Microservices Engine:AI Registry Skill Quick Start

Create a Skill

1. In the left-side navigation pane, choose AI Resource Management > Skill. In the workspace drop-down list at the top of the page, select the target workspace.

2. In the upper-right corner, click + New Skill.

3. Set the basic information:

   Parameter	Required	Description
   Skill Name	Yes	Up to 64 characters. Only letters, digits, underscores (_), and hyphens (-) are supported.
   Description	Yes	Up to 1,000 characters. Describes the functionality and purpose of the Skill.
   Tag	No	Press Enter to confirm each tag. Tags are used for business categorization and search.

4. Expand Advanced Options to set the visibility: Private (default, visible only to you) or Public (visible to everyone).

5. In the editor, write the SKILL.md file (required). The SKILL.md file uses Markdown format, with editing on the left and a real-time preview on the right. The basic structure is as follows:

6. name (required): the Skill name, used for identification.

7. description (required): a purpose description that explains the functionality and applicable scenarios of the Skill.

8. instructions (required): the core instructions that define how an agent uses the Skill.

9. Optional fields: tools (tool list), dependencies (dependencies), and more.

10. After you finish editing, click Create Now in the upper-right corner. The system generates a draft version of the Skill in the Draft state, and the Skill appears on the Skill list page.

Import an existing Skill

In the left-side navigation pane, choose AI Resource Management > Skill. On the Skill list page, click Import Skill. Two import methods are supported:

• Import Preset Skill: The system provides built-in official preset Skill templates. Select the Skills that you need and import them with one click.

• Upload File: Package your local Skill directory into a ZIP file and upload it.

After the import, the system automatically generates a draft version.

Choose an integration method

Configure VPC private network access

If you select public network access, skip this section and use the endpoint domain name from the public network access configuration guide as the host/serverAddr value.

If you select VPC private network access, associate a VPC with one click in Provisioning in the AI Registry console. The system automatically completes the PrivateLink endpoint configuration. You do not need to create an endpoint manually in the VPC console. For more configuration details, see the dedicated VPC private network access topic in See also.

The following prerequisites apply. If any resource is missing, create it first:

• A VPC is created in the target region. For more information, see Create or delete VPCs and vSwitches.

• A VPC and a vSwitch for deploying your application are created. If they are not created, see Create VPC and vSwitch.

• A security group is created. If it is not created, see Security groups.

Procedure:

1. Log on to the MSE console. In the left-side navigation pane, choose AI Registry > Provisioning.

2. In the VPC Access section, click + Associate VPC in the upper-right corner.

3. In the dialog box, set the following parameters:

   Parameter	Description
   Select VPC	Select the target VPC that you created.
   Zone and vSwitch	Select a zone in the VPC and the corresponding vSwitch. Click + Add vSwitch to configure multiple vSwitches. Select at least two zones for high availability.
   Security Group	Select a security group that allows access from within the VPC.
   Description	Optional. Helps you identify this access record.

4. Click Confirm Association. The system automatically creates the VPC endpoint, which typically takes effect within 1 to 2 minutes. When the status changes to Available, the access configuration is complete.

Verify connectivity:

After the access configuration is complete, the private access domain name of the current region appears at the top of the VPC Access section. If the VPC already contains an ECS instance, run the following command on the ECS instance to verify connectivity:

# Test network connectivity (Hangzhou region as an example)
ping -c 4 airegistry.cn-hangzhou.mse.aliyuncs.com

Method A: Nacos CLI

Configure credentials: On the device that runs the client, run the following command to configure the logon credentials:

npx @nacos-group/cli profile edit

Provide the following information as prompted:

Verify the connection:

npx @nacos-group/cli skill-list

The configuration is successful if the command returns a Skill list. The list can be empty.

Download a Skill: Multiple AI coding tools are supported, including QwenPaw, Qoder, Claude, Codex, and Cursor. On the Skill details page in the console, select the corresponding tool. The page automatically generates the command. Copy the command and run it:

# Download the latest version
npx @nacos-group/cli skill-get <skill-name>

# Download a specific version
npx @nacos-group/cli skill-get <skill-name> --version 1.0.0 -o ~/my-skills

# Download by label
npx @nacos-group/cli skill-get <skill-name> --label latest

Method B: Java SDK

Add the dependency: For the nacos-client version that supports Skills, see Access AI Registry with the Nacos Java SDK.

<dependency>
    <groupId>com.alibaba.nacos</groupId>
    <artifactId>nacos-client</artifactId>
    <version>${nacos.version}</version>
</dependency>

Initialize AiService:

Replace SERVER_ADDR with the public endpoint domain name or the private domain name.

Properties properties = new Properties();
properties.setProperty(PropertyKeyConst.SERVER_ADDR, "airegistry.cn-hangzhou.mse.aliyuncs.com:80");
properties.setProperty(PropertyKeyConst.NAMESPACE, "xxxxxx-xxxx-xxxx-xxxx-xxxxxx");
properties.setProperty(PropertyKeyConst.ACCESS_KEY, "your-access-key-id");
properties.setProperty(PropertyKeyConst.SECRET_KEY, "XXXXXXXXXXXXXXXXXXXXxx");

AiService aiService = AiFactory.createAiService(properties);

Download a Skill ZIP file:

The SDK provides three download methods: by name, by version, and by label. Each method returns a ZIP byte array (byte[]) that contains SKILL.md and all resource files. Binary resources are automatically decoded from Base64.

try {
    // Download the latest version
    byte[] skillZip = aiService.downloadSkillZip("my-first-skill");

    // Download a specific version
    byte[] skillZipByVer = aiService.downloadSkillZipByVersion("my-first-skill", "1.0.0");

    // Download by label
    byte[] skillZipByLabel = aiService.downloadSkillZipByLabel("my-first-skill", "latest");
} catch (NacosException e) {
    e.printStackTrace();
}

Subscribe to Skill changes:

Use subscribeSkill to subscribe to Skill updates. When the ZIP content changes, the listener callback is triggered. version and label are optional and narrow the subscription scope:

try {
    aiService.subscribeSkill("my-first-skill", null, null, new AbstractNacosSkillListener() {
        @Override
        public void onEvent(SkillEvent event) {
            byte[] newZip = event.getSkillZip();
            System.out.println("skill updated, size: " + newZip.length);
        }
    });
} catch (NacosException e) {
    e.printStackTrace();
}

Method C: AgentScope

Use the agentscope-extensions-nacos-skill module to use AI Registry Skills in the AgentScope framework.

Add the dependency: For the dependency version, see Integrate AgentScope with Nacos AI Registry to Use Skills.

<dependency>
    <groupId>io.agentscope</groupId>
    <artifactId>agentscope-extensions-nacos-skill</artifactId>
</dependency>

Complete example (copy and run):

Before you run the example, set the DASHSCOPE_API_KEY environment variable to the API key of your model service. Replace SERVER_ADDR with the public endpoint domain name or the private domain name.

import com.alibaba.nacos.api.PropertyKeyConst;
import com.alibaba.nacos.api.ai.AiFactory;
import com.alibaba.nacos.api.ai.AiService;
import io.agentscope.core.ReActAgent;
import io.agentscope.core.model.DashScopeChatModel;
import io.agentscope.core.message.Msg;
import io.agentscope.core.nacos.skill.NacosSkillRepository;
import io.agentscope.core.skill.AgentSkill;
import io.agentscope.core.skill.SkillBox;
import java.util.Properties;

public class QuickStart {
    public static void main(String[] args) throws Exception {
        // 1. Connect to AI Registry
        Properties props = new Properties();
        props.put(PropertyKeyConst.SERVER_ADDR, "airegistry.cn-hangzhou.mse.aliyuncs.com:80");
        props.put(PropertyKeyConst.NAMESPACE, "xxxxxx-xxxx-xxxx-xxxx-xxxxxx");
        props.put(PropertyKeyConst.ACCESS_KEY, "your-access-key");
        props.put(PropertyKeyConst.SECRET_KEY, "your-secret-key");
        AiService aiService = AiFactory.createAiService(props);

        // 2. Load Skill
        NacosSkillRepository repo = new NacosSkillRepository(aiService, "public");
        AgentSkill skill = repo.getSkill("my-first-skill");

        // 3. Bind to Agent
        SkillBox skillBox = new SkillBox(repo);
        ReActAgent agent = ReActAgent.builder()
                .name("MyAgent")
                .model(DashScopeChatModel.builder()
                        .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                        .modelName("qwen-max")
                        .build())
                .skillBox(skillBox)
                .build();

        // 4. Chat
        Msg response = agent.call(Msg.of("Hello"));
        System.out.println(response.getText());

        aiService.shutdown();
    }
}

You can filter a specific version by version (AGENTSCOPE_NACOS_SKILL_VERSION) or by label (AGENTSCOPE_NACOS_SKILL_LABEL). The repository is read-only.

For more configuration options and advanced usage, see Integrate AgentScope with Nacos AI Registry to Use Skills.