Skip to content

claude-code-cheatsheet-tutorial.md

Latest commit

 

History

History
209 lines (174 loc) · 7.41 KB

claude-code-cheatsheet-tutorial.md

File metadata and controls

209 lines (174 loc) · 7.41 KB
title Claude Code Cheatsheet Tutorial
tags
tools
claude-code
last_updated 2026-04-22

Claude Code Cheatsheet Transcription and Tool Tutorial

Claude Code Cheatsheet (Text Transcription)

Claude Code Cheatsheet

Slash Commands

  • /clear — Clears chat history
  • /agents — Access and manage agents
  • /model — Select the AI model
  • /mcp — Manage MCP servers
  • /cost — Show token usage
  • /permissions — View permissions
  • /bashes — List background tasks
  • /hooks — Hook configs for tools
  • /memory — Edit CLAUDE.md files
  • /plugin — Manage Code plugins
  • /init — Initialize project
  • /review — Request code review
  • /config — Clears chat history
  • /sandbox — Enable sandbox bash tool

Keyboard Shortcuts

  • ! — Run commands directly
  • Shift + Tab — Auto-Accept
  • @ — File path mention
  • Shift + Tab + Tab — Planning mode
  • Esc — Interrupt Claude process
  • Cmd + Esc — Quick launch in IDE
  • / + Esc — Quick Escape
  • Ctrl + T — Toggle syntax highlighting
  • Ctrl + R — Reverse search history
  • Ctrl + U — Delete entire line
  • Ctrl + V — Paste Image from clipboard
  • Alt + T — Toggle Extended Thinking
  • Esc + Esc — Rewind the code
  • Ctrl + B — Background running tasks

Best Plugins

  • Superpowers — Claude Code core super skills library — github.com/obra/superpowers
  • Context-7 — Up-to-date code documentation — github.com/upstash/context7
  • Exa Search — Web search and web crawling! — github.com/exa-labs/exa-mcp-server
  • Playright — End-to-end testing for web apps — github.com/lackeyjb/playwright-skill
  • Beads — A memory upgrade for your coding agent — github.com/steveegge/beads
  • Front-end design — Improve agents’ design skills — github.com/anthropics/claude-code/blob/main/plugins/frontend-design
  • PR Review Toolkit — App PR Automation agents — github.com/anthropics/claude-code/tree/main/plugins/pr-review-toolkit
  • Code Simplifier — Simplifies large code bases — claude plugin install code-simplifier

Essential Commands

  • claude -p --max-turns 3 "query" — Limit conversation turns (In this example 3)
  • claude --continue — Continue the current session
  • claude --add-dir /path/to/project — Validate directory paths
  • claude -p "query" --output-format json — Print output in different output formats
  • claude -r — Resume a previous conversation
  • claude --version — Check the current version
  • claude update — Update the Claude into its latest version
  • claude config set -g theme dark

Best Practices to use Claude code

Parallel Execution

  • Run 5 claudes in parallel in numbered terminal tabs
  • Use System notifications (iTerm2) for input alerts
  • Run 5–10 or more on claude.ai/code alongside local
  • Hand off with & or --teleport

Slash Commands

  • /aggregatel for aggregating logs (example not real)
  • /visualz for creating webpage from images (example not real)
  • Create Commands for repeated workflows
  • Store in .claude/commands/, check into git
  • Example: /commit-push-pr used dozens of times daily
  • Use inline bash to pre-compute for speed

Permissions

  • Don’t use --dangerously-skip-permissions
  • Use /permissions to pre-allow safe commands
  • Store in .claude/settings.json, share with team
  • For sandboxed: --permission-mode=onAsk

Operating on Claude.md

  • Share single Claude.md per repo, checked into git
  • Whole team contributes to add mistakes on multiple times/week

Verification

  • Use Chrome extension to test UI changes for your apps
  • Verify according to your domain: bash, tests, browser, simulators

Long-Running tasks

  • Use Bg Agent to very when done
  • Stop hook for deterministic verification
  • Use Ralph Wiggum for autonomous task execution

Plan Mode

  • Start most sessions in Plan mode
  • Iterate until you like the plan
  • Switch to auto-accept edits (Claude will 1-shot it)

Key File Locations

  • .claude/settings.json — Project Settings
  • .claude/agents/ — Your subagents for your projects
  • .claude/managed-settings.json — Your enterprise managed settings
  • .claude/commands/ — Your custom claude commands
  • .claude-plugin/plugin.json — Data about your plugins

Claude Code Tool Tutorial

1) Define the outcome and constraints

Start with a crisp goal, then add guardrails so Claude Code can act decisively.

Prompt template

TASK: [one-sentence goal]
CONTEXT: [feature area, repo path, or bug symptoms]
CONSTRAINTS: [no new deps, keep API stable, time/size limits]
FILES TO TOUCH: [paths or "unknown"]
DEFINITION OF DONE:
- [expected behavior or test]
- [docs or edge case requirement]
REQUEST: Provide a short plan, then implement.

2) Initialize the workspace

  • Open the repo root in your terminal before launching Claude Code.
  • Use claude --add-dir /path/to/project when the CLI needs to validate the workspace.
  • If you work with multiple repos, create a separate CLAUDE.md per repo to store stable context.

3) Plan before edits

Ask for a plan first so you can approve the approach before Claude Code changes files.

Example

Please propose a 3-step plan and confirm before editing any files.

4) Use slash commands to manage context

  • /memory to edit CLAUDE.md and lock in project conventions.
  • /permissions to pre-allow safe commands.
  • /review to request a post-change code review.

5) Execute in small, reviewable batches

  • Ask for a short diff summary after each batch.
  • Keep tasks scoped to 1–3 files when possible.
  • If the change is large, split it into checkpoints ("Phase 1: refactor", "Phase 2: features").

6) Validate outputs

  • Require tests or verification steps in the response.
  • If tests are skipped, ask for the reason and a manual verification alternative.

Verification checklist

  • Unit tests or local checks run
  • UI or integration behavior validated
  • Any warnings or TODOs captured

7) Capture reusable commands

Turn frequent tasks into custom commands stored at .claude/commands/.

Example

/commit-push-pr

8) Extend with plugins

  • Use plugin installs to scale specialist workflows (reviews, testing, design).
  • Keep a short list of approved plugins so teams stay consistent.

9) Troubleshooting quick hits

  • Claude misses context → add a short repo summary in CLAUDE.md.
  • Too many changes at once → ask for smaller batches or a plan checkpoint.
  • Permissions blocked → open /permissions and pre-allow the safe commands.

10) Example end-to-end session

TASK: Add a dry-run flag to scripts/deploy.sh.
CONTEXT: scripts/deploy.sh handles prod deploys.
CONSTRAINTS: Do not change default behavior; keep bash only.
FILES TO TOUCH: scripts/deploy.sh
DEFINITION OF DONE:
- --dry-run prints actions without executing
- help text documents the flag
REQUEST: Provide a short plan, then implement.

Expected response outline

### Plan
- Inspect scripts/deploy.sh for argument parsing and deploy steps.
- Add a --dry-run flag and guard execution blocks.
- Update the help output.

### Changes
- scripts/deploy.sh: parse --dry-run and conditionally skip execution
- scripts/deploy.sh: document the new flag in usage text

### Tests
- manual: run `./scripts/deploy.sh --help`
- manual: run `./scripts/deploy.sh --dry-run`

References