Before you write a single line of agent code, run this. It walks through four decision points to determine whether your project needs LangChain's basic agent loop, LangGraph's custom control flow, or Deep Agents' full orchestration layer with built-in planning and file management. The framework choice dictates everything else, from which other skills to load to how you structure your codebase. I've seen too many projects start with the wrong foundation and hit walls later when they need persistent state or subagent delegation that their chosen framework can't handle cleanly.
npx skills add https://github.com/langchain-ai/langchain-skills --skill framework-selection