Building Threat Models with MCP and AI Agents | Dark Hacker News