Configure CLAUDE.md files with appropriate hierarchy, scoping, and modular organization
What you should be able to do
- Place instructions correctly across the user, project and directory levels of the memory hierarchy.
- Diagnose why a teammate is not receiving instructions that work on your own machine.
- Keep configuration modular with @import and with topic files under .claude/rules/.
- Use /memory to confirm which memory files are loaded before theorising about behaviour.
Exam traps (4)
It works locally, so the project configuration must be right — the new joiner set something up wrong.
User-level memory is not version-controlled, so instructions that live there are invisible to everyone else. The author is the last person who can reproduce the problem.
Anything the team must share belongs at project level. "Works for me" is the signature of a user-level rule.
Keep one CLAUDE.md and grow it as conventions accumulate.
It is loaded in full for every task, so every session pays for every convention and the relevant lines sit in the middle of a long file. Splitting it into @imports organises the file and saves nothing — imported files load at launch too.
Split by topic for legibility, and scope with paths on a rule so a convention only loads in sessions that touch matching files.
Add the instruction again, higher up, when behaviour looks inconsistent.
Duplicating a rule across levels produces contradictions that resolve differently by directory, and the inconsistency gets worse rather than better.
Check what is loaded first with /memory; fix the placement rather than adding another copy.
Put a package's standards in the root file so nothing is missed.
Every package then loads every other package's conventions, and the model applies rules from a codebase area it is not working in.
Import the standards each package actually needs into that package's own file.
Where the guide and the current docs differ (1)
which command proves what loaded
Answer this: /memory verifies which memory files are loaded.
The docs now say: /memory lists file locations, including entries for files that do not exist yet; /context is what reports which files actually loaded into the session. Read it
Primary sources
- Claude Code — memorychecked 2026-08-15
- Claude Code — settingschecked 2026-08-15
Know cold
- Project memory file
- Memory @import
- /memory command
- Skill
18 practice questions in the bank are tagged to this task statement.
Practise this task in context: open it inside the interactive study guide, which carries the concept cards, the mock quiz and the practice simulation.