v1.6.0 ยท Works with Claude Code

One standard your AI coding agents actually follow.

One shared rulebook you install once, so every agent plans, writes, tests, and ships the same way. Your CI and review rules make it stick.

run in Claude Code
/plugin marketplace add DouglasVulcano/zeroth-ai
/plugin install zeroth

The problem

Good standards are easy to write and easy to ignore

AI agents only follow the instructions you give them. Left to each project, those instructions drift.

A different setup in every repo

Each project grows its own instructions, so the same assistant behaves differently from one repository to the next.

Rules nobody checks

A guidelines document is easy to skip. When nothing verifies it, quality quietly slips over time.

Too much manual wiring

Getting it right means setting up templates, checks, and reviews by hand, again and again, for every repo.

The fix

One shared standard, installed once

A single source of truth that any AI agent applies the same way, on any project and in any language.

A clear rulebook

A readable spec that defines how work gets planned, built, tested, and shipped.

A skill that applies it

It plugs into Claude Code and brings the right rule at the right moment, loading only what a task needs.

A setup tool

One command adds the templates, checks, and guardrails to a repository for you.

What it covers

Four areas, one consistent standard

Applied whenever an agent starts, plans, or reviews work.

1

How work flows

Every change starts as a ticket and ships as a reviewed pull request. Clear history, no surprise commits.

2

How interfaces feel

Smooth screens: loading states, tasteful motion, and respect for accessibility and reduced-motion settings.

3

How quality is kept

Consistent checks for formatting, types, dead code, and tests, on any tech stack, before code is merged.

4

Which tools to use

A curated toolbox the agent reaches for, with a safety review before anything new is added.

Why it works

Guidance you can rely on

The standard is the advice. Your project's own checks and review rules are what enforce it, and the setup tool helps you turn those on.

Setup

Governance in one command

Adds issue and PR templates, code owners, a checks workflow, agent guides, and safety rules. It never overwrites your files without asking.

Enforcement

Helps you turn on the guardrails

It prints, and can apply for you, the branch-protection settings that require checks and a review before anything merges.

Safety

Sensible guardrails

Blocks clearly destructive commands and edits to secret files while an agent works, a simple safety net.

Any stack

Not tied to one toolset

The quality checks are defined as goals, then mapped to your language: JavaScript, Python, Go, Rust, Java, or .NET.

Trust

Tested on itself

The package checks itself on every change, and this project runs on its own standard from end to end.

Distribution

Version it, share it

Ships as a Claude Code plugin, so a whole team stays on the same version, and it works with any agent that reads a shared guide file.

Get started

Up and running in minutes

Pick the scope that fits, then complete the short setup path so the standard is not just advice.

1
Add the guardrailsRun the setup tool on a repository to add templates, code owners, a checks workflow, and safety rules. Preview first with a dry run.
2
Fill in your detailsAdd your stack's commands and your real code owners, the two things the tool cannot guess.
3
Turn on the checksRequire the checks and a review before anything merges. The tool prints the exact command, or applies it for you.

Plugin, recommended

/plugin marketplace add \
  DouglasVulcano/zeroth-ai
/plugin install zeroth

Personal install

git clone https://github.com/DouglasVulcano/\
zeroth-ai.git
cd zeroth-ai
bash install-skill.sh

FAQ

Common questions

What is this, in one sentence?

A shared rulebook for AI coding agents that you install once, so they plan, build, test, and ship code consistently across all of your projects.

Who is it for?

Anyone using AI coding assistants like Claude Code who wants consistent, reviewed, and tested output, whether you are a solo developer or a whole team.

Do I have to change my tools or language?

No. The quality checks are goals mapped to your existing stack, so it works with JavaScript, Python, Go, Rust, Java, .NET, and more.

Does it change my repository automatically?

No. It sets up the pieces and shows you the commands; you decide when to turn on the checks and review rules that enforce the standard. Nothing is changed without your confirmation.

Which AI agents does it work with?

The plugin and skill install into Claude Code. For other agents (Cursor, Copilot, Codex), point them at the AGENTS.md the setup tool generates (it imports zeroth.md) and they read the same standard. Enforcement lives in your repo (CI plus branch protection), so it applies the same way whichever agent writes the code.

Bring consistency to how your agents ship code

Install the plugin, set up a repo, and let your checks do the enforcing.