The one-character upgrade for your CLAUDE.md

There is a file in every Claude Code project that starts out as a tidy one-page briefing and ends up as the drawer in your kitchen with the batteries, the takeout menus, and a single mystery key. Mine crossed 400 lines before I admitted it. Everything in it was true and useful; I just couldn't find any of it, and worse, neither could I tell what was safe to change.

The fix is one character, it's documented, and almost nobody uses it. Not "almost nobody" as a figure of speech — there are dozens of videos about the file itself and, as far as I can find, none about this feature. That's the kind of gap this field note exists for.

One flag before we start: this is a Claude Code feature, and Claude Code comes with Anthropic's paid plans. If you're not using it yet, read on anyway — the filing-cabinet idea at the center of this works in any tool that lets you give it standing instructions.

Quick read: Claude Code reads a markdown file named CLAUDE.md at the start of every session and treats it as standing instructions for the project. A line containing @ followed by a file path — for example, @docs/brand-voice.md — imports that file's contents into the briefing at launch. Imports accept relative and absolute paths, can nest up to four hops deep, and are ignored inside backticks and code blocks. Importing organizes instructions across multiple files but does not reduce context usage, because imported files still load in full at session start. The syntax is documented in Claude Code's memory documentation.

Three words to name before they land cold. CLAUDE.md is the briefing file — plain text you write once, that Claude re-reads at the start of every session so you stop re-explaining your business. An import is a line that says "go get that other file and read it too." And your repo — if you have one — is just the folder that remembers every change ever made to it; the briefing file lives at the top of it.