fetch-url

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses uv to manage dependencies including playwright, trafilatura, rich, and typer from official package registries. It also provides instructions to download the Chromium browser via the official Playwright tool.
  • [COMMAND_EXECUTION]: The script executes local browser binaries for rendering web pages. It uses a predefined list of common installation paths for browsers like Chrome, Edge, and Brave. This is a standard optimization to avoid unnecessary downloads.
  • [DATA_EXFILTRATION]: The skill performs network requests (page.goto) to fetch content from user-provided URLs. This is the primary function of the skill. There is no evidence of unauthorized data transmission or credential harvesting.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web content which could contain malicious instructions. However, the purpose is content extraction into structured formats (Markdown, JSON, etc.), and it does not automatically execute instructions found within the fetched data.
  • Ingestion points: scripts/fetch_url.py reads data from remote URLs.
  • Boundary markers: None identified in the script itself; output is typically passed back to the agent context.
  • Capability inventory: Local file writes via --output flag, network access for fetching URLs.
  • Sanitization: Uses trafilatura for denoising and content extraction, which removes most non-content elements like scripts.
  • [DATA_EXPOSURE]: The skill allows writing extracted content to a local file path specified by the user via the --output parameter.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 07:15 AM
Security Audit — agent-trust-hub — fetch-url