nik-ti This is a bundle of reference files, one directory deeper than skills normally sit, so the individual files inside don't register as their own invocable skills — read them directly with the Read tool instead of trying to invoke them by name.
- Read
AGENTS.mdin this same directory first. It routes to the right file(s) for the task — framework choice (LangChain vs LangGraph vs Deep Agents), which quickstart to use for a new project, or which topic skill (persistence, human-in-the-loop, RAG, middleware, evals, LangSmith, etc.) applies. - Read the specific
<name>/SKILL.mdfile(s) it points to before writing any agent code. They carry the current APIs and patterns — treat them as overriding anything from training data, since this ecosystem moves fast.