The repo I use to ship
agentic code without slop.
The actual structure behind PalinDragon, Bartley Editions, and the Encyclopedia of Agentic Coding Patterns. Eleven steps, four canonical agents, the policies, the planner / reviewer / coder / critic loop, all in one repo. Free. Yours in a minute.
No spam. One email a day for five days, then quiet. Unsubscribe in one click.
Five pieces. All free.
Not a brochure. Not a PDF that wishes it were a book. The actual working materials, indexed so you can clone, read, and ship inside an hour.
The Starter Repo
Public GitHub template, plus a direct-download zip. The full eleven-step methodology, four canonical agents (planner / reviewer / coder / critic), policies catalog, append-only log discipline, the /kickoff orchestrator skill, cross-harness wiring for Claude Code and Codex. Clone it, run it, ship with it. MIT-style.
"How to Read This Repo in 25 Minutes" PDF
Where to look first, what to skim, what to memorize, what to wire into your editor settings. Skips the "I cloned it and don't know where to start" trap that kills most dev MIFGEs.
12-Minute Working Session · Wolf on Video
Not a polished webinar. Wolf opens a fresh idea, runs /kickoff, you watch the planner → reviewer → coder → critic loop in motion. The first time you see this, the methodology stops being abstract.
5-Day Email Walkthrough
One short email per day for five days. Each one points at a specific file or pattern, with the reasoning. Day 1: the slop problem and where structure starts. Day 5: the one thing the repo alone can't give you. Read in 3 minutes, apply in 10.
Alumni Discord · Guest Tier
Read-only access to the channels where Liftoff alumni post what they're shipping. Watch builders thirty days ahead of you ask the questions you're about to ask. Promote yourself to full tier any time.
Total · Yours by morning
One email field. Repo + reading guide land in your inbox inside a minute. Video and Day 1 of the walkthrough arrive later today.
The starter is the foundation. The cohort is the application.
The repo encodes what works: the methodology, the policies, the loop. Anyone who can read English and clone a repo can use it. I want every serious builder to have it.
Where it stops short, and where I sit a few weeks a year with a small group of builders, is the application. Watching someone use the methodology on their own fresh idea. Catching the failure mode that's specific to their problem, before they spend two months in it. That's the Liftoff Clinic. The starter is the bait. I'm not pretending otherwise.
Two kinds of builders.
If you've tried "vibe coding"
You let an agent vibe its way through your project. You ended up with something fragile that breaks when you ask for anything new, and you can't tell why. The methodology is the antidote. Structure the agent reads before it writes, a critic loop that challenges its work before it lands in the project, a log so you never re-derive yesterday's decisions.
If you code already, but you're new to agents
You can write the code. What you're new to is delegating the typing while keeping the judgment. The four canonical agents and the planner / reviewer / coder / critic loop is the discipline that lets the agent handle the easy 80% without you losing control of the 20% that matters.
Wolf McNally.
I've shipped agentic code on four projects this year. The methodology in this repo is what I actually use, not a clean-room version. The same CLAUDE.md structure runs Hubert, PalinDragon, Bartley Editions, and the Encyclopedia of Agentic Coding Patterns. Before this I spent twenty years shipping software at Apple, Disney, EA, and a handful of other places. I finally got tired of watching agents write code we couldn't keep, so I started writing down what I learned. That's the starter.
FAQ
What is this, exactly?
A working repository template for building software with AI coding agents under a structured methodology. Eleven steps, four agent roles, a phase orchestrator, an append-only log, a policies catalog. Harness-agnostic: works with Claude Code, Codex CLI, and any other agent host that reads project-level instructions.
Why give away the actual stack?
Because the repo can't make decisions for you. It encodes the structure; the judgment is yours. I want every serious builder to have the structure. The handful who want a few hours with me applying it to their own fresh idea, that's the Liftoff Clinic. Fair trade.
Do I need Claude Code, or does it work with Codex / other agents?
Any harness that reads project-level instructions and agent definitions. The repo ships canonical files under .claude/ with Codex mirrors under .codex/ and .agents/. Adding a third harness is documented.
Am I going to get pitched constantly?
One email a day for five days. Then I'm quiet. After that you hear from me when a new cohort opens or when I ship something I think you'd want to see. Unsubscribe in one click, any time.
What if I already have a project halfway built?
The starter works on anything. Drop the structure into an existing repo and start running phases from where you are. The clinic, on the other hand, is greenfield-only: the methodology is easiest to internalize on an idea that isn't already fighting half-baked patterns.
Is this for solo builders or teams?
Both. The methodology was developed running solo across four projects, but every piece (briefs, policies, phase logs, plan/INDEX.md) is built to be read by other humans as well as agents. It scales to teams by being unambiguous about what's binding and where the source of truth lives.
Get the starter. Start tonight.
Repo + reading guide + working-session video + 5-day walkthrough + Discord access. One email. Five free days, then quiet.
No spam. Unsubscribe any time.