Agentic Dev Environment
A reproducible, AI-augmented development setup: version-controlled dotfiles, custom Claude Code commands, and a multi-model workflow.
- 6 custom Claude Code commands
- Reproducible fresh-machine bootstrap
- Multi-model Claude + Codex workflow
Context
A dev environment is a system, and an un-versioned one is a liability. A new laptop, a wiped install, or context lost between sessions all cost time that good tooling buys back. I’d rather pay that cost once. So I treat my own setup like any other project: versioned, reproducible, documented well enough that a fresh machine comes up through a documented, repeatable bootstrap.
What I built
An AI-augmented development setup in two halves. A version-controlled dotfiles repo that reproduces my environment on a clean machine, and a structured Claude Code workflow that makes AI assistance consistent across every project instead of starting cold each time.
Architecture
Configuration lives in the repo and symlinks into place, so the repo is the
single source of truth and edits flow straight back to version control. The Claude
Code half adds structured context files (a per-project CLAUDE.md, session notes)
plus custom slash commands for the recurring work: project init, commits, session
handoff. A multi-model setup brings Codex in alongside Claude where it fits.
Technical highlights
- Reproducible by symlink. A documented bootstrap (clone, symlink, Brewfile) links every config into place. The repo is the source of truth, so there is no drift between what runs and what’s checked in.
- Context as a first-class artifact. Structured
CLAUDE.mdand session-note files keep AI assistance grounded across long-running work, so the model resumes with state instead of guessing. - Six custom commands. Slash commands wrap the repetitive workflows (init, commit, handoff) into one consistent, repeatable interface I can trust to behave the same way every time.
Tradeoffs
Personal infrastructure only pays back over time. Measure a single afternoon and it’s pure overhead. Measure across every project since and the leverage compounds. I made that bet deliberately.
Outcome
A development environment that rebuilds itself on a new machine and keeps AI-assisted work consistent and repeatable. It’s the backbone behind everything else in this portfolio.