The complete guide to how AI agents discover and orchestrate Skills, Tools, and Protocols — with practical patterns for Claude Code, Cursor, and Cline.
If you just need to operate an agent, harness engineering (the implementation patterns of Agent Engineering / Context Engineering) is enough. But in AI-driven development, you also need to design, maintain, extend, and hand off agents.
This site is not a "how to operate" manual — it's a map for design. How to compose Skills, MCP, Sub-agents, and Doctrine; what to write as MUST vs SHOULD; how to make components reusable. The goal is structuring the entire development process, not automating one-off tasks.
| Verb | Goal | Primary Reference |
|---|---|---|
| Operate | Complete today's task | Harness engineering frameworks |
| Design | Build reusable structures and judgment criteria | 👈 This site (ai-agent-architecture) |
| Understand | Grasp the structural constraints of LLMs | understanding-llm-through-claude-code |
💡 For readers who came here searching for "harness engineering" — Harness is the mechanism for operating; this site is the map for designing. For the mapping between the two and the layers harness doesn't cover (Skills layer / Doctrine layer), see Harness Engineering Mapping.
A 3-phase learning path: "Know LLMs → Know Agent Design → Apply to Systems."
| Phase | Project | Focus |
|---|---|---|
| 1. Know LLMs | understanding-llm-through-claude-code | LLM structural constraints and the why behind configuration design |
| 2. Know Agent Design | 👈 This site | MCP, Skills, and Agent composition with implementation patterns (what/how) |
| 3. Apply to Systems | Management-of-software-systems-and-services | Coming soon — System operations in the AI era |
💡 For readers who learned "what are Skills?" or "Skills vs MCP" here — if you want to understand why the Skills design is necessary from LLM structural constraints, read understanding-llm / Part 5: On-Demand Context alongside this site.
Note: This documentation reflects the author's practical insights gained through building and operating AI agent systems with Claude. It is not official documentation from Anthropic or any other organization. Contributions and discussions via GitHub Issues are welcome.