A Skeleton Course for Building Agents With Your AI Partner

What it does

This repo is a 22-chapter skeleton course on designing, building, and operating production AI agents. Rather than prescribing a specific framework or walking through a single project, it provides architectural load-bearing concepts—tool loops, memory systems, multi-agent delegation, observability, safety—meant to be read with an AI coding partner at your side. The goal is to ship your own project while understanding every line of it.

The interesting bit

The material is explicitly written for dual consumption: humans read the durable skeleton, while your AI partner “puts the muscles on it” by generating examples, quizzes, and implementation advice tailored to your preferred stack. There is even a built-in agentic-system-reviewer skill that lets your agent audit your designs against the course. It is essentially a meta-curriculum that tries to defeat framework rot by offloading perishable details to the AI.

Key highlights

• 22 chapters progressing from single tool calls to proactive, self-evolving agents
• Framework-agnostic by design; never mandates LangChain, Pydantic AI, or similar
• Includes AGENTS.md and CLAUDE.md behavioral guides so your AI assistant knows how to teach the material
• Built-in agentic-system-reviewer skill for course-grounded design reviews
• Optional reference systems (OpenCode, Hermes Agent, etc.) for grounded code examples

Verdict

Developers who already use an AI coding assistant and want to move beyond one-off prompts into durable agentic systems will get the most from this. If you are looking for a traditional, finished-project tutorial, the skeleton format will feel abstract.