AI coding agents are quickly becoming part of the everyday developer workflow. Whether you're using GitHub Copilot, Gemini, Cursor, or other agent-driven tools, the quality of the results increasingly depends on one thing:

Good, project-specific instructions.

With the latest release of Blade, Liferay now makes this dramatically easier by introducing markdown-based rule files designed specifically for major AI coding agents, right out of the box.

That said, adding AI agent support to the workspace does not mean you can remove developers from the project and expect an agent to generate production-ready artifacts for you. AI agents can help bootstrap a new module, client extension, or configuration (often taking you further than the standard Blade templates ever could) but they are not a substitute for experienced developers. No current agent is capable of consistently delivering production-quality, supportable, and secure Liferay solutions on its own.

It's best to think of AI agents as a force multiplier for your development team. At their best, they accelerate experienced developers and reduce repetitive work. At their worst, they can create technical debt faster than you can review it. Used properly, they help your team move faster. Used blindly, they can slow you down.

Let's take a look at what's new, how it works, and how you can adopt it in both new and existing Liferay workspaces.

What's New in Blade

The latest Blade release introduces AI agent rule files generated automatically as part of a Liferay workspace.

After updating Blade, every new workspace you create will include:

• Agent-specific folders
• Base (shared) rules in the .workspace-rules folder.
• Markdown rule files tailored for popular AI coding agents

These rules describe:

• Workspace structure
• Liferay conventions
• Build and deployment expectations
• Common do's and don'ts for generating Liferay-compatible code

Because the rules are written in plain markdown, they're:

• Easy to read
• Easy to customize
• Easy for AI agents to consume and follow

This is a big step toward making AI-generated code actually usable in real Liferay projects.

Updating Blade

Before you can take advantage of this feature, make sure Blade itself is up to date. Use the command blade update, it's that simple.

Once updated, any newly created workspace will automatically include the AI agent rule files, no extra configuration required.

If you haven't updated Blade in a while, this alone is a good excuse to do it.

New Workspaces: Zero Effort

For new Liferay workspaces, adoption is effortless:

1. Update Blade
2. Create a new workspace
3. Start using your AI coding agent of choice

The agent rule files are already there, and your AI assistant immediately understands:

• What a Liferay workspace looks like
• How modules and client extensions are structured
• Which tools and conventions to follow

This significantly reduces hallucinated APIs, incorrect build setups, and "almost right" code.

Existing Workspaces: Two Practical Options

If you already have an existing workspace, you have a couple of good paths forward.

Option 1: Copy the Agent Rules into Your Existing Workspace

The fastest approach:

1. Create a new workspace using the updated Blade
2. Copy the AI agent folders and markdown files
3. Paste them into your existing workspace

This gives your current project immediate AI support with minimal disruption.

Option 2: Use This as a Toolchain Refresh Opportunity

If your workspace is older (or you've been meaning to modernize anyway) this is a great moment to do some cleanup.

1. Create a new workspace with the latest Blade
2. Copy over Existing modules. Client extensions, and gradle.properties
3. Validate builds and dependencies
4. Keep the new AI agent rule files intact

You end up with:

• A refreshed workspace
• Updated tooling
• First-class AI agent support

For long-lived projects, this is often the better long-term move.

Bonus for Gemini CLI Users: Shirley

If you're using Gemini CLI, there's some extra good news.

Our own Kris Patefield has just released a repository of additional Gemini rule files, affectionately nicknamed "Shirley." The rules should also work for other generative AI models and tools, such as Claude Sonnet and OpenAI tools, though you will need to test and modify as required.

These rules go beyond basic guidance and can generate:

• Enhanced Liferay assets
• More complete demo setups
• Solution-oriented starting points

While originally intended to speed up demo creation, they're just as useful for:

• Solution developers
• Proof-of-concept work
• Rapid prototyping

The repo is being updated a couple of times a week. So make sure you do a pull regularly to ensure you get all the latest updates. For DXP customers with a valid support license, you can access Kris for further training sessions on how to use the rules, through your Account Manager or your TAM. Partners can also access the service though their Partner Account Managers.

Why This Matters

AI agents are only as good as the context you give them.

By baking AI-specific rules directly into the Liferay workspace:

• Developers get better results immediately
• Teams get more consistent output
• Liferay-specific knowledge becomes part of the workflow

This is a small feature with a huge impact, and it's exactly the kind of practical innovation that makes day-to-day development smoother.

It's also important to remember what these rule files are and what they are not.

They are designed to help developers get started faster. They can scaffold, suggest structure, and generate useful starting points.

But they cannot replace developer judgment.

Everything an AI agent produces must be reviewed carefully and often refined, whether to prevent accidental data exposure, avoid performance pitfalls, ensure proper security practices, or simply make the solution production-ready.

These tools can accelerate your development process, but they do not eliminate the need for experienced developers to do the real engineering work.