Claude Code Starter Kit

A minimal template for getting a new project set up well with Claude Code: a best-practices CLAUDE.md, one example skill, and pointers to Anthropic's published skills so you get useful capabilities working on day one. This kit isn't tied to any particular language or framework — it's meant to be forked into a fresh project.

Getting started / verify your setup

Before touching anything in this repo, confirm Claude Code itself is installed and authenticated.

1. Install (native install is recommended and auto-updates in the background):

# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

Homebrew (brew install --cask claude-code) and WinGet (winget install Anthropic.ClaudeCode) also work, but don't auto-update — you'll need to brew upgrade / winget upgrade yourself.

2. Authenticate — pick whichever matches how you're using it:

• Subscription login (recommended for individual use): run claude and follow the browser login prompt (Pro, Max, Team, or Enterprise account). Use /login inside a session to switch accounts later.
• API key (Console access, CI, or scripting): create a key in the Claude Console, then:

  export ANTHROPIC_API_KEY=sk-ant-...

  Gotcha: if you also have an active subscription login, a set ANTHROPIC_API_KEY takes precedence once approved. Run unset ANTHROPIC_API_KEY to fall back to your subscription, and check /status inside a session to see which credential is actually active.

3. Verify it works — run a quick smoke test:

claude -p "which model are you, and what's the current working directory?"

A response that correctly names the model and directory confirms install + auth are both working. Or start interactively with claude and ask what does this project do? — that also doubles as your first real read of this repo's CLAUDE.md.

Using this template

1. Clone/fork this repo (or copy CLAUDE.md and .claude/ into an existing project).
2. Open the project in Claude Code and run /init if you want Claude to scan your actual codebase and draft a starting point — then replace the example content in CLAUDE.md with whatever /init finds (or with your own knowledge of the project). Don't just keep /init's raw output; treat it as a first draft.
3. Prune ruthlessly. For every line, ask: "would removing this cause Claude to make mistakes?" If not, cut it.

What belongs in CLAUDE.md

✅ Include	❌ Exclude
Bash commands Claude can't guess	Anything Claude can figure out by reading code
Code style rules that differ from defaults	Standard language conventions Claude already knows
Testing instructions and preferred test runners	Detailed API documentation (link to docs instead)
Repo etiquette (branch naming, PR conventions)	Information that changes frequently
Architectural decisions specific to your project	Long explanations or tutorials
Developer environment quirks (required env vars)	File-by-file descriptions of the codebase
Common gotchas or non-obvious behaviors	Self-evident practices like "write clean code"

Keep CLAUDE.md itself short — it's loaded on every session. For domain knowledge or workflows that only come up sometimes, add a skill instead (see below); Claude loads those on demand.

Skills

This repo ships one example skill, changelog-entry, so you can see the SKILL.md format working locally. Run /skill-creator (see below) or write a new folder under .claude/skills/ to add your own.

For broader capabilities, Anthropic publishes a set of ready-made skills at anthropics/skills. Install what you need through the plugin marketplace:

/plugin marketplace add anthropics/skills
/plugin install document-skills@anthropic-agent-skills
/plugin install example-skills@anthropic-agent-skills

Category	Skill	What it does	License
Document	docx	Create/edit Word documents	Source-available
Document	pdf	Extract from and generate PDFs	Source-available
Document	pptx	Create/edit PowerPoint presentations	Source-available
Document	xlsx	Create/analyze Excel spreadsheets	Source-available
Creative/design	algorithmic-art	Generative/algorithmic art	Apache 2.0
Creative/design	canvas-design	Visual layout design on a canvas	Apache 2.0
Creative/design	theme-factory	Generate cohesive visual themes	Apache 2.0
Creative/design	frontend-design	UI/frontend design guidance	Apache 2.0
Creative/design	brand-guidelines	Apply brand/style guidelines	Apache 2.0
Dev/technical	mcp-builder	Scaffold MCP servers	Apache 2.0
Dev/technical	webapp-testing	Test web apps end-to-end	Apache 2.0
Dev/technical	web-artifacts-builder	Build web-based artifacts	Apache 2.0
Dev/technical	claude-api	Work with the Claude API/SDK	Apache 2.0
Dev/technical	skill-creator	Scaffold new skills	Apache 2.0
Enterprise/comms	internal-comms	Draft internal communications	Apache 2.0
Enterprise/comms	doc-coauthoring	Collaborative document drafting	Apache 2.0
Enterprise/comms	slack-gif-creator	Generate GIFs for Slack	Apache 2.0

The four document skills (docx/pdf/pptx/xlsx) are source-available reference implementations of what powers Claude's built-in document capabilities, not open source — install them rather than copying their code.

Beyond this kit

A few other Claude Code features are worth knowing about, even though this starter kit doesn't configure them:

• Subagents — specialized assistants with their own context, defined in .claude/agents/.
• Hooks — deterministic scripts that run at specific points (e.g. lint after every edit), configured in .claude/settings.json.
• Plugins — bundles of skills/hooks/agents/MCP servers, browsable with /plugin.

See the official best practices guide for more.