Claude Code Workflow Cheatsheet

Getting Started

Install Claude Code, initialize a project, and generate your starter memory file.

curl -fsSL https://claude.ai/install.sh | bash

cd your-project
claude
/init

Scans your codebase and proposes an initial project memory.
Best first step before prompting on a large repository.

Understanding CLAUDE.md

`CLAUDE.md` acts like persistent project memory and is loaded automatically at session start.

What Why How

What it stores Tech stack, directory map, architecture, gotchas, and team-specific conventions.

Why it matters It reduces repeated explanation and keeps workflows consistent across sessions.

# Project: MyApp FastAPI REST API + React SPA + Postgres ## Commands npm run dev npm run test npm run lint ## Architecture /app Next.js App Router pages /lib shared utilities /prisma DB schema & migrations

Memory File Hierarchy

Global

~/.claude/CLAUDE.md Shared defaults across all projects.

Root

./CLAUDE.md Parent project memory for the repo.

Child

./frontend/CLAUDE.md Scoped context for a subfolder or domain.

Keep each file concise, ideally under 200 lines.
Subfolder files append context instead of replacing parent rules.
Never overwrite higher-level memory with narrow implementation detail.

Best Practices

Define the desired outcome before asking for changes.
Be explicit about constraints and non-goals.
Reference docs with `@filename` when exact context matters.
Add workflow rules to memory once you repeat them twice.
Commit to Git often and keep generated context shareable.

Project File Structure

your-project/
├─ CLAUDE.md
├─ .claude/
│  ├─ settings.json
│  ├─ settings.local.json
│  └─ skills/
│     ├─ code-review/
│     │  └─ SKILL.md
│     ├─ testing/
│     │  └─ SKILL.md
│     └─ helpers/
├─ commands/
│  └─ deploy.md
├─ agents/
│  └─ security-reviewer.md
├─ arc/
└─ .gitignore

Adding Skills

Skills are markdown guides Claude can auto-invoke using natural language descriptions.

Project skill

.claude/skills/<name>/SKILL.md

Personal skill

~/.claude/skills/<name>/SKILL.md

name: testing patterns description: Jest testing patterns allowed-tools: Read, Grep, Glob # Testing Patterns Use `describe + it + AAA` pattern Use factory mocks

Skill Ideas For Engineers

Code review
Testing patterns
Commit messages

Docker deploy
Codebase visualizer
API design

Setting Up Hooks

Hooks provide deterministic callbacks for linting, validation, and notification flows.

PreToolUse PostToolUse Notification

"hooks": { "PreToolUse": [ { "matcher": "Bash", "hooks": [ { "type": "command", "command": "./scripts/src.sh", "timeout": 5 } ] } ] }

Exit 0 = allow Exit 2 = block

Permissions & Safety

{ "permissions": { "allow": ["Read:*", "Bash:git:*"], "deny": ["Read:env:*", "Bash:sudo:*"] } }

Prefer explicit allows for repeat-safe tasks.
Block environment secrets and privileged shell commands by default.

The 4-Layer Architecture

L1 — CLAUDE.md Persistent context and project rules.

L2 — Skills Auto-invoked knowledge packs and playbooks.

L3 — Hooks Safety rails, checks, and deterministic automation.

L4 — Agents Subagents with focused scopes and their own context.

Daily Workflow Pattern

cd project && claudeStart in repo context

Shift + Tab + TabPlan Mode

Describe feature intentState outcome and constraints

Shift + TabAuto Accept

/compactCompress context

Esc EscRewind

Commit frequentlyKeep work checkpointed

New session per featureReduce context drift

Quick Reference

/initGenerate `CLAUDE.md`

/doctorCheck installation

/compactCompress session context

Shift + TabChange modes

TabToggle extended thinking

Esc EscOpen rewind menu