mcp-builder

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

• Third-party content exposure detected (high risk: 1.00). The skill's Phase 1.2 and 1.3 (and the final Notes) explicitly instruct the agent to fetch and load public web pages and raw GitHub files (e.g., https://modelcontextprotocol.io/sitemap.xml, https://raw.githubusercontent.com/modelcontextprotocol/..., and https://github.com/anthropics/skills/tree/main/skills/mcp-builder/reference), meaning it ingests untrusted, user-controlled third-party content that can materially influence tool design and agent actions.