Tune editing ergonomics with vim mode, key shortcuts, and /config

There’s one layer below the commands and the inline moves: the prompt input itself, and the keys you press a thousand times a day to drive it. None of this changes what Claude Code can do — it changes how little friction sits between a thought and a submitted prompt. This is the last and smallest polish, and the place where shaving a half-second off a constant motion actually adds up.

Vim mode, if that’s how your hands think

If you edit everything else in vim, editing prompts in a plain readline buffer is a constant small tax — every time your fingers reach for Esc, dd, ciw and nothing happens. Claude Code has a full vim editing mode for the prompt. Turn it on under /config → Editor mode, per the interactive-mode docs, and the input gains NORMAL and INSERT modes with the motions and operators you’d expect — h/j/k/l, word motions, dw/cw/ciw, text objects like ci", visual mode, the lot.

Two interactions are worth knowing so it doesn’t surprise you. When the cursor is at the top or bottom edge of the input in NORMAL mode and can’t move further, j/k start navigating command history instead — so vim motion and history browsing share the same keys, which is what you’d want. And / in NORMAL mode opens history search, the same as Ctrl+R. If vim isn’t your default, skip it entirely; this is ergonomics, not a requirement, and forcing modal editing on hands that don’t already speak it is friction, not relief.

The shortcuts that actually earn their keep

The keyboard reference is long, and most of it you’ll never touch. A small set, though, is worth committing to muscle memory because they come up constantly:

Esc            interrupt Claude mid-turn — redirect without losing the work done so far
Esc Esc        on empty input, open the rewind menu; with text, clear the draft (and recall it)
Ctrl+R         reverse history search (last lesson)
Ctrl+G         open your current prompt in your real $EDITOR, for when it's getting long
Ctrl+O         toggle the transcript viewer to see full tool calls and output
Ctrl+L         redraw the screen when the terminal display gets garbled
Shift+Tab      cycle permission modes — the ladder from the permissions chapter

A few of these connect back to moves you already know. Esc is the gentle interrupt — it stops the current response or tool call so you can correct course, and Claude keeps everything it had already done, which is far better than Ctrl+C-ing the whole thing. Double-Esc on an empty prompt is the same rewind menu you met in sessions. And Shift+Tab is the Shift+Tab ladder from permissions — the single most-used shortcut in a real session, because you’re constantly loosening and tightening the leash as the task changes.

Ctrl+G deserves a specific call-out: when a prompt grows past a couple of lines — a spec, a careful instruction, a block you’re editing word by word — pop it into your real editor instead of wrestling it in the input buffer. You get your full editing setup, save, and it lands back in the prompt.

/config, and rebinding the keys themselves

Most of the friction that’s left lives in /config. It’s worth opening once, deliberately, and reading down the list rather than discovering each toggle the hard way — editor mode, session recap, the external-editor and prompt-suggestion behaviours, theme, and more all live there. Tuning it once to your taste is a five-minute investment that quietly removes annoyances for months.

And if a default key itself is wrong for you — it collides with your terminal multiplexer, or you simply want Ctrl+R somewhere else — the keys are not fixed. Run /keybindings to open ~/.claude/keybindings.json, where you can rebind actions per context or unbind a default outright; the full action list and syntax are in the keybindings docs. A handful are reserved — Ctrl+C, Ctrl+D — but nearly everything else is yours to move. Most people never need this; the few who do are usually fixing a real conflict, and it’s good to know the escape hatch exists.

You’ve reached the end

That’s the course. Look back at the arc you’ve walked: you installed Claude Code and shipped a first reviewed change, then learned to keep a session alive across days and manage its context like the resource it is. You set the permission leash long where it’s safe and short where it isn’t, made the agent research and plan before touching hard changes, and wrote down the context it needs as rules and memory. You picked the right model for the task, pushed noisy work into subagents, packaged repeatable procedures as skills, extended the tool with MCP and hooks, and automated whole jobs to run without you. And in this last chapter you smoothed all of it down into reflex.

The whole point was never any single command. It was the running habit underneath them — treating context, permissions, and plans as things you steer on purpose — and the fluency to do it without thinking. You have that now.

Two places to go from here. When you want the spec-level reference — the exhaustive “what is this primitive and every option it takes” view that this course deliberately kept light — that’s Foundations; it’s the other half of the same picture, lookup-shaped where this was a story. And if you work across more than one agent, the Codex and opencode courses are on the way, built on the same teach-by-doing spine you just followed.

Now go close the gap between you and the agent. That’s the whole job.