Learn Claude Code

Claude Code Skills

Our own skills track for working with Claude Code day to day: the CLI, configuration, extension points, automation and the API basics. It is not an Anthropic exam; it pairs with the lessons on this site.

Take the 30-question mock examPractise by domain
At a glance

Five practical domains covering everyday Claude Code usage through to the API, with 100 questions set in four scenarios and some select-TWO questions. The mock exam uses the same 720/1000 benchmark so you can compare progress across tracks.

The domains and their weights

Claude Code Foundations & CLI25%
Configuration & Memory20%
Extending Claude Code25%
Automation, CI & Agent SDK15%
Claude API Fundamentals15%

Claude Code Foundations & CLI 25%

  • Install, run claude in your project folder, and describe what you want in your own words.
  • Use @ to reference files, ! to run a shell command, and Esc to interrupt (press it twice to rewind).
  • Resume work with claude -c (latest) or claude -r (pick a session).
  • Shift+Tab cycles permission modes, including plan mode.

Practise 25 Claude Code Foundations & CLI questions →

Configuration & Memory 20%

  • CLAUDE.md levels: user (~/.claude/CLAUDE.md), project (committed) and sub-directory.
  • settings.json: user, project (shared) and .local (personal, git-ignored).
  • Permission rules allow or deny specific tools and commands.
  • Use /memory to see and edit what is loaded.

Practise 20 Configuration & Memory questions →

Extending Claude Code 25%

  • Custom slash commands are Markdown files in .claude/commands/.
  • Skills live in .claude/skills/<name>/SKILL.md and load when relevant.
  • Subagents are defined in .claude/agents/ with their own prompt and tools.
  • Hooks (PreToolUse, PostToolUse, Stop and more) run scripts deterministically; exit code 2 from PreToolUse blocks the action.
  • Add MCP servers with claude mcp add or a shared .mcp.json; bundle it all as a plugin.

Practise 25 Extending Claude Code questions →

Automation, CI & Agent SDK 15%

  • claude -p "..." runs headless for scripts and CI.
  • --output-format json or stream-json makes output machine-readable.
  • The Claude Code GitHub Action brings Claude into PRs and issues.
  • The Claude Agent SDK gives you Claude Code's agent loop for your own products.

Practise 15 Automation, CI & Agent SDK questions →

Claude API Fundamentals 15%

  • Messages API: send user/assistant turns with an optional system prompt.
  • Tool use loop: run tools while stop_reason is "tool_use".
  • Prompt caching reuses long prefixes for lower cost and latency.
  • Batches for bulk, non-urgent work; streaming for responsive UIs.

Practise 15 Claude API Fundamentals questions →

Try three sample questions

Pick your answer, then open the reveal. The full bank has 100 questions.

Sample 1Claude Code Foundations & CLI
You've installed Claude Code. How do you start an interactive session in your project?
  1. Open the project folder in a terminal and run claude
  2. Run claude init --interactive from your home directory
  3. Upload the project as a zip to claude.ai and start a chat
  4. Run npx anthropic start from inside the project folder
Reveal answer
Answer: A. Running claude in the project directory starts an interactive session with that folder as its working directory. The other commands are not how Claude Code is launched.
Sample 2Extending Claude Code
You create .claude/commands/changelog.md. What do you get?
  1. A hook that runs the file's steps after every git commit
  2. A subagent named changelog with its own context window
  3. An MCP server called changelog exposing the file as a tool
  4. A /changelog slash command that runs the file as its prompt
Reveal answer
Answer: D. Markdown files in .claude/commands/ become slash commands named after the file, shared with anyone who has the repo.
Sample 3Claude API Fundamentals
In the Messages API, where do you put a system prompt?
  1. As a 'system' role message in the array
  2. In a custom HTTP header on the request
  3. In the top-level system parameter
  4. As the very first assistant message
Reveal answer
Answer: C. The Messages API takes the system prompt as a top-level system parameter; the messages array uses 'user' and 'assistant' roles.

A four-week study plan

  1. Week 1, foundations. Complete Levels 1 and 2, then practise the Claude Code foundations domain.
  2. Week 2, configuration. Level 3: CLAUDE.md, settings and permissions. Set them up in a real project.
  3. Week 3, extending. Level 4: build one skill, one subagent, one hook and connect one MCP server.
  4. Week 4, automation & API. Level 5, then take the timed mock exam until you score 800+ consistently.

Frequently asked

Is this an official Claude certification?

No. Anthropic offers a Claude Certified Developer – Foundations exam, which focuses mainly on building with the Claude API, and Claude Code is a small part of it. This track focuses on Claude Code itself. For the exam, use our Developer – Foundations track.

Should I do this track or the Architect track first?

If you are newer to Claude Code, start here. It builds the hands-on knowledge the Architect exam assumes. If you already build agentic systems, go straight to Architect.

Do I need to know how to code?

Basic familiarity helps, but the lessons explain each concept from scratch. Many questions test configuration and judgement rather than writing code.

Independent practice material, not affiliated with Anthropic. This is our own skills track, not an Anthropic certification.