scrapingant-web-fetch

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow exposes an LLM tool (get_web_page_markdown via ScrapingAnt’s hosted MCP at https://api.scrapingant.com/mcp) that ingests outsider-authored text when the user/agent supplies arbitrary URLs to fetch and the fetched page content is converted to Markdown at runtime.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill registers and calls ScrapingAnt MCP/REST endpoints at runtime (e.g., https://api.scrapingant.com/mcp and https://api.scrapingant.com/v2/markdown), which return fetched page content (Markdown/HTML) that is injected into agent prompts, so external content directly controls the agent's prompt context.