build-mcp-use-agent

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to fetch and interpret open/public web content (e.g., "Recipe 4 — Browser Automation Agent" which navigates to TARGET_URL like https://news.ycombinator.com and extracts page title/headlines, and dynamic-server recipes that add web-search servers such as the Brave search server), meaning untrusted third-party page content is read and used to drive tool calls and follow-up actions, which could enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill repeatedly invokes npx to fetch and run external MCP server packages at runtime (e.g., npx -y @modelcontextprotocol/server-everything), which downloads and executes remote code and exposes tools/prompts that directly control the agent's behavior, and those servers are required for the agent flows—this meets the criteria for a runtime-executed external dependency.