add-integration

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow includes “Search for existing MCP package” by querying the npm registry and known MCP server directories at runtime, which can ingest outsider-authored free-form text (e.g., package descriptions/readmes) into the agent’s LLM context during evaluation/presentation.

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

• Potentially malicious external URL detected (high risk: 0.90). This skill explicitly queries and may install/run pre-built MCP packages from the npm registry (https://registry.npmjs.org) and configures .mcp.json to use npx to start those packages during connectivity testing, which causes fetched remote code to be executed at runtime.