cloudflare-mcp-server

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly instructs connecting to arbitrary external MCP servers and OAuth providers (see SKILL.md's "await manager.connect('https://external-mcp.com/sse')" which "auto-discovers tools, resources, prompts", and the GenericOAuthHandler example that fetches "https://provider.com/oauth/userinfo"), meaning the agent is expected to ingest and interpret untrusted third-party content which can change available tools/prompts and thus influence actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill contains an explicit runtime example that connects to an external MCP server which "auto-discovers tools, resources, prompts" — e.g. manager.connect("https://external-mcp.com/sse") — meaning that fetching from that URL at runtime can inject prompt/tool definitions that directly control agent behavior.