uses BeautifulSoup to remove potentially dangerous elements like script and style, and the trafilatura library provides high-quality text extraction that filters out non-content elements.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 20, 2026, 12:20 PM
Security Audit — agent-trust-hub — web-scraping