6.1 The Daily Drivers: Slash Commands and Shortcuts

You know /clear, /usage, /exit, and ++esc++. This lesson completes the driver's toolkit and introduces the move that previews Chapter 7: packaging repeatable workflows.

What you will learn

• Use the core slash commands fluently, each at its right moment.
• Steer sessions with interrupts, mode toggles, and quick redirects.
• Create your first custom slash command — a reusable prompt with a name.

Builder principle

Fluency isn't knowing every control. It's never breaking flow to think about the controls.

The controls that matter

You've met the session managers. Here's the full daily set, organized by what they operate on:

On the conversation:

Command	Use it when
/clear	Between tasks — the whiteboard eraser (5.5)
/compact	You want to keep going but shed the bulk — compaction on your schedule, at a checkpoint you chose, not mid-thought
++esc+esc++ at an empty prompt	Open the checkpoint menu to restore code, conversation, or both to an earlier point

On your awareness:

Command	Use it when
/usage	Before/after large sessions — plan allowance for subscribers, local token and cost estimates for API users
/context	You're curious what's actually on the whiteboard — seeing the real context breakdown once will cement everything from 2.2 and 5.5
/help	You forget any of this — no shame in the lookup

On the work:

• ++shift+tab++ — cycle permission modes, including plan mode (4.2). The toggle you'll touch most.
• ++esc++ — interrupt mid-action. Steering, not scolding (4.1).
• ! prefix — run a shell command yourself without leaving the session (!git status, !ls). Small thing; changes the rhythm completely — you stop context-switching between two terminals.
• @ — reference a file by path with tab-completion (@docs/style.md), the fastest way to point the gather phase at exactly the right thing.

Ten controls. That's the whole cockpit. Fly it for a week and your hands know it.

Your first custom command

Here's the door to Chapter 7, opened a crack. .claude/commands/ remains compatible for custom slash commands. Skills in .claude/skills/ are the recommended format for reusable capabilities and richer supporting files; Chapter 7 moves you there. For this small exercise, a command keeps the first step visible.

Remember the librarian prompt from 5.4 — the one you're supposed to run at the end of every significant session? Make it a command:

$ mkdir -p .claude/commands

Create .claude/commands/wrap.md:

Update docs/log.md with a dated entry for this session: what we did,
what's unfinished, what to pick up next. If we made any decision that
future sessions should honor, add it to docs/decisions.md in the
standard four-line format. Then show me both updates.

When the command is available and current, typing /wrap requests the routine in a compact, consistent form. Confirm that each step ran; the shortcut does not guarantee completion.

The pattern generalizes immediately. Strong early candidates from what you already do:

• /wrap — the librarian routine (above)
• /ship-check — run your done-means verification pass from 4.4
• /fresh — read docs/log.md's top entry and summarize current state + next step

Commands can also take arguments and live at the user level (~/.claude/commands/) so they follow you across projects — the same layering logic as memory (5.3). Chapter 7 builds this into full Skills; for now, one good /wrap is plenty.

🌱 No-code lane⚙️ Engineer lane

Don't collect commands — earn them. The rule: type a prompt twice, command it the third time. Three earned commands beat twenty downloaded ones you don't remember owning.

Your commands directory is code — commit it. .claude/commands/ in the repo means the whole team gets /wrap and /ship-check for free, and improvements flow through review like everything else. Team-shared session tooling is one of those quiet force multipliers: it standardizes the practices, not just the prompts.

Try it now

1. Run /context in a mid-task session once today — look at what's actually in the window. Correlate with what you'd want there after Chapter 5.
2. Create /wrap exactly as above. Use it to close today's session.
3. Earn one more command from your own repetition — check your journal for the prompt you've retyped most.
4. Journal line: which of the ten controls changed how you work? ! and @ are useful candidates to test.

Check your understanding

• When would you reach for /compact instead of /clear, and what do you control by running it manually?
• What do ! and @ each do, and which loop phase does @ accelerate?
• What's the earn-a-command rule, and why does it beat collecting?
• Where do commands live for one project vs. all your projects — and what chapter-5 concept is that mirroring?