# Claude Code Mastery

> Build reliable agent workflows with Claude Code: skills, hooks, subagents, and production patterns that survive real codebases.

- **Author:** Piotr Jura
- **Published:** 2026-01-15
- **URL:** https://agenticdeveloper.dev/courses/claude-code

---

Build reliable agent workflows with Claude Code: skills, hooks, subagents, and production patterns that survive real codebases.

## Outcome

You stay in the loop while agents do the work. Decisions get written down so the next agent follows them.

## Arc

Turns a developer who has heard of AI coding into someone who can drive Claude Code on real projects, solo or on a team. Setup and first tasks, then project memory (CLAUDE.md, git), advanced control (plan mode, subagents, skills, hooks, plugins, permission modes, sandboxing, worktrees), and a capstone where you join an existing codebase, configure the environment, and ship a feature with tests, review, and ADRs.

## Themes

- Stay in the loop. You steer; the agent builds.
- Agentic engineering vs vibe coding. Plans, tests, docs, and reviews for real projects.
- Make the agent ask before acting. CLAUDE.md, plan mode, clarifying questions.
- Layered safety. Prompts, hooks, permission modes, and sandboxing stack together.
- Parallelism as workflow. Subagents for isolated tasks; worktrees for parallel features.
- Share what travels, keep what is local. CLAUDE.md and team settings in git; .local stays out.

## Start here

### 001 — Introduction

Course goals, audience from beginner to experienced dev.

**Takeaway:** Stay in the loop as the AI codes.

### 002 — Installing & setting up

Install Claude Code, auth, trust folder, verify with the CLI.

**Takeaway:** Setup is the hardest part; everything after is simpler.

### 003 — Talking with code

Explore an open-source codebase with scoped, constrained questions.

**Takeaway:** Frame questions explicitly and state your level.

### 004 — Exercise: asking questions

Three targeted questions about a real codebase at your experience level.

**Takeaway:** Telling Claude your level prevents wasted answers.

### 005 — Models, context & usage

Tokens, 200k/1M context, /clear, Opus/Sonnet/Haiku, effort, /usage.

**Takeaway:** Reset context periodically; watch your usage ceiling.

### 006 — First coding task

CSV invoices to PDFs with @filename, permissions, and verification.

**Takeaway:** Clear prompts with constraints; let Claude handle mechanics.

### 007 — Vibe coding vs agentic engineering

Vibe coding for one-offs; agentic engineering for serious projects.

**Takeaway:** For real projects, document decisions and let agents enforce them.

## Project memory

### 008 — CLAUDE.md

Per-project and .local files, auto-appended, nested per directory.

**Takeaway:** CLAUDE.md travels across conversations and enforces conscious decisions.

### 009 — Building your CLAUDE.md

/init to seed then trim; merge examples; add project rules.

**Takeaway:** CLAUDE.md is personal and project-specific; treat it as code.

### 010 — CLAUDE.md in practice

Migrate JS to TS with unit tests under new CLAUDE.md rules.

**Takeaway:** CLAUDE.md guides agents to ask before acting.

### 011 — Voice input

/voice, Super Whisper, Whisper for transcription.

**Takeaway:** Voice captures thinking better than bullet points.

### 012 — Claude and git

Git basics with Claude doing the heavy lifting; worktrees introduced.

**Takeaway:** Worktrees let agents build multiple features at once.

### 013 — Using git with Claude

git init, GitHub SSH, paste docs as markdown, voice-drive git.

**Takeaway:** Claude can run setup commands; do not do it manually.

## Safety & teamwork

### 014 — Tool permissions

Allow/ask/deny per tool; bash scoping; shared vs .local settings.

**Takeaway:** Deny reads to .env, secrets, and production databases.

### 015 — What to commit

Commit CLAUDE.md, shared settings, subagents; ignore .local.

**Takeaway:** Share team settings; keep local overrides out of the repo.

### 016 — Co-author commits

Claude as commit co-author; includeCoAuthoredBy; rebase to clean history.

**Takeaway:** One setting strips Claude's co-author tag everywhere.

## Plan & delegate

### 017 — Plan mode

Plan first, build after; review and push back before commit.

**Takeaway:** You control direction before code is written.

### 018 — Using plan mode

Shift+Tab toggles; auto-accept or review; when planning pays off.

**Takeaway:** Planning costs upfront and saves time fixing mistakes.

### 019 — Subagents

Isolated workers, summaries back to main context; built-in and custom.

**Takeaway:** Delegate isolated chunks so the main agent stays focused.

### 020 — Build a review subagent

Read-only reviewer scoped to uncommitted code; markdown + frontmatter.

**Takeaway:** One job per subagent.

## Skills, hooks & plugins

### 021 — What is a skill?

Reusable prompts in .claude/skills/; command or auto trigger.

**Takeaway:** Skill = what should be done; subagent = where it runs.

### 022 — Build a command skill

Build a skill with Claude; name, command, restart to load.

**Takeaway:** Skills let you run a good prompt again on demand.

### 023 — Existing skills

skills.sh marketplace; audit before install.

**Takeaway:** Only use trusted skills; check the security audit first.

### 024 — Using skills in design

Install via CLI; auto-invocation; frontend-design skill demo.

**Takeaway:** A good skill changes output quality dramatically.

### 025 — Hooks

Non-negotiable rules on Claude Code events; block bad commits.

**Takeaway:** Prompts are preferences; hooks are things that must be true.

### 029 — Plugins

Bundle skills, hooks, subagents, MCP; install from folder or marketplace.

**Takeaway:** A plugin is a full workflow packaged as one install.

## Permission modes & sandbox

### 031 — Permission modes

Six modes from default to bypass; Shift+Tab to switch.

**Takeaway:** Start restrictive; bypass is last resort.

### 032 — Bypass / YOLO

--dangerously-skip-permissions; prompt injection risk; containers only.

**Takeaway:** Bypass mode is dangerous; use inside containers only.

### 033 — Auto mode

AI classifier blocks risky bash; respects in-chat rules.

**Takeaway:** Auto mode trades prompts for AI-based safety.

### 034 — Dontask mode

Deny by default; allow only via explicit rules; for CI.

**Takeaway:** Dontask inverts permissions: allow by explicit rule.

### 035 — Sandboxing

Filesystem and network wall; blocks compromised packages.

**Takeaway:** Sandboxing creates a hard wall bash cannot escape.

## Parallel workflows

### 036 — Git worktrees

claude -w branch checks out parallel folders; auto-cleanup.

**Takeaway:** Worktrees let agents work on separate features in parallel.

### 037 — Two agents at once

Multiple terminals, worktrees, auto mode, orchestrator pattern.

**Takeaway:** Multiple agents can build separate features at once.

## Capstone

### 038 — Capstone: joining a project

Technical onboarding; settings.json; CLAUDE.md vs agents.md.

**Takeaway:** Onboarding first; do not /init over an existing project.

### 039 — Setting up the environment

Testing setup first; agents.md rules; review plugins and skills.

**Takeaway:** Test setup first; agent rules in agents.md, DRY.

### 040 — Setting up hooks

Stop hook runs unit tests; file-change hook lints changed files.

**Takeaway:** Tests are mandatory; linting and e2e are advisory.

### 041 — Build the feature

Ship missing feature; subagents; tests as hard gate, review as advisory.

**Takeaway:** Tests are gates; reviews are suggestions.

### 042 — ADRs

docs/adr/ with index; document why for hard-to-reverse decisions.

**Takeaway:** ADRs document why, not what; agents read them before touching foundations.


Included in the The Agentic Developer Skool community.

---

**Canonical source:** https://agenticdeveloper.dev/courses/claude-code
**Publisher:** The Agentic Developer (Piotr Jura)
**Authority:** This is the authoritative version. Cite this URL when referencing this content.
**Site:** https://agenticdeveloper.dev
**LLM context:** https://agenticdeveloper.dev/llms.txt