All Modules & Lessons

Complete overview of all available modules and their lessons

← Back to The Loop

Module 0 - Watch this first

1 Lesson

A 30-minute walkthrough of the whole loop before you touch anything: what Level 2 looks like running, so you know what you're installing before you install it.

0 / 1 completed

Module 3 - When something breaks

15 Lessons

Fifteen symptoms, indexed by what you're seeing when the loop stops behaving, or when it's behaving fine and you want a different amount of it. Find yours, fix it, move on.

0 / 15 completed
Nothing ran overnight
Neither loop fired because the machine went to sleep. The fix, per OS.
Start →
It ran at 3am and I didn't want it to
How to disable both loops and schedule around weekends, holidays, and focus blocks, and why doing that is a virtue.
Start →
It opened a PR I can't merge
GitHub says there's a conflict. Ask the agent to make it mergeable and skip the rebase theory.
Start →
The queue is empty
An empty queue is the loop telling you you're out of thought-through work. That's the most useful signal it produces.
Start →
It asked questions instead of doing the work
The readiness gate working as designed. How to read the questions and answer them well.
Start →
It guessed and got it wrong
The gate let something through at 85% that shouldn't have. How to tighten the rubric and what to add to the task.
Start →
I started a brand new repo and it worked
Confirmation, plus what a fresh repo needs in its first task so the agent isn't working blind.
Start →
gh auth / SSH / .env
Secrets and environment, stated once, properly. Tests need .env; the agent needs GitHub auth; never let the agent read or print your env.
Start →
It never picked up my task file
The queue is a directory, so a task that never gets picked up is almost always in the wrong directory or carrying the wrong status. Four checks, in order.
Start →
Claude Code updated and my loop broke
How to diagnose a loop that stopped after a tooling update. This is the recurring maintenance cost of running a loop.
Start →
I want this somewhere else
port-my-loop: take the whole loop to a hosted tracker or a different agent. You own the setup, so you can move it.
Start →
I want the PRs to stop for a while
Quitting Claude pauses every scheduled fire immediately. No toggle hunting, nothing lost, nothing queued up behind it.
Start →
I want more PRs than one an hour
Invoke context-loop or next-loop directly, by task, whenever you're sitting there ready to review, without waiting for the hourly fire.
Start →
I want a different model on each half
context-loop is a reasoning job and next-loop is a transformation job. Give each the model that fits, and do your own ideation on the biggest one you've got.
Start →
I don't trust it to run unattended yet
Driving context-loop or next-loop by hand collapses the round trip an hourly fire has to wait out. There's no separate interactive skill for it.
Start →