Contents — domains, guide and mocks CCDV-F Claude Certified Developer — Foundations Domains
1 Agents and Workflows 14.7% 1.1 Agent architecture and tradeoffs 1.2 Building agents with the Agent SDK 1.3 Loops, subagents, memory and context 3 of 3 lessons written
2 Applications and Integration 33.1% 2.1 From business need to requirements 2.2 Life cycle of a Claude system 2.3 Claude API mechanics 2.4 Engineering foundations 2.5 Designing a Claude application 2.6 Configuration management 6 of 6 lessons written
3 Claude Code 3.1% 3.1 Operating Claude Code 1 of 1 lessons written
4 Eval, Testing, and Debugging 2.6% 4.1 Debugging and error handling 1 of 1 lessons written
5 Model Selection and Optimization 16.8% 5.1 LLM fundamentals 5.2 SDKs, REST and streaming 5.3 Choosing a model 5.4 Cost and token management 4 of 4 lessons written
6 Prompt and Context Engineering 11% 6.1 Context engineering 6.2 Prompt engineering 6.3 Output handling 3 of 3 lessons written
7 Security and Safety 8.1% 7.1 Prompt injection and data leakage 7.2 Guardrails and safe deployment 7.3 Hooks as safety controls 7.4 Identity, secrets and keys 4 of 4 lessons written
8 Tools and MCPs 10.6% 8.1 Tool implementation 8.2 MCP server development 8.3 Agentic customization 3 of 3 lessons written
CCDV-F · Practice · Domain 3 Claude Code CCDV-F · Domain 3 3 questions · 3.1% of the exam
Answer everything, then check. Each result links back to the lesson for the objective it came from.
0 of 3 answered0:00
Question 1 · 3.1
A monorepo has CLAUDE.md at the repository root and another in services/billing/. An engineer launches Claude Code from services/billing/ and asks why the root instructions still seem to apply.
What is the correct explanation?
A Only the nearest CLAUDE.md loads; the root file must have been imported with @. B Both files load and are concatenated, root first and the nearer file last. C The root file wins because broader scope has higher precedence. D The nearer file replaces the root file for any heading they share. Question 2 · 3.1
A team wants a nightly job that reviews the previous day's merged pull requests and writes findings to a file, with no human available to answer prompts.
Which two choices make that run behave predictably? (Select 2.)
A Run claude -p and pre-approve only the tools the job needs with --allowedTools. B Use --output-format json so the script can read the result and session id programmatically. C Add an instruction to CLAUDE.md telling Claude not to ask for permission. D Rely on the default text output and grep the response for keywords. E Pass --resume with no session id so the job starts clean each night. Question 3 · 3.1
A security team wants Claude Code sessions across the company never to run a particular family of shell commands, whatever an individual engineer configures locally.
Where should that rule live?
A In each engineer's ~/.claude/settings.json, distributed by a setup script. B In the repository's .claude/settings.json, committed to version control. C In managed settings deployed by the organisation. D In the root CLAUDE.md, stated as a prohibition. Check my answers You can change answers until you check. Nothing is saved or sent anywhere.