Bootstrap a new project with a working CLAUDE.md, settings.json, and .gitignore for Claude Code
Add defensive hooks that block rm -rf on protected paths.env writes, and forced pushes
Drop in read-only code-auditor and security-auditor sub-agents for scoped reviews
Install slash commands like /ship, /audit-quick, /standup, and /coach in any repo
Mostly copy-paste, but you must chmod the hook scripts and edit CLAUDE.md to match your real stack before Claude Code picks it up correctly.
This repository ships a bilingual guide for configuring Anthropic's Claude Code, the coding assistant built into Anthropic's official CLI and IDE integrations. The guide comes in two 33-page PDFs, one in English and one in French, with the same content in both. The English version is titled the Claude Code Handbook and the French version is titled Le Code du Claudeur. The guide is split into eight chapters and includes a 30-minute checklist on the last page for readers who want to jump straight in without reading the full document. The PDFs are checked into the repository alongside their HTML source files. The other half of the project is a set of copy-paste templates, organised under a templates folder. They cover the core files Claude Code reads at startup: a CLAUDE.md that holds the project's permanent context, a .gitignore sample with lines to append, a settings.json with permissions and hooks, optional COMMANDS.md and PATTERNS.md indexes, a hooks folder with five shell scripts (pre-tool-guard, post-edit-format, session-start, activity-log, and a coach-suggest for the Stop event), an agents folder with two read-only audit agents (code-auditor and security-auditor), a commands folder with slash commands like /ship, /audit-quick, /standup, /coach, /coach-mute, and /coach-on, and a small agent-memory folder describing a cross-session memory pattern. Installation is four steps from a project root: copy CLAUDE.md and the .claude folder in, append the .gitignore additions, mark the hooks executable with chmod, and edit CLAUDE.md to reflect the project's actual stack. After that, launching Claude Code from inside the project picks everything up automatically. The author lists what the reader gets in return: permanent project context loaded at every session, defensive hooks that block dangerous commands like rm -rf on protected paths or .env writes or forced pushes, auto-formatting after edits, a git status recap when a session opens, scoped sub-agents for security and code-quality reviews, one-line slash commands, and an optional Coach mode that suggests the next step without polluting the conversation. The guide applies to all six Claude Code surfaces (terminal, VS Code, JetBrains, desktop, web, iOS). The project is MIT licensed.
Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.