Agent Security Center:Getting started

Connection process

• Purchase Agent Security Center: Activate Agent Security Center with pay-as-you-go billing to obtain agent asset discovery and risk detection capabilities.

• Connect OpenClaw: Complete the service-linked role authorization and install the Security Center client, so that OpenClaw assets such as agents and model services are synchronized to Agent Security Center automatically.

• View and handle agent risks: Review the risk detection results and handle risk events such as prompt injection and data leakage.

• Install the defense plugin: Install the defense plugin for the target agent to intercept prompt injection attacks and malicious URL access in real time.

• Skills detection (optional): Upload skill files through the Skills CLI or an SDK to detect security risks and handle them.

Step 1: Purchase Agent Security Center

For more information, see Purchase and unsubscribe from Agent Security Center.

1. Access the Security Center console - Agent Security Center - Agent Overview. In the upper-left corner of the page, select the region of your assets: Chinese Mainland or Outside Chinese Mainland.

2. In the Protection Status area in the upper-right corner of the page, click Buy Now.

3. In the Enable Agent Security Center dialog box that appears, read the Security Center (Pay-as-you-go) Terms of Service and Authorization for Security Center Service-linked Role instructions, and select the checkbox to confirm your agreement.

4. Click Activate Now. After activation, the indicator in the protection status area changes from Basic to Pay-as-you-go.

Step 2: Connect OpenClaw

OpenClaw requires no separate connection information. After you complete the service-linked role authorization and install the Security Center client on the server, assets are synchronized automatically. For more information, see Connecting to Agent Security Center.

1. Role authorization (first time):

   1. Access the Security Center console - Agent Security Center - Agent Overview. In the upper-left corner of the page, select the region of your assets: Chinese Mainland or Outside Chinese Mainland.

   2. In the Connected Platforms section, click OpenClaw.

   3. In the OpenClaw connection panel, click Authorize.

      Note

      The system automatically creates the service-linked role AliyunServiceRoleForSasAI for Agent Security Center and configures the AliyunServiceRolePolicyForSasAI permission policy, which allows Agent Security Center to access resources in other cloud services.

2. Install the client: On the server where OpenClaw is deployed, install the Security Center client. Install the client.

3. After installation and authorization, the system automatically synchronizes OpenClaw assets.

   Important

   Agent Security Center supports real-time protection for OpenClaw. During the public preview, free defense is available for a limited time. Agent risk detection and protection.

Step 3: View and handle agent risks

After the connection is complete, the system automatically runs one risk detection per day on the connected OpenClaw assets. The detection is a read-only analysis and does not affect running workloads. The time it takes depends on the complexity of your assets. Detection results reflect the risk status at the time of detection, so we recommend that you run a new detection after you change an agent configuration.For more information, see Detect and Remediate Risks.

1. Access the Security Center console - Agent Security Center - Agent Risks. In the upper-left corner of the page, select the region of your assets: Chinese Mainland or Outside Chinese Mainland.

2. On the Risk Events page, in the Actions column for the target event, click Handle.

3. On the risk event details page, review the details. After assessing the risk, choose one of the following handling methods:

   • Manual handling: Follow the Recommended Solution or Suggestions on the details page to fix the issue in the Agent application's code or configuration.

   • Ignore: If the current risk is acceptable, click Ignore. The risk event will no longer appear in the pending list. If the same risk is detected in the next scan, a new risk event is generated.

   • Rescan: To verify the result of manual handling or re-detect the current AI asset, click Rescan.

     Important

     • This applies only to risks originating from Agent Security. Detection takes time; you can view progress in the console.

     • A rescan runs a full detection on the AI asset associated with the current risk event, verifying whether the current risk is resolved and checking for additional risks.

Step 4: Install the defense plugin

After you install the Agent Security Center defense plugin, the plugin monitors agent runtime in real time and automatically detects and intercepts security threats such as prompt injection attacks and malicious URL access. OpenClaw supports both automatic installation and manual installation:

• Automatic installation: applicable only to OpenClaw assets that have been detected by Agent Security Center.

• Manual installation: applicable when automatic installation fails, or when the OpenClaw asset has not been automatically detected by Agent Security Center. After installation is complete, related data is automatically synchronized to Agent Security Center.

Step 5: Skills detection (optional)

Skill files can contain injected malicious scripts, risky configurations, or malicious prompts. Skills detection focuses on the supply chain security of agent skills and covers risks such as dangerous commands, data exfiltration, prompt injection, privilege escalation, insecure configurations, and sensitive information leaks, so that you can intercept problematic files before a skill is deployed. The main process is as follows:

1. Choose an upload method:

   • Skills CLI: Use the alibabacloud-aisc-skill-inspection skill provided by the Agent Skills portal to start a detection through the Skills CLI. No integration code is required, which suits a quick validation of the detection capability.

   • Multi-language SDK: Integrate a multi-language SDK such as Java, Python, or Go, and call the operation that creates a skill detection task to upload skill files. Integration code is required.

2. Configure Alibaba Cloud access credentials: The Skills CLI obtains credentials from the Alibaba Cloud default credential chain. Write the AccessKey pair or role information to environment variables or the ~/.aliyun/config.json configuration file.

3. After you start the detection, complete the following operations on the SkillsCheck tab of the Agent Risks page:

   • View the results: view the number of detected files, the number of unhandled files, and the risk severity distribution.

   • Handle risks: click Details in the Actions column of the target file, fix the issue manually based on the remediation suggestions, or ignore risks that are acceptable.

   • Add trusted files to the allowlist: files confirmed as trusted by a security assessment can be excluded from detection through SHA256 allowlist rules.

For SDK installation and invocation examples, API parameters, CI/CD integration methods, and the risk severity criteria, see Agent Risk - Skills Detection.

Billing

• Defense plugin: Starting August 25, 2026, the defense plugin that Agent Security Center provides for OpenClaw and Hermes enters official commercial billing. The fee consists of a defense seat fee and an excess Credits fee. For the billing start time, prices, and the free trial extension rules for users who already installed the plugin, see Agent Security Center Defense Plugin Billing Announcement (Seat Fee/Credits).

• Skills detection (optional): Skills detection is billed based on the number of detected skill files and requires an activated paid instance. Each detected skill file counts as one detection. If a single API call contains multiple files, detections are accumulated by the total number of files. Fees are incurred only when a detection task completes successfully. The price is $1.42 per detection.

Next steps

• Connecting to Agent Security Center: Bring AI assets from platforms such as Model Studio, Platform for AI (PAI), Dify, VolcEngine Agent Kit, and Agent Run under unified management.

• Detect and Remediate Risks: Learn about all detection items, recommendations for choosing a detection method, and the available risk handling methods.

• Install the defense plugin: Install the defense plugin for the Hermes platform, or disable or uninstall an installed plugin.

• Agent Risk - Skills Detection: Automatically scan skill files for supply chain risks in a CI/CD pipeline.

• Agent Security Center Billing (Skills Detection/AI Red Teaming/Audit Logs): Learn about the metering logic of each billing item, overdue payment handling, and data deletion rules.