build-mcp-use-server

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

• Third-party content exposure detected (high risk: 0.80). The skill's included references and recipes (e.g., references/examples/server-recipes.md Recipe 3's ghFetch calling the public GitHub API, Recipe 6's fetch to a weather API, the webhook handler that stores arbitrary incoming payloads, and references/guides/advanced-features.md describing server.proxy() with remote URLs) show the server code will fetch and ingest open/public, user-generated third‑party content and then read/interpret it as part of tool workflows, which can materially influence tool behavior and thus 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 contains server.proxy() examples that run npx to fetch and execute remote packages at runtime (e.g., "npx -y @modelcontextprotocol/server-filesystem" / "npx -y @modelcontextprotocol/server-github"), which clearly executes remote code during skill runtime and is a required dependency in those proxy examples.