ClaudeMap

My Claude Code Setup

A complete, working CLAUDE.md memory-bank starter: config plus the hooks, commands, and subagents that make it run.

Prompt Librariesclaude-mdsetupmemory-bankhooks

centminmod's personal Claude Code configuration, kept runnable end to end: a CLAUDE.md memory-bank structure — now with 3 modernized templates following Anthropic best practices (progressive disclosure, dual-memory architecture) — plus the supporting hooks, custom commands, statuslines, and subagents, each documented. Less a template to admire, more one to clone — the author exercises all of it daily (repo still actively updated, last pushed October 2026).

What it does

  • CLAUDE.md as a memory-bank system: a set of CLAUDE-*.md files the model updates after each completed task, so context survives across sessions — after a successful task you just say "update memory bank"
  • 3 modernized CLAUDE.md templates (2026-05) following Anthropic best practices: progressive disclosure (memory files read on-demand), dual-memory architecture, standalone behavioral rules — the live CLAUDE.md uses Template 3; a migration prompt converts existing files
  • Documented subagents: memory-bank-synchronizer, code-searcher (with an opt-in Chain-of-Draft mode claiming ~80% token reduction), get-current-datetime, ux-design-expert, zai-cli and more — each with purpose and usage
  • A statusline bash script that reads Claude Code's stdin JSON: model, context-window metrics, cost, progress bar, git branch
  • Curated MCP server list, plugin marketplaces (Safety Net destructive-command guard, Z.AI usage query, Cloudflare skills), and git-worktree launchers for parallel Claude/Codex sessions

When to use it

Clone it when "how should I structure CLAUDE.md?" has stalled your setup: the memory-bank loop is the most complete worked example of context persistence for Claude Code. The template section doubles as a migration path — paste the provided prompt and Claude Code converts your existing CLAUDE.md to the new format while preserving project-specific content.

Getting started

Clone the repo and copy what you need into your project's .claude/ and CLAUDE.md:

git clone https://github.com/centminmod/my-claude-code-setup.git

Then launch Claude Code in your project and run /init — it analyzes the codebase and populates the memory-bank files per the CLAUDE.md instructions. macOS users keeping the bundled settings.json need terminal-notifier; faster tools are assumed: brew install ripgrep fd jq

Things to know

It is one person's daily driver, not a neutral framework: the Brisbane (GMT+10) timezone is baked into the datetime subagent, several pieces are macOS-specific, and the README mixes in Z.AI promotional content (discount codes, GLM plan pitches) alongside the Claude setup — take the structure, audit the rest. The documented memory-bank loop ends with committing the updated memory files to git.

Repository

centminmod/my-claude-code-setup

Charted

Related on the map

  • Version-by-version teardown of Claude Code's system prompts and 27 built-in tool descriptions, updated with every release.

    • system-prompt
    • claude-code
    • analysis
    • prompt-engineering
  • A single-file CLAUDE.md distilling Karpathy's coding observations into four working rules (211k stars).

    • claude-md
    • rules
    • prompt-engineering
    • best-practices
  • prompts.chat

    Prompt Library

    The de-facto standard ChatGPT role-prompt collection (f.k.a. Awesome ChatGPT Prompts).

    • prompts
    • awesome-list
    • roles
    • community