Here's Brownfield Trick Number One: the Readers' Digest approach to documentation. This is all about optimizing your context usage.

When you're working with a large codebase, you don't want to load hundreds of files into context. Instead, create condensed documentation that gives Claude just what it needs to know.

Use the `/init` command to create documentation files in your project. For example, if you're working on a complex system, create a doc that explains the architecture, the key components, how data flows through the system, and where the important files are. Think of it like a compressed knowledge base - the Reader's Digest version of your codebase.

This serves two purposes. First, it saves context space. Instead of loading twenty files to understand the system, Claude can read one well-written doc. Second, it helps onboard new team members. When someone new joins your team and starts using Claude Code, they get instant access to this condensed knowledge.

The key is to focus on the "what" and "why" rather than the "how." Don't reproduce the code - explain what each subsystem does, how they interact, and where to find things. Then Claude can read the actual code files only when it needs the implementation details.

This is probably the single most effective trick for working with large legacy codebases. It turns context from a liability into an asset.