Claude Code in Action
By the end of this week you will have Claude Code wired into your real project: a CLAUDE.md that holds, a custom slash command you use daily, a hook that catches your own mistakes, and a PR review bot that runs on every push. 4 modules, 21 lessons, all free.
Taught by Edwin Ashdown. Founder of AIFirstPartner, 12 production AI systems shipped.
More background on the instructor on the about page.
Four real scenarios covered end to end in the free first module. Each is a different shape of work; the common thread is tool use.
Chalk optimisation
3.9x throughput on a package pulling 429M weekly downloads
Profile, benchmark, hypothesise, re-measure. A closed optimisation loop, driven end to end by tool calls.
Churn analysis in Jupyter
An iterative notebook session, cell by cell
Run a cell, read the result, write the next cell based on what came back. Data-scientist workflow, not a single pre-written script.
UI styling via Playwright MCP
A browser-driven screenshot-evaluate-update loop
A Playwright MCP server gave Claude a real browser. It took screenshots, compared styles, updated prompts, and iterated.
PR review that caught a PII leak
A Terraform-aware review that a linter could not have flagged
Adding an email field to a Lambda looked harmless. Claude followed the data through DynamoDB streams into a shared S3 bucket and flagged the exposure before merge.
This is for developers who already use a terminal, a git workflow, and at least one codebase they care about not breaking. If you have opened Claude Code once, got something useful, got something weird, and closed the tab, this is the course that turns that gamble into a reliable daily tool.
This is not for people looking for an intro to AI coding assistants in general (Claude Code 101 is better for that), and not for people who only want to use Claude Code through Cursor or another IDE. This course is the CLI, hooks, and GitHub integration.
The gather-plan-act loop, the tool use bridge that makes it work, and four demos that show it in the wild: a 3.9x speedup on the Chalk library, a Jupyter churn analysis, a Playwright-driven styling loop, and a Terraform-aware PR review that caught a PII leak before it shipped.
4 lessons
Context management that actually holds, slash commands written in a markdown file, MCP servers that let Claude drive a browser or a database, and a GitHub Actions integration that turns every PR into a reviewed one. Ten lessons, each paired with a working example.
10 lessons
Hooks run before or after Claude uses a tool. That simple rule lets you block it from reading .env, auto-run tests after edits, catch missed call sites with tsc, and spawn a second Claude to catch duplicate queries. Eight lessons, including an exercise where you build a working .env blocker end to end.
8 lessons
An eight-question quiz to test what you retained, plus a guided prompt for picking the one workflow you commit to automating this week.
2 lessons
Edwin Ashdown founded AIFirstPartner and has shipped 12 production AI systems for clients in government, legal, and enterprise. He is fractional CTO at Speculo, a cyber risk software company, and manages cloud and data platform engineering for a large Australian government department supporting 20+ agencies across the portfolio.
He publishes his build process at 150days.aifirstpartner.com. This course is the playbook he uses when onboarding a team onto Claude Code.
No. Claude Code in Action is independent. This course is not an Anthropic product and has not been reviewed or endorsed by them.
The docs are a reference. They tell you what hooks are; they do not walk you through building the .env blocker on Monday, catching the common mistakes, and then stealing two production-ready hook patterns for Tuesday. This course is the path that gets you from "I have read about hooks" to "I have one running on my repo" in an afternoon.
A CLAUDE.md that actually works on your project, at least one custom slash command you use daily, a working .env hook, a tsc auto-fix hook, and a GitHub Actions setup that reviews every PR. You can either build these on the provided sample project or on your own codebase.
Parts of it will. Claude Code ships features monthly. The gather-plan-act mental model, the tool use framing, the PreToolUse and PostToolUse asymmetry, the feedback-loop pattern: that stuff outlasts specific commands. When a new feature lands, the framework Module 4 gives you for evaluating it still works.
No. The course starts with fundamentals: how coding assistants work and what makes Claude Code different. You should be comfortable with a terminal, a code editor, and basic development workflows.
The course uses JavaScript and TypeScript for examples and the sample project. The concepts (context management, hooks, GitHub integration) apply to any language or stack you use with Claude Code.
Yes. The sample UI generation project is provided for convenience, but the course explicitly encourages following along with your own codebase. The techniques are project-agnostic.
Claude Code 101 is a broader introduction covering the Explore-Plan-Code-Commit workflow. Claude Code in Action goes deeper into practical integration: custom commands, MCP servers, GitHub Actions, hooks, and the SDK.