Claude Code Harness & Graph Engineering: Advanced Claude Code CLI Harness Engineering in Practice

Claude Code can be fun in a GUI.

However, Claude Code pays off in CLI harness design.

This course does not introduce Claude Code CLI as merely an AI coding tool. Instead, it covers how to establish Claude Code within real backend and DevOps workflows through the lenses of Context, Permission, Verification, Tool, Subagent, and Debugging Harness. By connecting log analysis, bug fixes, test recovery, CLAUDE.md design, session management, and parallel worktree operations, you will learn how to design a work system that enables Claude to work safely on real projects.

Recommended for those who:

After taking the course,

After taking this course, you will no longer see Claude Code as merely a “tool for entering good prompts.” Instead, you will be able to determine what Claude should see, how far it is allowed to act, how to verify the results, when to clear/compact/restart if a session becomes polluted, and how to manage worktrees and parallel sessions.

Specifically, you will be able to structurally codify project rules and prohibitions in CLAUDE.md, organize logs, environment variables, and configuration scope issues into a form that Claude can solve, and review Claude’s proposed changes through testing, diff, and counterexample validation loops. In other words, you will take Claude Code beyond being a GUI-based assistant and bring it to a level where it can be integrated into actual backend and DevOps workflows.

Features of this course

1. Rather than explaining features, it teaches from a “harness engineering” perspective based on research papers.

This course does not list Claude Code’s features like official documentation. Instead, it explains how to design Context, Permission, Verification, Tool, Subagent, and Debugging Harnesses so that Claude Code can work safely in real projects. As a result, you’ll be less affected by rapidly changing version information and retain practical principles that can be applied to real projects.

2. It is based on large codebases and real-world work problems

Rather than simply building demo apps, this course covers how to keep Claude from getting lost in large codebases, reduce session contamination, safely manage worktrees and parallel sessions, and connect the GitHub Issue → PR workflow with Claude. It starts with the points where people actually get stuck in real-world work.

What you'll learn

Getting Started with Claude Code CLI in Practice /

Architecture / Memory / CLAUDE.md / Plan Mode / Skills

In the first section, we clearly lay out the differences between GUI-based Claude and the CLI, starting with the basics. Rather than simply comparing interface differences, we explain why GUI-based workflows tend to stop at “asking questions and receiving answers,” while the CLI can connect all the way to actual logs, tests, Git, and deployment preparation. We then use Claude Code architecture to understand why the harness is more important than the model, and explore the role of memory across sessions and CLAUDE.md, as well as the differences between Plan/Thinking Mode and Subagents, and between Skills and Tool use.

Harness configuration file to copy and use immediately

The harness configuration you create directly in the course is provided in full as completed files.
After taking the course, you can copy them into your own project and apply them immediately without having to write everything again from scratch.

The provided settings are as follows.

• CLAUDE.md
  Defines the work rules Claude must follow in the project.

• codecraft-harness/SKILL.md
  Transforms a short, ambiguous one-sentence request into a concrete design and implementation plan.

• solid-review/SKILL.md
  Before starting implementation, check whether responsibilities are properly separated and whether the design is unnecessarily complex.

• block-sensitive-paths.sh
  Blocks access to sensitive files and paths, such as .env, SSH keys, and parent project folders.

• run-tests-after-edit.sh
  Automatically runs the relevant tests every time Claude modifies a file.

• block-stop-without-tests.sh
  It prevents Claude from reporting a task as complete when tests have not been run or have failed.

• guard.sh
  It blocks even unauthorized external network requests to create a secure execution environment.

The Hooks included in the course are not merely sample code.
We ran them all in real environments and verified that they work.

Once you register the configuration file, tests run automatically every time Claude modifies the code, and when the tests fail, it cannot arbitrarily end by saying, “I’ve completed the task.”

In other words, you will learn how to go beyond simply asking Claude to verify through prompts and make the system forcibly execute the verification process.

3. Model-Specific Prompt Playbook

Even when using the same harness, each model differs in which instructions it follows well and which it is likely to miss.

In this course, instead of using the same prompt for every model, we provide five copy-ready prompt blocks that outline which instructions to remove and which to add based on each model’s characteristics.

• Opus 5
  Reduces verification instructions that the model performs on its own to prevent unnecessary duplicate work.

• Opus 4.8
  Clearly specify the tool execution conditions so that the necessary tools are used at the right time.

• Sonnet 5
  This section covers how to describe the files that need to be modified and the scope of the changes in specific, literal terms.

• Haiku 4.5
  It lays out the execution procedure step by step so you can work reliably even with models that do not have an effort option.

• Fable 5
  We provide, for reference, an approach that avoids giving the model too many instructions and retains only the necessary prescriptions.

Each playbook also includes the following content.

• Parameter combinations that cause a 400 Error when requested

• Default parameters and behavioral differences by model

• A record table comparing results while changing the effort value

• Run the same prompt on multiple models and compare the results in the Lab 04 exercise

Rather than simply explaining “which model is better,” you will learn how to find the optimal settings for each model yourself by repeatedly running the same task and recording the results.

Graph Engineering Practice Materials — Lab 05

You will directly compare the approach of handling complex tasks sequentially in a single pass with the Graph Engineering approach of dividing and processing multiple tasks according to their dependencies.

Interactive Learning Sheet

You can use it by opening the HTML file in a browser, without any separate installation or server setup.

You can play back step by step how the same request is processed differently in the following two approaches.

• Repeating a single flow with loop-based execution

• Graph-based execution that divides a task into multiple nodes and dependencies

When you select each step, you can also view the configuration file that created the corresponding action, allowing you to immediately understand the relationship between the execution result and the configuration.

Step-by-step learning guide

We provide a learning guide totaling 645 lines.

Rather than explaining the theory at length from the beginning, it is organized in the following order.

Check the conclusion → 30-minute practice → core concepts → advanced theory

There are also separate learning paths for non-majors and for students with development experience.

Even if you’re using Claude Code for the first time through vibe coding, you can start with hands-on practice, while developers can delve deeper into its internal workings and design principles.

3 demo projects you can run yourself

Rather than simply reading configuration files, you will run claude in an actual project folder and see firsthand how Hooks block and validate tasks.

Each project intentionally includes the following issue.

• Type error

• Test failure

• Logic bugs that appear normal on the surface

Students will directly experience the following workflow.

Detect the issue → Block the task → Modify the code → Run the tests → Pass validation

Loop-based and graph-based projects use the same code, with only the harness configurations set up differently.

Therefore, you can accurately compare how results differ due not to code differences, but to differences in how the work is organized.

Reference Documentation

We also provide practical documentation that you can refer to when configuring projects or solving problems even after completing the course.

• Environment Setup Guide
  This section covers preparing the execution environment, how to register Hooks, and common errors and their solutions.

• Distinguishing the Roles of Skills / Hooks / CLAUDE.md / Tests
  Explains what each component is responsible for and which tool to use for which problem.

• Harness Inspection Checklist
  You can check that nothing is missing from the prompt, design, implementation, tests, or completion criteria.

• Question and test cards
  Provides questions to use when requirements are unclear, test design criteria, and verification table templates.

• Model-Specific Playbook Documentation
  Organize each model’s parameters, behavioral characteristics, and prompt-writing guidelines in one place.

• Model evaluation log
  You can run the same task with multiple models and configurations, then compare and record the results.

These materials are not merely course supplements; they comprise a Claude Code harness operations manual that you can reuse repeatedly in real-world projects.

The person who created this course

• Silicon Valley Survivor | American Snail 

  Drawing on experience and expertise gained at the forefront of the Global Tech Scene, this course presents a path for non-majors to overcome technical barriers and become the owners of their businesses.

  • Current) Founder of a Silicon Valley AI coding agent startup

    • Operating the self-developed AI tool 'Snailer CLI' (18K+ downloads)

    • Selected for the Google for Startups Program

  • Former engineer at U.S. big tech companies and promising startups

    • Gave up the final stage at Amazon to start a company

    • Silicon Valley AI Fintech Startup Engineer

    • OpenAI / Meta / Apple / Adobe / Amazon Full-Stack Fellowship

    • Domestic search engine portal, fintech development

    • AI startup AR/B2B/SDK development

  • Proven teaching expertise

    • Double major in Computer Science and Business Administration at a four-year university in Seoul, with extensive startup experience

    • Produced 1,900+ students cumulatively; 4.5K+ followers on SNS Threads and 480+ Substack followers.

    • Threads: https://www.threads.com/@rich_l_2024?igshid=NTc4MTIwNjQ2YQ==

    • Substack: https://open.substack.com/pub/1234373801

Do you have any questions?

Notes Before Taking the Course