How to Use AI Skills to Build Faster, Cleaner EDS Components

There’s never been a more exciting time to be a developer. AI is rapidly transforming the way we work, sometimes for the better and sometimes in ways that require more thought. Over the past few years, many of us have adopted AI in different ways, some cautiously and others more freely. In this post, I’ll explore a new approach to using AI in development, one that integrates it into existing code patterns, standards, and workflows through the concept of AI skills.

AI Skills

AI skills are groups of instructions designed to perform specific tasks. What sets them apart from other AI concepts is two key characteristics. First, a skill is narrowly focused on a single task. In simple terms, it represents a well-defined task that an AI can perform effectively.

Second, AI skills can be understood as workflows—a structured series of steps required to complete a task. This structure helps guide the model, resulting in more accurate, consistent, and predictable outputs.

The concept of skills was introduced by Anthropic in 2025 and has since been rapidly adopted by many popular AI models. Skills open the door to a new paradigm for AI-powered code assistants, making it easier to create specialized capabilities tailored to different domains. In the software world, this means defining skills that align with our specific needs, patterns, and standards. As a sign of this shift, Adobe has officially published skills for Edge Delivery Services development tasks.

AEM Skills

AI skills offer exciting new possibilities for developers. Although there aren’t yet any official skills for traditional AEM development, creating a custom skill is relatively straightforward, and many community-driven skills will likely appear soon. If this topic interests you, I’ll be publishing a follow-up post soon that walks through how to create AEM skills step by step. For now, let’s focus on the AI Skills currently available for EDS development.

Edge Delivery Services (EDS) Skills

One of the most significant advantages of Edge Delivery Services (EDS) is the simplicity of its stack. Using vanilla JavaScript and CSS—with no additional frameworks by default—allows AI to fit naturally into the development workflow when building blocks or individual components for EDS. However, as is well known, AI models often struggle to provide accurate responses when dealing with AEM or EDS–specific topics due to limited context.

To address this, the Adobe team has shared a curated set of AI skills, along with additional AI guidance, to enrich context and provide more efficient and reliable assistance. These skills help AI better understand EDS patterns and conventions, resulting in higher-quality output and a smoother developer experience.

You can explore more in Adobe’s guide, Developing with AI Tools.

Installing EDS Skills

AI skills adhere to the Anthropic specification, which is rapidly gaining adoption as a standard across AI providers. As a result, installing skills is simply a matter of making the Skill.md files available in a location where the AI model can discover and load them.

Traditionally, since Claude introduced the skills paradigm, skills are stored under the .claude/skills directory. Other providers have adopted similar approaches with slight variations. For example, GitHub Copilot uses .github/skills, while still maintaining interoperability by also scanning and reading from .claude/skills.

Skills can be applied at different scopes. They can live at the root of a specific project, enabling them only for that codebase, or they can be installed at the user or system level, making them available across all projects. This flexibility offers a significant advantage in terms of reuse and consistency across projects.

For this blog demo, I added the AGENTS.md file from the EDS boilerplate to my VS Code workspace. In addition, I cherry-picked several AI skills provided by Adobe for EDS development, including building-blocks, content-driven-development, testing-blocks, and docs-search. With these in place, the setup in my editor looks like this:

Please note that the Skills functionality in VS Code is still an experimental feature and, as of December 2025, must be manually enabled. It is expected to become stable by January 2026. To enable it, go to Settings in VS Code and search for “Skills.”

Comparing Custom Promo Block Creation: With vs. Without EDS Skills

Let’s see it in action. I ran a small experiment using GitHub Copilot in VS Code to create a custom “promo” block for EDS and UE, providing little context to observe how it behaved and what results it would naturally produce. I then repeated the same exercise after adding the three skills mentioned above, along with the AGENTS.md file for block development provided by Adobe, to compare the outcomes.

This is the prompt I used:

Below are the results of the experiments comparing the creation of a custom promo block with and without EDS Skills

Screenshoots

As the comparison illustrates, EDS Skills significantly improve the AI-assisted development process. Without skills, AI can generate code quickly, but it often introduces errors, misstructured files, or unintended changes that require manual fixes. With skills, the AI follows a structured workflow, detects and corrects issues iteratively, and produces a component that is more accurate and closer to being ready for testing.

Final Thoughts

The difference between using AI Skills for EDS and not using them is clear. Integrating these skills into your workflow is simple and well worth the effort, though it’s important to remember they aren’t a silver bullet—effective prompting is still essential. Equally important is understanding what each skill does. The better you know a skill, the more effectively you can craft prompts, resulting in more accurate and reliable output. Overall, AI Skills is a powerful tool that can significantly enhance and streamline development workflows.