Claude Code Configuration and Workflows
Free Claude Certified Architect - Foundations practice — 6 questions on Claude Code Configuration and Workflows, with explanations. No sign-up. Full 12-question mixed test →
A large monorepo has a CLAUDE.md at the repo root and additional CLAUDE.md files inside packages/backend/ and packages/frontend/. The architect wants backend-specific conventions to apply when Claude is working in that directory, without discarding the global root conventions. How does Claude Code actually handle this setup?
Claude Code builds project memory by walking from the current directory up to the repo root (and beyond, to user-level memory), loading every CLAUDE.md it finds along the way and merging them, with nested files adding/overriding context relevant to that subdirectory.
An architect needs a reusable capability that Claude autonomously decides to invoke based on conversational context (not an explicit user command), bundles multiple supporting files (scripts, reference docs) under a folder, and uses progressive disclosure so the full content isn't loaded into context until needed. Which Claude Code feature fits this requirement?
Agent Skills are defined by a SKILL.md with metadata and can reference bundled scripts/resources in the same folder; Claude autonomously decides when a skill is relevant and only pulls in the full skill content (progressive disclosure) when needed, rather than loading everything upfront.
An architect wants Claude Code to be allowed to run any git command EXCEPT git push, and wants file edits to be applied without prompting for confirmation each time. Which settings.json configuration achieves this?
Allowing Bash(git:*) permits all git commands, the deny rule for Bash(git push:*) overrides that allow specifically for push, and defaultMode acceptEdits auto-applies file edits without per-edit confirmation prompts.
An architect is tasked with having Claude Code refactor a critical payment-processing module across roughly 40 files in a production codebase that has no existing automated test coverage. What is the architecturally sound approach?
Plan mode lets Claude explore the codebase and propose an approach with zero file mutations until the architect reviews and approves it, which is the appropriate risk control for a high-stakes, untested, multi-file refactor before allowing any execution.
An architect wants to integrate Claude Code into a GitHub Actions pipeline to automatically triage new issues and post a labeled summary comment, with no human interactively approving each tool call during the run. Which approach fits this headless CI use case?
The -p flag runs Claude Code in non-interactive print mode suitable for scripting and CI, and pairing it with a tightly scoped permission configuration lets it act (read issue, post comment) safely without a human approving each step, since none is available in the pipeline.
In the Claude Code 'explore, plan, code, commit' workflow pattern, what is the primary architectural reason for instructing Claude to explore and plan before writing code, rather than jumping straight to code generation?
Exploring the codebase and producing a plan grounds Claude in the actual structure, conventions, and constraints of the project before it writes code, which reduces guesswork and produces changes that fit the existing system more accurately than jumping directly to generation.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →