dot-ai

Personal backup for my agentic coding setup.

This repo stores the config for:

• Claude Code (~/.claude)
• OpenCode (~/.opencode)

Everything is versioned with Git so I can sync my setup across machines like dotfiles.

---

AI Coding Stack

A clean AI coding setup works best when each layer has a distinct role: orchestration, context delivery, code execution, graph awareness, and quality control.

Layer	Tooling	Primary Role
Session orchestration	gstack	Coordinates the coding environment and workflow scaffolding
Context compression	lean-ctx + Caveman plugin	Keeps prompts lean while preserving useful project state
Context retrieval	context7	Supplies structured external or project context
Persistent memory	claude-mem	Stores and retrieves relevant observations across sessions
UI motion and polish	emilkowalski/skill	Improves animations, design, code quality, and UI performance
Frontend taste system	impeccable.style	Adds stronger frontend design judgment and higher-quality UI decisions
Reference-first anti-slop	tasteskill.dev	Forces reference-driven design choices before building UI
Semantic code search	ck	Local-first hybrid search to replace regular grep calls
Codebase mapping	Graphify	Builds relationship awareness across files and symbols
Quality gates	Posttool hooks + Ruff/ESLint	Enforces hygiene after generation or modification

---

Structure

$HOME/dot-ai
├── .claude/       # mirror of ~/.claude
├── .opencode/      # mirror of ~/.opencode
├── scripts/
│   ├── backup.sh  # one-shot backup script
│   └── restore.sh # one-shot restore script

---

Prerequisites

The backup and restore scripts require:

• rsync — for syncing directories
• git — for version control

Installing dependencies

Arch Linux

sudo pacman -S rsync git

Ubuntu / Debian

sudo apt update && sudo apt install rsync git

Fedora

sudo dnf install rsync git

macOS

# rsync is pre-installed on macOS
brew install git   # if not already installed

Windows (WSL)

sudo apt update && sudo apt install rsync git

---

First-time setup (local)

1. Create the repo directory:

mkdir -p ~/dot-ai
cd ~/dot-ai
git init
git checkout -b main

2. Add the scripts (copy backup.sh and restore.sh into ~/dot-ai) and make them executable:

chmod +x ~/dot-ai/scripts/backup.sh
chmod +x ~/dot-ai/scripts/restore.sh

3. Run an initial backup:

~/dot-ai/scripts/backup.sh

4. Commit the initial state:

git add .
git commit -m "chore: initial Claude + OpenCode backup"

---

Connecting to GitHub (or any remote)

1. Create a private repo called dot-ai on GitHub / GitLab. [web:34][web:38]

2. Add it as a remote and push:

cd ~/dot-ai
git remote add origin git@github.com:<your-username>/dot-ai.git
git push -u origin main

Now this repo acts as your cloud backup and source of truth. [web:34][web:38]

---

Backing up configs (daily use)

Whenever you tweak Claude Code or OpenCode settings, skills, or agents:

~/dot-ai/scripts/backup.sh

The script:

• Copies ~/.claude into ~/dot-ai/.claude/
• Copies ~/.opencode into ~/dot-ai/opencode/
• Commits any changes with a standard message

---

Restoring on a new machine

On a fresh machine:

git clone git@github.com:<your-username>/dot-ai.git ~/dot-ai
cd ~/dot-ai
./scripts/restore.sh

The script:

• Pulls latest changes
• Restores .claude/ → ~/.claude
• Restores .opencode/ → ~/.opencode

Then restart your editor / agents and you should have the same setup.

---

Security notes

• Do not store API keys or secrets directly in these folders.
  • Prefer .env files, OS keychains, or a separate secrets manager.
• If you must store any config with tokens, add them to .gitignore before committing. [web:38]

---

Extending later

You can later add more agent configs:

• Add cursor/, aider/, etc. directories
• Update backup.sh and restore.sh to sync those folders too

This keeps all your agentic coding tools under one versioned umbrella.