dev-manual

You can minimize token usage by selectively loading only the relevant development manual chapters instead of feeding the entire guide to the AI. This targeted approach provides task-oriented guidance for new features, debugging, or refactoring while significantly cutting token consumption.

Selective chapter reading is a workflow mechanism that loads specific bookmarked sections from a chapter library to guide AI work. It prevents unnecessary token usage by ensuring only context-relevant documentation informs the AI during tasks like API development or refactoring.

Set up selective reading by placing your development manual chapters under the .cwm/dev-manual/chapters directory. Once organized there, you can specify and load only the bookmarked chapters relevant to your current task to guide the AI.

Yes, selective manual reading applies directly to debugging and refactoring tasks. By loading only the chapters covering coding standards or debugging protocols, the AI receives targeted workflow guidance without wasting tokens on unrelated documentation.

The best way to provide project context within token limits is through selective chapter access. Loading only task-relevant sections from the chapter library, rather than the full manual, delivers lightweight self-service context that reduces cognitive load and saves tokens.