ai-ready

AI Ready

An Agent Skill that analyzes your repository and generates the configuration files AI agents need to contribute correctly. GitHub-native — it auto-discovers your repo's context, community health, and PR review patterns without you explaining anything.

Works in GitHub Copilot, Claude Code, OpenAI Codex, Cursor, and any other tool that supports the Agent Skills standard.

Quick Start

Any agent, one command

The open skills CLI installs into Claude Code, Codex, Cursor, Copilot, and 70+ other agents:

npx skills add johnpapa/ai-ready

That's it — no per-tool setup. Restart your agent afterward.

Native plugin installs

Prefer your tool's own plugin system? Each of these installs the same skill.

Tool	Install
GitHub Copilot CLI	copilot plugin install ai-ready@awesome-copilot
Claude Code	/plugin marketplace add johnpapa/ai-ready/plugin install ai-ready@johnpapa-ai-ready
OpenAI Codex	codex plugin marketplace add johnpapa/ai-readycodex plugin add ai-ready@johnpapa-ai-ready
Cursor	Copy skills/ai-ready/ into ~/.cursor/skills/ai-ready/
Anything else	Copy skills/ai-ready/ into ~/.agents/skills/ai-ready/

Claude and Codex each need two commands — the first registers the marketplace, the second installs the plugin from it.

Copilot CLI installs from the built-in awesome-copilot marketplace. Installing straight from a repo (copilot plugin install johnpapa/ai-ready) still works but is deprecated, so prefer the marketplace form.

~/.agents/skills/ is the vendor-neutral Agent Skills directory that Codex and Cursor both read, so the last row works for most tools.

If a marketplace mirrors this skill, it should point to a release tag such as v1.4.0, not a raw commit SHA.

Then type

make this repo ai-ready

Or invoke it explicitly:

Tool	Invoke
Claude Code, Cursor, Copilot CLI	/ai-ready
OpenAI Codex	$ai-ready

When installed as a plugin from a marketplace, Claude namespaces it as /ai-ready:ai-ready.

The skill analyzes your code, CI, tests, docs, and structure, then generates assets customized to your project — not generic templates.

Run it again anytime

The skill is safe to re-run. On the first run, it creates missing assets. On subsequent runs, it audits your existing AI-ready files against the current state of your codebase — flagging drift like outdated build commands, stale repo structure, or new PR review patterns that should become conventions. It never overwrites your files — it suggests updates and lets you decide.

Keeping updated

Tool	Update
skills CLI	npx skills update
GitHub Copilot CLI	copilot plugin update ai-ready
Claude Code	/plugin marketplace update johnpapa-ai-ready
Codex	codex plugin marketplace upgrade

Troubleshooting

1. Restart your agent. Most tools only scan for skills at startup.
2. Confirm it's installed. Run npx skills list, copilot plugin list, claude plugin list, or codex plugin list depending on how you installed it.
3. Check you finished both steps. Claude and Codex need a marketplace command and an install command.
4. Ask for it by name — /ai-ready (or $ai-ready in Codex) — to rule out matching problems.

The marketplace entry is pointing at a commit SHA. Plugin installs clone with git clone --branch <ref>, which only accepts a branch or tag, so a raw SHA always fails. The registry entry needs to use a release tag such as v1.3.0. See #26.

Installing directly from this repo (copilot plugin install johnpapa/ai-ready) avoids the problem entirely.

If /plugin marketplace add fails with a Git authentication error, use the full HTTPS URL:

/plugin marketplace add https://github.com/johnpapa/ai-ready

Or tell Git to prefer HTTPS for GitHub:

git config --global url."https://github.com/".insteadOf git@github.com:

Skip what you don't need

Add exclusions to your prompt and the skill will respect them:

make this repo ai-ready but skip CI and issue templates

just generate AGENTS.md and the per-tool pointer files

Report only

Just want to see where you stand? Ask for a report without generating any files:

how ai-ready is this repo?

score this repo

What to Expect

After you run the skill, you get a full AI-readiness report — analysis, proposed changes, and a projected score. Here's what it looks like for vscode-peacock:

🔗 View the interactive version — collapsible sections, responsive layout, works on mobile.

The report shows:

1. Your Repo Today — current score, what's nailed, what's missing, and why it matters
2. What I'd Like To Do — proposed files to create (nothing changes until you say so)
3. If You Accept — projected score with category breakdown
4. Ready? — offer to create a PR with all the changes

Issue/PR messages generated by this skill always include explicit attribution to the AI Ready skill. It also keeps related docs aligned to repo standards and proactively attempts conflict resolution before opening PRs (or asks when manual help is needed).

📄 Also available as terminal output — same content, rendered in the CLI.

Scoring

Your score is how many of the 11 tracked assets are nailed — no formulas, no weights. Two of the medals have one extra requirement.

Medal	Name	Count	Also required	What it means
🥉	Getting Started	1–3 of 11	—	A few of the files exist. Nothing in the repo tells an agent how it works
🥈	On Track	4–6 of 11	—	Real scaffolding is in place, but the conventions are still in people's heads
🥇	Solid	7–9 of 11	AGENTS.md nailed	The conventions are written down, in the one file every tool reads
🏆	AI-Ready	10–11 of 11	every 🤖 AI Context asset nailed	Conventions, boundaries, reviewers and procedures all live in the repo

Why the two extra requirements. A repo can reach seven nailed assets on a changelog, docs, issue templates, a PR template, CI, and two reviewer agents — with no AGENTS.md at all. That repo is well maintained. It is not AI-ready, and a plain count would hand it 🥇. So the count is a ceiling rather than a score: without AGENTS.md the repo stops at 🥈 no matter how high the count climbs. When a medal is capped, the report says which prerequisite is missing and what would lift it.

What the score doesn't tell you. It measures what's in place, not whether agents write better pull requests in your repo. Nobody has measured that, here or anywhere, and this skill won't quote you a number.

Why

Contributors (human and AI) show up to your repo and don't know the conventions. They submit PRs that miss tests, break patterns, skip docs. You leave the same review comments on every PR. AI agents make this more challenging — they generate PRs faster, but without context, those PRs create more review burden.

It's the same gap from both sides: contributors don't know what maintainers expect, and maintainers keep re-teaching it. This skill closes that gap by generating repo-level configuration that teaches everyone — human and AI — how to work in the repo correctly. It even mines your PR review comments for repeated feedback and turns them into written conventions, so the thing you keep saying in reviews is in the repo instead of in your head.

Built from Real Maintainer Experience

This skill isn't theoretical — it's shaped by John Papa's experience maintaining popular open source projects and repos at large enterprises. It prioritizes the things maintainers keep re-explaining: maintenance matrices for the files contributors forget, conventions mined from the PR feedback you're tired of repeating, and CI that catches problems before a person has to.

How It Works — GitHub-Native by Default

You shouldn't have to explain to an AI tool that you're in a GitHub repo. This skill assumes it, and leverages everything GitHub already knows about your project.

Auto-Discovery (zero user input)

The skill starts by pulling context directly from GitHub — no questions asked:

What it discovers	How	Why it matters
Repo description, topics, languages	GitHub API	Knows what your project is without reading every file
Community health score	GitHub API	GitHub already tracks which community files are missing — no need to look
Contributors	GitHub API	Team size, contribution patterns
Recent merged PRs	GitHub API	Understands what typical contributions look like
PR review comments	GitHub API	Turns your repeated review feedback into automated conventions
CI/CD workflows	GitHub Actions API	Knows your build/test pipeline
Releases	GitHub API	Understands versioning and release cadence

It then scans your local codebase for deeper details — manifest files, test configs, directory structure, existing AI configuration — and combines both into a complete picture.

PR Review Mining

The skill reads your recent PR review threads and looks for repeated feedback — the same comments you leave on every PR:

• "Please add tests for new features" → becomes a test convention rule
• "Use the X pattern instead of Y" → becomes a coding convention rule
• "Don't forget to update the changelog" → becomes a maintenance matrix entry
• "This breaks on mobile, check responsive layout" → becomes a screen size rule

These mined conventions go directly into AGENTS.md, where every tool reads them. The next AI-generated PR follows those rules automatically. You stop repeating yourself.

What Gets Generated

Every file is customized to your repo's actual language, framework, and patterns — not generic boilerplate. The skill only creates files that don't already exist.

File	What It Does
AGENTS.md	The single source of truth. Repo structure, build/test commands, architectural decisions, how to add features, coding conventions, conventions mined from your PR reviews, and the maintenance matrix. Read by Copilot, Claude Code, Codex and Cursor alike
Per-tool pointer files	.github/copilot-instructions.md, CLAUDE.md, .cursorrules — three lines each, pointing at AGENTS.md. One file holds the content; everything else points at it, so nothing drifts
.github/agents/*.agent.md	Three reviewer agents you assign to a PR — one checks the diff against what the issue actually asked for, one checks whether the tests would have failed against the old code, and one checks what a revert would not undo. Each asks a question that can come back no, works from the diff alone, and has no way to say ship it