Claude Code Usage: CCDV-F Developer Exam Domain Guide

Claude code usage sits in Domain 3 of the CCDV-F blueprint at 3.1%, making it the second-smallest domain on the 53-question paper. That weight means roughly 1 to 2 items per sitting. Those items cost the same scaled points as any Domain 2 question, so they are worth preparing for specifically, just not over-preparing. This guide maps what the exam actually tests, how to allocate time across all eight domains, and where Domain 3 knowledge compounds with higher-weight areas.

What is the CCDV-F exam and how is it structured?

The Claude Certified Developer, Foundations (CCDV-F) exam costs $125 per attempt, runs 53 items over 120 minutes, and scores on a 100-to-1000 scale with 720 as the passing bar. It launched 12 March 2026 as part of the Claude Partner Network, a $100 million programme. By 3 June 2026, the network had recorded 10,000 or more certified individuals across all four tracks.

Every item on the CCDV-F is scenario-based and tests practical judgment. Unlike the CCAR-F Architect exam, CCDV-F does not draw items from a pre-built scenario bank; items are written directly against the skills in each domain.

The eight domains and their official blueprint weights are:

Domain 2 alone carries more weight than Domains 3, 4, and 7 combined. That ratio should drive your study plan directly.

What does Domain 3 actually test about Claude Code usage?

Domain 3 tests applied judgment about configuring, invoking, and integrating Claude Code into a developer workflow. Items present a realistic problem and ask for the correct configuration decision or root cause. No item tests recall of CLI flag names in isolation.

The skills the domain targets:

• Choosing between interactive and headless (non-interactive) modes for a given context
• Reading and writing configuration at each level of the three-level hierarchy
• Setting permissions that allow or deny specific tool calls
• Writing CLAUDE.md files that provide accurate, persistent context to the model
• Recognising when hook-based enforcement is required versus prompt instructions

Claude Code Configuration & Workflows covers the configuration model that Domain 3 scenario items draw on. Reviewing it before your sitting addresses the realistic question surface without excess preparation.

How should you allocate study time across all eight domains?

Proportionate to the blueprint weights. Domain 2 (33.1%) deserves more than ten times the preparation of Domain 3 (3.1%). Spending equal time per domain is one of the most common inefficiencies in CCDV-F preparation.

A weight-proportional allocation for a 40-hour study plan:

For Domain 3, 1.5 hours spent on the configuration model and two or three scenario types covers the realistic exam scope. Extending past that point yields diminishing returns against a domain worth roughly 1 to 2 questions.

What is the three-level configuration hierarchy for Claude Code?

The three-level configuration hierarchy is the central concept for Domain 3 scenario questions. The three levels are, from highest to lowest precedence:

1. Enterprise policy (system-level, set by the organisation, cannot be overridden by users or projects)
2. Project settings (.claude/settings.json, committed to version control, applies to all contributors and CI runs)
3. User settings (~/.claude/settings.json, personal, applies only to the individual developer)

The exam tests which level to use for a given requirement. A typical item presents a team or organisational requirement and asks which configuration level reliably enforces it.

{
  "permissions": {
    "allow": ["Bash(git:*)"],
    "deny": ["Bash(rm:-rf*)"]
  }
}

If the requirement says "all developers must" or "the organisation requires", the correct answer will always be the project or enterprise level. User settings cannot be relied on because each developer controls their own file. The exam consistently rewards deterministic enforcement over probabilistic approaches.

What is headless mode and when does the exam test it?

Headless mode runs Claude Code non-interactively and is invoked with --print (or -p). It writes output to stdout, making it the correct choice for CI pipelines, automated review workflows, and any context where no human is present in the session.

claude --print "Review this diff and output a JSON list of issues"

A typical Domain 3 scenario asks: a developer needs Claude Code to validate pull requests in CI and fail the build on specific findings. The correct pattern is headless mode with structured output parsing, not an interactive session. The --allowedTools flag restricts which tools Claude Code may call during the run, useful for read-only CI contexts:

claude --print --allowedTools "Read,Bash(git:status)" "Summarise recent changes"

Expect one item in the headless-versus-interactive decision space, and understand the output contract that headless mode provides.

What version control considerations does Domain 3 cover?

The version control implications concept addresses which Claude Code files should be committed and which should not. Project-level settings (.claude/settings.json) belong in version control so that all contributors and CI runs share identical permissions and deny rules. User-level settings are personal and must not be committed.

A common exam scenario: a team sets a deny rule in user settings, a new contributor joins without the restriction, and a CI pipeline running as a service account also lacks it. The root cause is always the same: only project-level settings committed to the repository propagate reliably to all parties.

CLAUDE.md files follow the same pattern. A project-level CLAUDE.md is committed to the repository and provides persistent context to every invocation, interactive or headless. An item might present two developers getting different Claude Code behaviour and ask why; if one has a personal CLAUDE.md override that conflicts with the committed file, that is the cause.

How does Claude Code usage connect to the higher-weight domains?

Domain 3 knowledge appears in items formally assigned to other domains, which is why understanding it pays returns beyond the 3.1% weight alone.

Tool Design & MCP Integration (Domain 8, 10.6%): Claude Code is extensible via MCP servers. Questions about adding a custom tool to Claude Code draw on MCP integration patterns alongside Domain 3 configuration knowledge. The MCP scoping hierarchy, environment variable expansion, and tool description quality all apply.

Agentic Architecture & Orchestration (Domain 1, 14.7%): Headless Claude Code invocations are components in larger agent pipelines. Structured output from a headless run can feed downstream agents, connecting Domain 3 to agentic workflow design.

Security and Safety (Domain 7, 8.1%): Permission gates, hook-based policy enforcement, and scope restrictions carry security implications. A Domain 7 item about a secure CI deployment pattern may involve correct Claude Code configuration as part of the correct answer.

AI Skill Certs' CCDV-F prep includes adaptive study, Archie tutoring, and practice exams scored on the same 100-to-1000 scale as the real exam, with 720 as the passing bar. AI Skill Certs is an independent platform and is not affiliated with or endorsed by Anthropic.

What do strong practice items for Domain 3 look like?

Effective Domain 3 practice items present a developer or team scenario and ask for the root cause of a misconfiguration or the correct enforcement decision. They test the configuration model, not memorisation of flag syntax.

A representative scenario: a security team requires that Claude Code never reads files outside the project directory during CI runs. A developer has added the restriction to their personal ~/.claude/settings.json. The CI pipeline runs as a service account and does not have the restriction. What is the correct fix?

The answer: move the deny rule to .claude/settings.json at the project level and commit it to the repository. The CI service account reads the committed project file; user settings are invisible to it. This combines Domain 3 configuration knowledge with Domain 7 security reasoning, which is the profile of a real CCDV-F item.

Per Anthropic's official CCDV-F exam guide (published 8 July 2026), every item is scenario-based and tests practical judgment. That framing applies equally to the smallest domain and the largest.

What should you review the day before the exam for Domain 3?

A short pre-exam review for claude code usage should confirm three things. First, you can explain which configuration level enforces a given requirement and why. Second, you can distinguish headless from interactive mode and name the flag. Third, you know which files belong in version control and which do not.

Beyond those three, the pre-exam day is better spent on Domain 2 (Applications and Integration, 33.1%) or Domain 5 (Model Selection and Optimization, 16.8%), which together carry roughly half the exam. Domain 3 is a domain to master early and not revisit; the concepts are stable and the question surface is small.