CLAUDE.md, rules and memory, from the Claude Code documentation

Domain 3 examines where configuration lives and who receives it. The exam guide names three levels. The documentation describes four, plus a second memory system the guide does not mention at all.

Every CLAUDE.md location

Listed in load order, broadest first, so a project instruction lands in context after a user instruction.

Files are concatenated, not overridden. Across a directory tree, content is ordered from the filesystem root down to your working directory, so the file closest to where you launched Claude is read last. Within a directory, CLAUDE.local.md is appended after CLAUDE.md.

Subdirectory files are discovered but load on demand, when Claude reads a file in that directory — not at launch.

The limits worth memorising

Imports

@path/to/import pulls another file in. Relative paths resolve against the file containing the import, not the working directory. Imported files can import further files, to a maximum depth of four hops.

Import parsing skips code spans and fenced blocks, so a path wrapped in backticks stays literal text while the same path outside them imports the file. That is the documented way to mention a path in a CLAUDE.md without pulling it in.

Path-scoped rules

.claude/rules/ holds topic files. A rule with no paths frontmatter loads at launch with the same priority as .claude/CLAUDE.md. A rule with paths loads only when Claude reads a matching file.

Brace expansion works — src/**/*.{ts,tsx} — and a rule's whole paths list shares a budget of 1,000 expanded patterns. User-level rules live in ~/.claude/rules/ and load before project rules, giving project rules the higher priority.

/memory and /context

/init generates a starting CLAUDE.md from the codebase, and suggests improvements rather than overwriting if one already exists.

What survives /compact

A project-root CLAUDE.md is re-read from disk after compaction and re-injected. Nested CLAUDE.md files and path-scoped rules reload as Claude reads files they apply to. An instruction that disappears after /compact was given only in conversation — which is the argument for writing it down.

Skills and commands, as documented

Two facts worth holding beside the guide's account.

Custom commands have been merged into skills. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy and work the same way. Existing command files keep working; skills add a directory for supporting files, frontmatter, and automatic invocation.

argument-hint does not do what the guide says. The documentation describes it as a hint shown during autocomplete to indicate expected arguments — not a prompt raised when a skill is invoked without any. The exam grades the guide's version.

The CLI flags this exam names

Sources

• How Claude remembers your project — https://code.claude.com/docs/en/memory (retrieved 2026-08-30)
• Extend Claude with skills — https://code.claude.com/docs/en/skills (retrieved 2026-08-30)
• CLI reference — https://code.claude.com/docs/en/cli-reference (retrieved 2026-08-30)