We taught an AI how to build on Magnolia

A powerful platform — and a faster way to build on it

Magnolia DXP gives you a lot to work with: light modules, content apps, decorations, multisite, headless rendering, personalization, publishing workflows, and Magnolia 6.4 comes with the AI Accelerator. That breadth is exactly why teams choose it — there's the right tool for almost any digital experience problem.

Breadth comes with depth, and depth means there's craft to learn. Experienced Magnolia DXP developers internalize the conventions — the YAML structure, the version-specific syntax, the patterns that make a clean, maintainable build. Newcomers work their way up that curve. That's normal for a serious enterprise DXP, and it's a curve worth climbing.

Here's what's new: you no longer climb it alone. Today, most of us build with an AI coding assistant at our side. The opportunity is enormous — faster scaffolding, less boilerplate, more time on the interesting problems. The catch is that a general-purpose assistant knows a little about everything and the specifics of nothing. To make it genuinely useful on Magnolia DXP, it needs Magnolia DXP expertise — the same expertise a senior developer carries.

That's what we set out to give it. And now we're giving it to you.

This release didn't start as a product. It started with a simple goal: we wanted our AI assistant to be a real Magnolia DXP expert, not an enthusiastic generalist.

So every time the assistant was about to write a particular kind of Magnolia DXP configuration, we had it consult correct, version-specific, battle-tested guidance *first* — the accumulated know-how of building on the platform, written down in a form the assistant could load exactly when it was relevant.

The result is a set of skills: focused, structured reference documents that an AI assistant invokes by name before it touches a particular kind of Magnolia DXP config. Building dialog YAML? It reads the dialog-fields skill first. Writing a REST endpoint? It reads the rest-services skill first. Setting up multisite, imaging, personalization, publishing? There's a skill for each, and the assistant automatically reaches for the right one.

The effect is immediate: the assistant stops guessing and starts building like someone who already knows Magnolia DXP — correct syntax, the right version's conventions, clean patterns, the first time.

What began as our own way of working turned out to be broadly useful. The same skills that made our assistant a strong Magnolia DXP developer work for anyone building on Magnolia DXP. So we're opening them up.

There's a happy truth here worth naming: good documentation doesn't only help people — it also helps AI. Magnolia DXP's documentation is public, so an assistant can read it directly while it works. And because it's been public for years, a lot of that knowledge is already part of what today's large language models have learned during training. Openness pays compounding dividends — every well-written public page makes every AI assistant a little better at Magnolia DXP.

Skills are built on that foundation rather than replacing it. Where documentation is broad and written for a human reader, a skill is focused, version-specific, and surfaced at exactly the right moment — the precise move the assistant needs for the precise task in front of it. Think of it as the difference between a reference library and an expert colleague leaning over your shoulder at just the right time.

The `magnolia-development-6` skill set is 39 skills spanning the platform end-to-end. Each one captures the proven way to do a category of work in Magnolia 6.4 — correct syntax, current conventions, and the patterns experienced teams rely on. Grouped by theme:

It adds up to tens of thousands of words of dense, example-driven guidance — hundreds of working YAML, FreeMarker, and Java snippets, not pseudocode.

The clusters show the coverage. Here's the payoff — what an assistant equipped with these skills actually does for you, on both sides of the job.

Ask for a dialog, a content app, a REST endpoint, a multisite setup, or an image variation, and the assistant writes it the way an experienced Magnolia DXP developer would — current 6.4 syntax, the right conventions, clean and maintainable. The combo box renders its options. The link field points to the right workspace. The page template carries everything the Visual Editor expects. You spend your review time on product decisions, not on correcting syntax.

This is where skills really shine. Because the assistant knows the correct shape of a working Magnolia DXP configuration, it can recognize when something's off and explain why. "The public site isn't showing my change" becomes a guided answer about how the author/public model and publishing actually work. A configuration that isn't taking effect becomes a quick diff between what you wrote and the pattern the skill describes. The assistant troubleshoots with you, grounded in how Magnolia DXP really behaves — not in a plausible guess.

The two together change the rhythm of a project: faster to build, and faster to get unstuck.

If you're new to Magnolia DXP, the skills are the fast, friendly start you'd want. Instead of piecing together tutorials of uncertain vintage, your assistant gives you a clear mental model of how the platform fits together, the exact working syntax for whatever you're building, and gentle guidance around the conventions that take newcomers time to learn. You climb the learning curve faster — and you learn the *right* patterns from day one.

If you're an experienced Magnolia DXP developer, the value is just as real. Your assistant produces config you don't have to rewrite — parameterized templates done well, decorations with correct merge behavior, version handlers that are safe to re-run, multisite URL resolution that respects domain mapping. The hard-won knowledge in your head now sits alongside you in the tool, working at machine speed, on every task in parallel.

Either way, the principle is the same: expertise that used to live in one person's head now travels with the work — and scales to the whole team.

The skills are designed for AI coding assistants that support the open skills format — Claude Code, among them. You point your assistant to the public repository, and from then on, it automatically reaches for the right skill whenever it works on the matching part of Magnolia DXP. No new platform, no lock-in — just a noticeably stronger Magnolia DXP developer inside the AI you already use.