bright-data-mcp

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS] (SAFE): The skill directs users to download and install the @brightdata/mcp package from npm. Bright Data is a recognized provider, and this is a standard installation procedure for an MCP server.
  • [COMMAND_EXECUTION] (LOW): The skill provides command-line instructions for installing and running the server locally (npm install -g, npx @brightdata/mcp). These commands are necessary for the skill's primary purpose but involve local execution. No high-risk command injection patterns were found.
  • [DATA_EXPOSURE & EXFILTRATION] (SAFE): The skill handles API tokens for the Bright Data service. It correctly uses placeholders (<YOUR_BRIGHTDATA_API_TOKEN>, YOUR_TOKEN) rather than hardcoding credentials. The network operations (connecting to mcp.brightdata.com) are the primary function of the skill.
  • [INDIRECT PROMPT INJECTION] (LOW): The skill implements various scraping and extraction tools (e.g., scrape_as_markdown, extract, web_data_*). These tools ingest untrusted content from the web and present it to the agent, creating a surface for indirect prompt injection.
  • Ingestion points: Tools like scrape_as_markdown and scraping_browser_get_text in references/mcp-tools.md.
  • Boundary markers: Not explicitly mentioned in the provided reference files.
  • Capability inventory: The skill allows for browser automation (scraping_browser_*) and structured data extraction.
  • Sanitization: Not specified; standard LLM guardrails are expected to apply to processed data.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:19 PM