We’re seeing a fundamental shift in how digital projects are delivered. AI isn’t just a “helper” anymore; it’s becoming the primary starting point for new projects and where developers explore. We want to ensure Umbraco is an Agent-Ready CMS, meaning that when your team reaches for an AI assistant, the right Umbraco knowledge is already there, available in the tools they use the most.
Imagine being able to use the Backoffice Extension Agent Skill to:
Turn a client’s idea into a working wireframe in hours, not days, to validate concepts early.
Upgrade older extensions to the modern backoffice using the latest patterns instead of guessing.
Use it as a digital "senior developer" to help your team master new patterns while they actually build.
Here's how it works and why it matters for Umbraco developers and your business.
What Are Agent Skills?
Agent Skills act as a helping hand for AI coding tools. These specific skills are designed to provide AI with structured, up-to-date knowledge and best practice on how to create extensions for the Umbraco backoffice.
Think of them as expert guides that sit alongside your AI assistant. When you ask your AI to build a dashboard, a custom section, or a property editor, the relevant skill is loaded into context and gives the AI everything it needs: the concepts, the patterns, working code examples, and links to official documentation.
Learn more about what Agent Skills are in our previous blog post, The Agent-Ready CMS: What are Agent Skills?
The Problem Agent Skills Solve
AI is probabilistic. It learns from everything it's seen, and there's a lot of content out there about Umbraco; articles, tutorials, and documentation spanning versions 8-13, 14, 15, 16, and 17.
When you ask an Agent to build a backoffice extension without guidance, it might have a good stab at it. But it will often:
Mix patterns from different Umbraco versions
Use deprecated APIs or outdated component names
Maybe even try to write AngularJS!?!!
These skills help to reduce that problem by giving the AI the right information at the right time.
What We're Aiming For with CMS Backoffice Skills
This project has several goals:
Help people learn how the backoffice works. Each skill contains details, documentation, and working examples that demonstrate real patterns. You can ask the AI to explain how the code works, walk you through the patterns, and clarify why things are done a certain way. It's documentation that actually runs with an expert on hand to explain it.
Pair programming with your AI assistant. These skills aren't only about generating code for you; they're about creating extensions together. Load the umbraco-backoffice skill, describe what you want to build, and the Agent will guide you through the process step by step. You stay in control while learning as you go. It's pair programming with a partner who knows the Umbraco backoffice inside out.
Enable rapid prototyping. Got an idea for a custom admin area? A data management tool? A hierarchical content browser? These skills help you go from concept to working prototype quickly.
Assist with upgrades. Whilst there are no skills specifically for the AngularJS backoffice, the detailed descriptions of what each extension type does and how it should behave can help when upgrading. Describe your existing extension, provide specs or even screenshots, and let the AI help you rebuild it with the latest patterns.
How The Umbraco Backoffice Skills Work
The skills are organised as a collection, with over 66 individual skills covering many different aspects of backoffice development. This grouping matches the documentation for extension types in the Umbraco backoffice.
Each skill provides:
A brief description of what the extension type does and when you'd use it
The fundamentals of how it works within the backoffice architecture
Code examples showing minimal implementations
Links to official Umbraco documentation for deeper detail
The key design principle is progressive discovery. Only the skills needed for your current task are loaded into the AI's context. When a skill is loaded, it contains links to related skills and other documents that provide more information as and when the AI needs it.
This matters because AI models have limited context windows. Loading everything at once would overwhelm the context and degrade quality.
The key is: Only what's needed, when it's needed.
By loading information progressively, the AI can maintain focus and produce better results.
The Umbraco Backoffice Routing Skill
At the heart of the collection is a routing skill called umbraco-backoffice. This is the "big picture" skill that explains how the backoffice works and how all the pieces fit together. It also routes down to relevant sub-skills for specific implementation.
It relates to a crucial concept: backoffice customisations are combinations of extension types. For example:
A custom admin area = Section + Menu + Dashboard
A data management tool = Section + Menu + Workspace
A hierarchical browser = Section + Menu + Tree + Workspace
The skill includes an extension map showing where all 40+ extension types appear in the backoffice UI. This helps it to understand what it's building and where everything will appear.
Most importantly, this skill contains fully working, runnable, and tested examples. These aren't snippets - they're complete extensions you can build and run. The AI can reference these examples to validate its output against known-good implementations.