Day in Our Data · Saturday, October 3, 2026, 11 a.m. to 3 p.m.

Claude Code cheat sheet

The commands, keys, and prompt patterns you will use all day, on one page. Print it or keep it open in a tab.

Starting and resuming

Type thisWhat it does
claudeStart a session in the current folder.
claude "explain this repo"Start a session with a first prompt already typed.
claude -cContinue your most recent conversation in this folder.
claude -rPick an earlier conversation to resume.
claude -p "summarize data/README.md"Ask one question, print the answer, exit. Handy in scripts.
claude --versionShow the installed version.
claude doctorDiagnose install and settings problems.
claude updateUpdate now instead of waiting for the automatic update.

Slash commands inside a session

Type / at the prompt to see every command, including ones added by plugins.

Type thisWhat it does
/helpList commands and get help.
/initRead the project and write a CLAUDE.md of notes Claude rereads every session.
/clearStart a fresh conversation. Use it when you switch tasks.
/compactSummarize the conversation so far to free up room for more work.
/contextSee how full the conversation's memory is.
/modelSwitch model, for example to Opus for hard problems.
/memoryOpen and edit the CLAUDE.md memory files.
/permissionsChoose which tools and commands Claude may use without asking.
/mcpSee and manage connected MCP servers (connectors).
/pluginBrowse, install, and manage plugins.
/agentsCreate and manage specialist subagents.
/resumeSwitch to an earlier conversation.
/rewindRoll the conversation and your files back to an earlier point.
/loginSign in, or switch accounts.

Keyboard shortcuts

Type thisWhat it does
EnterSend your message. While Claude is working, it queues the message for next.
Shift+EnterNew line without sending (most terminals). Or end a line with \ then Enter.
EscStop Claude mid-task. Your conversation is kept.
Esc EscClear what you typed, or rewind to an earlier message.
Shift+TabCycle permission modes: normal, auto-accept edits, and plan (research only, no changes).
@Mention a file or folder: @data/trees-oak-park.csv what columns are here?
!At the start of a line: run a shell command yourself and show Claude the output. ! ls data
Up / DownScroll through what you typed before.
Ctrl+RSearch your earlier prompts.
Ctrl+OShow the full transcript, including tool output.
Ctrl+TShow or hide Claude's task checklist.
Ctrl+CInterrupt or clear input. Twice to exit.
Ctrl+DExit Claude Code.

Permission modes, explained

Claude asks before it edits files or runs commands. Shift+Tab changes how often it asks.

Permission modes
ModeWhat Claude doesUse it when
NormalAsks before each edit and command.You are learning, or the change matters.
Accept editsEdits files without asking; still asks before commands.You trust the plan and want speed.
PlanReads and researches only, then presents a plan. Changes nothing.The start of every new feature. Read the plan, then say go.

CLAUDE.md: the project's memory

A plain text file at the top of your project. Claude reads it at the start of every session. /init writes a first draft; edit it as you learn what matters.

A starter CLAUDE.md for a Day in Our Data team
# Project: Urban forest diversity map (starter project 11)

## What we are building
A one-page map of Oak Park's public trees, coloured by genus,
flagging blocks that break the 10-20-30 diversity rule.

## Data
- data/trees-oak-park.csv (public, Village of Oak Park)
- Always cite the source file and date under the chart.

## How we work
- Python 3.11. pandas for cleaning, folium for the map.
- Plan first; keep each change small.
- Run `python -m pytest` before every commit.
- Never commit keys, tokens, or .env files.

Prompts that work

Prompt patterns
PatternExample
Give context and a goalI'm on starter project 4. Goal: a map of school routes with crash points. Audience: parents.
Ask for a plan firstBefore writing code, propose a plan with 3–5 steps and what you'd skip if short on time.
Make it smallJust step 1. Stop when there's something I can open and look at.
Point at filesUse @data/crashes-bike-ped-oak-park.csv. What are the columns and which are messy?
Ask it to check itselfWrite a test that proves the totals match the raw CSV, then run it.
Ask whyExplain that change like I'm new to Python.
RecoverThat broke the map. Undo your last change and try a simpler approach.
Wrap upWrite README.md with exactly these headings: The question, The data, How we got the numbers, How to use it, Status and next steps. Only describe what is really in these files. (why)