connect-glean

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

• Third-party content exposure detected (high risk: 0.75). The required workflow configures and authenticates a remote HTTP MCP server at a user-supplied Glean URL (e.g., <server-url>/mcp/<server-name>), so at runtime the agent will read outsider-authored free text from that remote server’s tool responses into the LLM context (indirect prompt injection risk).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires registering and calling remote MCP endpoints at runtime (e.g., /mcp/ and https://developers.glean.com/mcp), which the agent will invoke to access remote "tools" that can control prompts or execute actions.