Claude Code Version Control: CCDV-F Developer Exam Guide

The claude code version you install and configure directly shapes how your team develops with Claude, and it shapes a small but connected portion of the CCDV-F exam. Domain 3 (Claude Code) carries 3.1% of the 53-item test, yet the version-pinning and configuration-hierarchy concepts it covers support correct answers in Domain 2 (Applications and Integration, 33.1%), the heaviest domain on the exam. We treat these topics as linked, not siloed.

What does "claude code version" mean for CCDV-F candidates?

"Claude Code version" covers two related concerns on the CCDV-F: the version of the Claude Code CLI you are running, and the version-control implications of Claude Code configuration files in a project repository. Both appear as scenario contexts in Domain 3.

The CCDV-F exam, launched 12 March 2026, contains 53 items across eight domains. Passing requires a scaled score of 720 out of 1000. Each item is scenario-based and tests practical judgment rather than rote recall. Unlike the CCAR-F Architect exam, CCDV-F does not draw from a randomised scenario bank; items are written directly against the skills in each domain, which means Domain 3 questions test concrete Claude Code knowledge rather than general architectural reasoning.

What is the Claude Code configuration hierarchy?

Claude Code applies a three-level configuration hierarchy: system-level settings that apply across all projects on a machine, project-level settings stored in .claude/settings.json and CLAUDE.md files, and user-level settings that override project defaults for a specific developer.

The exam tests whether candidates know which level takes precedence and which files belong in version control.

Precedence flows from user to project to system: a developer's local settings override the project defaults, which override system defaults. The exam exploits this in scenarios where a new hire's local config suppresses a security-critical hook that the team lead assumed was always active. The correct answer is always to move enforced requirements to a level the individual developer cannot override at runtime.

The hierarchy also governs CLAUDE.md file composition. Claude Code reads CLAUDE.md files at multiple levels of a repository: the home directory, the project root, and sub-directories. Files are concatenated at load time. This means teams can scope rules to particular sub-directories by placing a CLAUDE.md in each folder, a pattern that appears in Domain 2 questions about multi-team repository management.

How do version control implications apply to Claude Code?

The version control implications of Claude Code configuration come down to one principle: commit the files that define shared behaviour, exclude the files that capture personal preferences or secrets.

In practice:

• CLAUDE.md files at project, sub-directory, and repository roots should be committed. They define the instructions Claude Code follows for every team member who opens the project.
• .claude/settings.json should be committed when it contains project-wide tool permissions and hooks.
• .claude/settings.local.json must not be committed. It holds individual overrides, including personal API keys or model preferences.
• Any environment file or secrets file referenced by configuration must be excluded via .gitignore.

A CCDV-F question will often describe a scenario where secrets are leaking into the shared repository via the wrong config file, or where a new developer joining the team gets different model behaviour from everyone else. The correct answer separates these concerns at the right configuration level.

To check your active claude code version and the configuration currently in effect:

# Check the active claude code version
claude --version

# List all configuration values currently in effect
claude config list

For CI pipelines, specifying the Claude Code version and model explicitly prevents silent upgrades from changing agent behaviour between pipeline runs:

# Headless run with explicit model version in CI
claude --model claude-sonnet-4-6 \
  -p "Run the integration test suite and output failures as JSON" \
  --output-format json

The --output-format json flag matters for CI integration: it makes Claude Code output machine-readable so downstream steps can parse results rather than pattern-match free-form text. Domain 2 questions test whether candidates select this flag correctly in automated pipeline scenarios.

What are all eight CCDV-F domain weights?

Distributing study time effectively requires knowing the exact domain weights from Anthropic's official exam guide. Domain 2 dominates at 33.1%; Domain 3 is the second-smallest at 3.1%.

Anthropic does not publish the raw-to-scaled conversion formula, so never plan preparation around hitting a specific item count per domain. A scaled score of 720 is required to pass.

How do Claude Code configuration concepts connect to Domain 2?

The connection between Domain 3 and Domain 2 (Applications and Integration, 33.1%) is direct. Production Claude Code deployments depend on correct configuration management, and Domain 2 tests exactly the kind of integration failure that misconfiguration causes.

Three scenario patterns are worth practising:

Pattern A: Leaked secrets in CI. A team deploys Claude Code to a CI server. A developer's personal .claude/settings.local.json was accidentally included in the Docker build context, exposing an API key in the image layer history. The fix requires understanding which config files must never be packaged into deployment artefacts.

Pattern B: Silent model upgrade breaks a pipeline. A pipeline that reliably passed last month now fails intermittently. Claude Code was updated between runs, and the new model interprets a prompt differently. The fix is to pin the model version explicitly with --model claude-sonnet-4-6 in the CI command. Domain 2 rewards candidates who know how to make agents deterministic under operational change.

Pattern C: Hook suppressed by user config. A compliance hook in .claude/settings.json is supposed to block file writes outside permitted directories. It fires correctly for most of the team, but a senior developer's local override in settings.local.json suppresses it. The fix is to move enforcement to a programmatic layer rather than relying on a configuration file a user can override.

These scenarios reward candidates who have studied Claude Code configuration systematically rather than memorised isolated facts.

How should I allocate study time for Claude Code on CCDV-F?

Because Domain 3 is 3.1% of the exam, study it in context rather than in isolation.

1. Anchor preparation on Domain 2 (Applications and Integration, 33.1%). This is where the most marks sit, and it subsumes many Domain 3 application scenarios.
2. Study Domain 3 configuration concepts for roughly 30 to 45 minutes: the three-level hierarchy, version control implications, headless mode, and hook configuration are the core topics.
3. Connect Domain 3 to Domain 1 (Agents and Workflows, 14.7%) when studying CI/CD agent pipelines. The configuration files you commit determine the tooling context for every automated run.
4. Connect CLAUDE.md writing to prompt engineering (Domain 6, 11.0%). CLAUDE.md files are project-level system prompts, and the same precision principles that produce reliable API outputs apply equally to project instruction files.
5. Use practice questions to test whether you can distinguish project-level from user-level settings under time pressure.

The adaptive study engine on AI Skill Certs tracks mastery at concept level using Bayesian Knowledge Tracing with a 0.90 mastery threshold. When you answer Domain 3 questions correctly in practice exams, the engine reallocates study time toward the gaps that will cost you the most marks on the actual test.

What is the full CCDV-F exam format?

The Claude Certified Developer, Foundations exam costs $125 USD per attempt and is delivered online-proctored or at a Pearson VUE test centre. The credential is valid for 12 months from the date it is awarded.

As of 3 June 2026, more than 10,000 individuals hold a Claude certification across all four active tracks in the Claude Partner Network programme. The CCDV-F is one of the two $125 Foundations-level exams alongside the CCAR-F Architect track.

AI Skill Certs is an independent prep platform. We are not affiliated with, endorsed by, or approved by Anthropic.