Progress will not be saved. This browser is blocking storage (often private browsing). You can still learn, but closing the tab loses your progress.
Config templates
16 templates for CLAUDE.md, skills, subagents, hooks, .mcp.json and plugins, tested with Claude Code. Copy or download, then adapt to your project.
CLAUDE.md
- CLAUDE.md for a Go project
Tells Claude how to build, test and vet the code and how errors are handled in Go style. Fits a backend service or a command-line tool written in Go.
CLAUDE.md - CLAUDE.md for a monorepo
Root CLAUDE.md for a monorepo with several packages. The root file covers only what is shared; each package keeps its own CLAUDE.md, which Claude reads only when it works there, so context stays small.
CLAUDE.mdpackages/{{package}}/CLAUDE.md - CLAUDE.md for a Node.js and TypeScript project
Short instructions telling Claude how to build, test and lint a TypeScript project and which conventions to follow. Replace the {{…}} spots with your real commands and folders, then delete what you do not use.
CLAUDE.md - CLAUDE.md for a Python project
Tells Claude which virtual environment to use, how to run tests and lint, and the typing conventions of a Python project. Change the commands to match your tools (uv, pip, poetry).
CLAUDE.md - CLAUDE.md for an iOS app in Swift
Tells Claude how to build and test an iOS app from the command line with xcodebuild, plus SwiftUI and concurrency conventions. Change the scheme and simulator to match your project.
CLAUDE.md
Hook
- Hook that blocks dangerous shell commands
Before Claude runs a Bash command, the hook checks it and blocks dangerous patterns such as recursive deletes at the root, force pushes, piping a download into a shell or dropping database tables. Claude is told why so it can change course.
.claude/hooks/block-dangerous.sh.claude/settings.json - Hook that formats code after each edit
Every time Claude edits or creates a file, the hook runs the formatter that fits its extension (Prettier, Ruff, gofmt), so code stays formatted without asking in the prompt.
.claude/hooks/format-file.sh.claude/settings.json - Protect .env files
Two layers that keep the secrets in .env away from Claude. Deny rules in permissions stop the file tools from reading or editing the usual secret files (.env, .env.local, .env.production…); the hook also blocks Bash commands that mention them, such as cat or source. .env.example stays readable.
.claude/hooks/protect-env.sh.claude/settings.json
.mcp.json
- Filesystem MCP server
The official MCP server that reads and writes files inside the folders you name. Claude Code already reads project files, so this template is mostly for learning how a local MCP server is configured; to add another folder, /add-dir is usually enough.
.mcp.json - GitHub MCP server
Connects Claude Code to GitHub's remote MCP server to read issues, pull requests and reviews inside a session. The token comes from an environment variable, so the .mcp.json file can be committed for the whole team.
.mcp.json - PostgreSQL MCP server (read-only)
Lets Claude query a PostgreSQL database through DBHub, a local MCP server started with npx. Use a database user that can only read, so Claude cannot change data.
.mcp.json
Plugin
- A minimal plugin
The smallest Claude Code plugin skeleton, a manifest and one skill. Use it to package your team's skills, subagents or hooks and share them through a marketplace.
{{plugin_name}}/.claude-plugin/plugin.json{{plugin_name}}/skills/hello/SKILL.md
Skill
- A first SKILL.md
Skeleton of a project skill. Claude reads the description to decide on its own when to use the skill, so say clearly what it does and which requests should trigger it. You can also run it yourself with /skill-name.
.claude/skills/{{skill_name}}/SKILL.md
Subagent
- Debugging subagent
A subagent that hunts down the root cause of a bug: reproduce, narrow it down, then propose the smallest fix with evidence. Handy for a hard bug you want to keep out of the main conversation.
.claude/agents/debugger.md - Code review subagent
A read-only subagent that reviews uncommitted changes and reports problems by severity. It cannot edit files, so it is safe to run after Claude finishes writing code.
.claude/agents/code-reviewer.md - Test-writing subagent
A subagent that writes tests for the code that just changed, runs them and reports. It only edits test files, never the code under test, so fixing what it finds stays your call.
.claude/agents/test-writer.md