Chapter 32 · Course Summary and Learning Roadmap

Chapter goals: Wire together the six modules of this course;画 the能力雷达图; clarify the后续 learning path (Agno / CrewAI / Pi course衔接); and answer frequently asked questions.

32.1 Six Modules Wired Together

This course围绕构建可靠的 AI 编码助手这一目标,覆盖了六个相互支撑的模块:

text

┌─────────────────────────────────────────────────────────────┐
│                  Agent Engineering Fundamentals              │
├──────────┬──────────┬──────────┬──────────┬──────────┬───────┤
│  Prompt  │ Harness  │   MCP    │  Skills  │   Loop   │AGENTS │
│ Engineering│Engineering│ Protocol │ Standard │Engineering│ .md   │
│  01–09   │  10–15   │  16–21   │  22–25   │  26–29   │30–31  │
└──────────┴──────────┴──────────┴──────────┴──────────┴───────┘
     ↓            ↓           ↓          ↓          ↓         ↓
  Write prompts  Build env   Connect tools  Package capabilities  Auto-run  Project docs

Core Takeaways Per Module

Module	Core Question	Key Output
Prompt Engineering	How to make the model understand andexecute the task?	Structured prompts, CoT, ReAct, RAG
Harness Engineering	Why do strong models still fail?	Five-subsystem model; diagnostic loop
MCP Protocol	How to let agents safely use external tools?	Tools/Resources/Prompts primitives; transport layer
Agent Skills	How to封装 reusable task capabilities?	SKILL.md; progressive disclosure
Loop Engineering	How to升级 from a single prompt to an自动 loop?	Five blocks; Loop Ready Score
AGENTS.md	How to write an agentinstruction manual for a project?	Standardized instruction file; nesting priority

32.2 Capability Radar Chart

After completing this course, you should be able to self-assess on the following dimensions:

                    Prompt Design ────────● 80%
                        /   \
                       /     \
             Harness Design      MCP Integration ────● 70%
                   |             |
                   |             |
             Skills Packaging ───● 75%
                   |             \
                   |              \
             Loop Design ──────────● 65%
                    \              /
                     \            /
                   AGENTS.md ────● 70%

● 80%+: can独立 design complex prompts; diagnose harness defects
● 70%+: can implement MCP servers; write high-quality skills; design basic loops
● 60%+: understand the原理; can完成实践 under guidance

32.3 Subsequent Learning Roadmap

This course covers the "principles and protocol layer." The suggested path to继续深入:

Path A: Framework实战

After completing this course,进入框架课程 to落地 theory:

Course	Focus	Connection to Prior Knowledge
Agno	Single agent + team + workflow	MCP tools, Skills packaging, Loop orchestration
CrewAI	Multi-agent collaboration	Skills auto-trigger, MCP integration, loop patterns
Pi	Terminal coding agent customization	AGENTS.md, Skills, Hooks

Path B: Testing and Verification

Strengthen reliability配合测试课程:

Course	Focus
pytest	Unit tests, fixtures, plugin development
Playwright	E2E tests, network mocks, auth reuse

Path C: Web Services

Build the agent's后端支撑:

Course	Focus
FastAPI	API development, security auth, database integration

32.4 Frequently Asked Questions

Q: I've finished the course; which framework should I use next? A: Depends on the goal:

Rapid prototype → Pi
Multi-agent collaboration → CrewAI
Deep customization + production deployment → Agno

Q: What's the difference between Prompt Engineering and Harness Engineering? A: Prompt solves "how to说 it"; Harness solves "where to say it, how to说 it, how to验 it." The former is a subset of the latter.

Q: What is the relationship between MCP and Skills? A: MCP provides通用 tool capabilities (search, database); Skills封装特定 task workflows. The two can叠加使用.

Q: Is Loop Engineering suitable for所有 projects? A: Not for greenfield projects (starting从零). It's最适合 for持续 maintenance scenarios with an existing codebase.

Q: Can I go to production directly after finishing this course? A: It's建议 to配合 Playwright and FastAPI courses,补足 testing and backend capabilities before deploying.

Chapter Summary

The six modules cover the完整链路 from prompt design to automatic loops;
The capability radar chart helps you定位 the current level and提升 direction;
Subsequently可进入框架课程 (Agno/CrewAI/Pi) or测试课程 (pytest/Playwright) to深化;
Remember the core principle: when something fails, check the harness first, then swap the model.

Back to Course Guide

Chapter 32 · Course Summary and Learning Roadmap ​

32.1 Six Modules Wired Together ​

Core Takeaways Per Module ​

32.2 Capability Radar Chart ​

32.3 Subsequent Learning Roadmap ​

Path A: Framework实战 ​

Path B: Testing and Verification ​

Path C: Web Services ​

32.4 Frequently Asked Questions ​

Chapter Summary ​