Claude Code Commands Quick Guide! Master These, Efficiency Soars! 2026 Latest Tested Version

Claude Code is the originator of the harness tool and one of my favorite Agent frameworks.

The reason I like Claude Code is partly because it’s powerful, and partly because I think this command-line interface (CLI) has a major advantage over GUI-based agent frameworks — it helps you stay focused on the task without distractions. GUI environments make it too easy to get sidetracked.

Here are some useful Claude Code commands, including CLI launch parameters, shortcuts, and backslash commands. Mastering these can greatly boost your productivity:

1. CLI Launch Commands (enter in terminal)

• claude → Start an interactive session (obviously the most used ^_^)

• claude "your question" → Start with a question directly

• claude -c or --continue → Continue the last session

• claude --resume → Resume a historical session

• claude --permission-mode auto → Start directly in Auto mode (recommended)

• claude --dangerously-skip-permissions → Dangerously skip all permission prompts (use with caution)

2. Common Shortcuts

• Shift + Tab → Cycle through permission modes (Normal → Auto → Plan) — most commonly used

• Esc → Interrupt current generation

• Esc + Esc → Clear input / open fallback menu

• Ctrl + C → Cancel generation (press twice to fully exit)

• Ctrl + G → Open system editor to write long prompts

• Ctrl + J → Insert newline (without sending)

• Ctrl + V (Mac) / Alt + V (Win) → Paste image

• ↑ / ↓ → Scroll through input history

3. Important Backslash Commands

• /init → Initialize project, generate CLAUDE.md (first step for new projects)

• /plan [description] → Enter planning mode (read-only, essential before large tasks)

• /memory → Edit project memory file CLAUDE.md

• /permissions → Manage permission rules

• /compact → Compact conversation history to reclaim tokens

• /context → View current context and token usage

• /clear or /reset → Clear current conversation history

• /model → Switch models

• /diff → Show code changes

• /code-review → Code review

• /btw <question> → Ask a side question (doesn’t pollute main history)

• /rewind → Rewind to the last checkpoint

• /help → Show all commands

4. Prompting Tips

• @filename → Bind specific file context

• #content → Write to CLAUDE.md project memory (persistent effect)

First step for any new project: always /init + /memory to refine; before large tasks: always /plan; for long conversations: use /compact + /context to monitor tokens.

Among the commands above, the ones I use most are /plan and Shift + Tab (switching to auto permission mode).

Why these two?

1. /plan (Planning Mode) — My biggest personal efficiency gain

I almost always do /plan first for large tasks (refactoring, adding features, architecture changes).

It forces Claude to output only a plan, not directly modify code. This lets me review the overall approach, risks, and step-by-step plan before deciding whether to execute. It avoids chaotic changes that lead to rework later.

Often after planning I’ll also do /rewind to roll back, then execute formally — the workflow is very clean.

2. Shift + Tab (Quickly switch to auto mode)

Honestly, like many people, I hate manually confirming permissions every time a file is changed.

Right after starting a project, I spam Shift + Tab to switch to auto, so it can freely read and write. After that, I don’t have to worry about permissions most of the time — the experience is much smoother (though I’m more cautious with important projects).

What’s the command you use most?

References:

Claude Official, https://code.claude.com/docs/en/commands

Blog Park, https://www.cnblogs.com/lf109/p/18974129