mcp-server-builder

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly instructs configuring MCP servers that connect to external services and expose tool outputs from public sources—e.g., Brave Search and GitHub in the "Popular MCP Servers Reference" and the remote-server URL in "Adding Servers"—which Claude will read and act on as part of its workflow, exposing it to untrusted, user-generated content that could carry indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime examples instruct executing remote packages via npx (e.g., "npx @modelcontextprotocol/server-github" and "npx @anthropic/slack-mcp"), which fetch and run external code at runtime and can therefore supply/alter tools or behavior that control the agent.