Loop & Gate

shane logsdon · loop & gate

The loop does the work.
You work the gates.

A small stack for working with an AI agent: one Foundation layer that remembers you across sessions, and three kits that stop an agentic loop at the points where a human has to decide.

Loop & Gate is the name for one specific way of working: the loop runs on its own, and it stops at a fixed set of gates where a human decides. It is the opinionated instance of what the field now calls loop engineering and human-in-the-loop control.

Start with the Foundation

MIT licensed, and it needs a paid Claude plan. Pro or Max; the free plan does not include Claude Code.

the idea

one pattern, four repos

Most people meet an AI agent as a blank chat box. It is brilliant, and it forgets you the moment the window closes. It will build whatever you ask, including the wrong thing, confidently. And it has no opinion about whether you actually shipped the work you meant to.

An agentic loop can run on its own most of the time. But it has to stop at a fixed set of gates where a human decides. The loop does the work between the gates; you work the gates. Each gate marks a place where the agent is weak: it can't tell a real problem from a loud one, it can't feel when an architecture will hurt in six months, it can't tell a demo from evidence, and left alone it will confidently build more than you asked.

The gates are the asset. The tools that fill the space between them are a snapshot, and every kit says so in its own reference files: named, dated, and disposable on purpose. When the tools change, the gates don't.

the four pieces

foundation first

The Foundation sits underneath all three kits. The kits work better when it is there, because they can read your voice and taste and write what they learn back into your memory. One of them requires it. The other two can start without it.

The Foundation
An agent that remembers: memory as plain markdown in an Obsidian vault, with no vector database, no server to stand up, and nothing to sign up for. It captures what you tell it, files it, and writes one honest lesson at the end of each session. A profile interview turns how you write and what good looks like to you into notes the kits can read. Set it up
The Build Kit
Eleven gates for an agentic build loop, sized by the master gate before any of them run. It sits on top of a build pipeline it does not ship: two free public plugins provide the loop between the gates. The Build Kit
The Grow Kit
The same method pointed at go-to-market: audience and the one claim, the design, the copy, the channel, and whether it worked. It requires the Foundation, because its design gate reads your taste profile and its copy gate reads your voice profile. The Grow Kit
The Accountability Kit
The loop pointed at your own follow-through: eight gates, five daily rituals, and a deferral engine that counts the focus you keep skipping and forces the call at three strikes. Runs on Claude alone, with nothing else to install. The Accountability Kit

how they fit together

build → grow → accountability

You build something with the Build Kit. You take it to market with the Grow Kit, whose feedback gate reads whether it actually landed and writes those signals back into the Foundation vault. The Build Kit's first gate reads the same signals to decide what is worth building next. The Accountability Kit runs underneath the whole thing, so a stalled build or a launch that never went out gets caught at the nightly reckoning instead of quietly slipping.

The Foundation is the shared bus in the middle. You do not need all four to get value: each kit stands on its own, with the one exception that Grow requires the Foundation. Start with what maps to your actual problem and add the rest when you feel the seam.

status

four public repos

All four repositories are public and MIT licensed. Each one installs with a /plugin marketplace add line of its own, or with a single pi install, and each can be cloned and used directly without a plugin at all.

A plugin install is not the full loop. The Foundation's improve skill rewrites the skill files as it learns, and that self-editing only persists in a clone, where the skills are your working copy. Installed as a plugin, the skills sit in a cache that is replaced on update, so the agent still evolves your memory but not its own instructions.

Two qualifiers the sources state and this page keeps: the Build Kit's own status note lists the profile glue and a final package pass as still ahead, and the Foundation's no-terminal walkthrough is described as new and still being validated with first users. The gate counts and ritual names on the kit pages are the parts that are settled.

where to go

six routes