Study guide for the CCA-F Claude Code domain: CLAUDE.md, settings and permissions, hooks, commands, subagents, plan mode and CI use.
Claude Code Configuration & Workflows carries 20% of the Claude Certified Architect – Foundations blueprint, about 12 questions on a 60-question form. It tests whether you can set Claude Code up for a team — so that conventions are shared, risky actions are controlled and repeatable work is packaged — rather than whether you can type commands quickly.
What the domain covers
Memory files and their scope. CLAUDE.md files carry instructions Claude Code loads automatically. A file at the repository root, committed to version control, holds team conventions: build and test commands, code style, architecture notes. The user-level file in your home directory holds personal preferences across all projects. CLAUDE.md files in subdirectories add guidance for that part of a codebase, which keeps a monorepo's root file short. Keep memory concise and current; move long reference material into separate files that are imported or read on demand.
Settings and permissions. Project settings in .claude/settings.json are shared with the team; a local settings file holds personal overrides that are not committed; user settings apply everywhere. Permission rules allow, ask for or deny specific tools and commands. For anything that must never happen, use deny rules or hooks, not instructions.
Hooks. Hooks run your commands at defined points in Claude Code's lifecycle. A PreToolUse hook can inspect a tool call and block it — for example edits to a protected directory. A PostToolUse hook can run a formatter or linter after every edit. The exam's recurring theme: hooks are deterministic, CLAUDE.md is advisory.
Commands, skills and subagents. Repeated prompts become project commands or skills stored under .claude in the repository, so the whole team invokes the same routine by name. Specialised assistants — a test writer, a security reviewer — are subagents defined with their own instructions and restricted tool list, running in their own context.
MCP in Claude Code. Servers the whole project needs go in a project-scoped .mcp.json that is committed to the repository, with credentials supplied through environment variables or each user's own authentication rather than written into the file.
Workflows. For unfamiliar or wide-reaching changes, explore and plan before editing: plan mode lets Claude read and propose without modifying files, and a human approves the approach. Between unrelated tasks, clear the session so stale context does not leak in; for one long task, compact it. In CI, run Claude Code non-interactively with a prompt flag and structured output, and give the job a tight allowlist because nobody is there to approve actions.
How questions are framed
Expect "where should this live?" questions (team versus personal, committed versus local) and "how do we guarantee this?" questions where the trap is an instruction in a memory file. Questions about pipelines often hinge on headless mode, restricted tools and not handing automation more credentials than it needs.
Study checklist
- Draw the hierarchy of memory and settings files and say who sees each one.
- Write one PreToolUse rule that blocks a path and one PostToolUse formatter step.
- Package a routine your team repeats as a project command or skill.
- Define a subagent with a narrow tool list and describe when the main session should delegate to it.
- Sketch a CI job that runs Claude Code headless with only the tools it needs.
Domain weights and the 60-question form size come from published summaries of Anthropic's exam guide, which is available to registered partners; see our exam guide page for what is confirmed by Anthropic's public FAQ.
Free sample questions: Claude Code
Original questions written for practice, not real exam items. Each one appears again, with fresh option order, in the full practice exam.
Sample question 1
Scenario: Code generation with Claude Code
A team wants every developer's Claude Code sessions in a repository to follow the same build commands, code style rules and test conventions. Where should these instructions live?
- A. In each developer's user-level ~/.claude/CLAUDE.md, kept in sync by a setup script
- B. In the repository README, which Claude Code reads first when a session opens
- C. In a CLAUDE.md at the repository root, committed to version control
- D. In a pinned message that each developer pastes at the start of every session
Show answer and explanation
Answer: C. A project CLAUDE.md in the repository is loaded automatically for everyone working there and is versioned with the code, so conventions stay consistent and reviewable. User-level files are for personal preferences and drift even with sync scripts. A README is not the memory file Claude Code loads, and pasted messages are neither automatic nor reliable.
Sample question 2
Scenario: Code generation with Claude Code
A developer prefers terse answers and a particular editor command, but these are personal preferences, not team rules. Where should they put them?
- A. In the project's .claude/settings.json, alongside the team's permission rules
- B. In the repository's root CLAUDE.md, under a heading with the developer's name
- C. In their user-level ~/.claude/CLAUDE.md, which applies across all their projects
- D. In a CLAUDE.md inside a subdirectory that only this particular developer works in
Show answer and explanation
Answer: C. The user-level memory file applies to that developer in every project and is not shared. Putting personal preferences in committed project files imposes them on teammates. settings.json holds configuration such as permissions, not prose preferences, and a subdirectory CLAUDE.md affects anyone working in that directory.
Sample question 3
Scenario: Code generation with Claude Code
A monorepo contains a frontend and a payments service with very different conventions. The root CLAUDE.md has grown long and mixes both. What is the better structure?
- A. Remove CLAUDE.md entirely and let Claude infer conventions from the existing code
- B. Keep everything in the root file but organise it under clear per-service headings
- C. Copy the full root CLAUDE.md into each service folder and delete the irrelevant parts
- D. Keep shared rules at the root and put service-specific rules in each service's CLAUDE.md
Show answer and explanation
Answer: D. Claude Code reads CLAUDE.md files hierarchically, picking up subdirectory files when it works in those areas, so guidance scoped close to its code keeps each context relevant and the root short. Headings do not stop irrelevant rules from loading. Full copies duplicate shared rules that then drift. Conventions not obvious from code are lost without any file. (Imports with @path can also help split a long file.)
Sample question 4
Scenario: Developer productivity with Claude
Your team runs the same multi-step review routine many times a day ("check the diff against our security checklist, then summarise risks"). What is the best way to package it in Claude Code?
- A. Package it as a project command or skill under .claude/ so anyone can invoke it by name
- B. Save it in a shared snippets document that developers paste into the chat when needed
- C. Turn it into a PostToolUse hook so that it runs automatically after every single file edit
- D. Add the routine to the root CLAUDE.md so it is in context for every single session
Show answer and explanation
Answer: A. Repeatable prompts belong in versioned, shareable form: a project command (.claude/commands/) or skill (.claude/skills/) is committed with the repo and loaded only when invoked. Putting it in CLAUDE.md spends context in every session. Pasting is error-prone. A hook after every edit would run the review far too often and is meant for deterministic commands.