Two paths, both doing the same thing at heart: load context so Claude knows who you are and what you're working on.
Every morning. And every time you open a new chat or terminal. Without context, Claude is blind. With context, the session knows what's in the workspace, what's on this week's plate, and who you are.
Type /prime into the chat window at the top. Loads context docs (ich.md, business.md, todos.md, key-metrics.md).
Lightweight. No file operations. Perfect for questions throughout the day.
Type cr in the terminal. Starts Claude Code with workspace context and continues like a /prime.
When real operations happen on the workspace: read files, write files, run commands.
cr starts Claude Code with --dangerously-skip-permissions. Every file operation runs without confirmation. Use deliberately, only in folders where you want this. Not in third-party repos, not in your home directory.
One Mac app, one window. Chat area on top, terminal at the bottom, folder pane in the bottom right. You don't leave this app all day.
The top chat now has access to your workspace context. Ask it anything that comes up throughout the day.
Claude Code starts, loads the same workspace context, and is ready for real work. This is where skills, workflows, and builds run.
You're live. Top for random, bottom for specialist tasks.
/prime reads predefined Markdown files from the workspace and loads them as context into the chat. cr does the same in the terminal and additionally starts Claude Code in permission-skip mode. Both load the same truth, just in different environments.