Loops, Not Prompts: Agentic Engineering with Claude Code

Most people use AI coding tools like a smarter autocomplete: type a prompt, read the answer, fix what's wrong, repeat. So you end up as the only quality gate. This course flips it. You'll set up Claude Code so it can check its own work, follow rules that can't be skipped, plan before it builds, split work across subagents and workflows, and run unattended with a budget. Six hands-on exercises, each with a « done when » check and a rubric, plus a final project, all in a real repository.

New

Offered by ITwithDom

Last updated November 2026
Language: English
2 h 42
E-learning

What you'll learn

Turn a one-shot prompt into a loop: change → check → read the output → fix → pass
Decide what belongs in CLAUDE.md (guidance) and what belongs in hooks and deny rules (enforcement)
Use plan mode and a spec, then review the diff against that spec in a fresh subagent
Split work across custom subagents, and know when a dynamic workflow is the better tool
Run Claude Code unattended with claude -p, budgets, /goal and scheduled tasks, and keep a human on the merge

Practical information

Format: E-learning
Location: Online
Schedule: Self-paced online, about 2 h 40 including a 60-minute final project. Recommended: one page per sitting, doing each hands-on step right away.

Course content

8 lessons, 2 h 42 in total

Why loops beat prompts10 min
Hand Claude a pass/fail signal15 min
Advice vs. guard rails: CLAUDE.md, hooks and deny rules18 min
Plan, then build, then review against the spec14 min
Graphs, not chats: subagents16 min
Dynamic workflows: when the plan becomes code16 min
Running without you: headless, budgets, schedules13 min
Final project: a repository that checks itself1 h

Description

A hands-on course on working with Claude Code as an engineering loop instead of a chat: checks Claude can run, rules the harness enforces, plans before code, subagents and dynamic workflows, and unattended runs with budgets.

Key information

Learning objectives: Configure a repository so Claude Code can verify its own changes with a pass/fail check
Write a Stop hook, a formatter hook and deny rules, and explain why each one is deterministic
Write a custom subagent and a reusable skill, and choose between subagents, skills and workflows
Run Claude Code non-interactively with a turn or budget cap and structured output
Who it is for: Developers, DevOps and platform engineers, tech leads who already tried Claude Code (or a similar coding agent) and want reliable results instead of lucky ones.
Prerequisites: Comfortable with git and a terminal. Node.js 22.6+ (24 LTS recommended) with npm, jq and a bash shell (macOS, Linux, WSL or Git Bash). Claude Code 2.1.295 or later on any paid plan: every exercise works with your subscription login. Optional: an Anthropic API key (platform.claude.com) if you want to try --bare on pages 7–8.
Duration: 2 h 42
Price: Free
Teaching format: E-learning
Access conditions and lead time: Free. Create a learner account on the course page and start immediately. Immediate access after enrolment.
Teaching methods: Short explanations, diagrams and an animated walkthrough. Six hands-on exercises, each with a « done when » check and a rubric, plus a final project. Starter kit and cheat sheet included.
Assessment: Checkpoint quiz on every page (70 % to continue), Six hands-on exercises, each with a « done when » check and a rubric, plus a final project, Final project graded against a rubric, reference solution provided
Accessibility for people with disabilities: Self-paced, no time limit. Every diagram has a text title and is explained in the text around it; the animation has captions; all code is provided as text. Questions: contact the instructor through the course page.
Contact: Dominik Heinrich (ITwithDom)