Skip to main content
RisingWave provides free skills for general development and CLI guidance, plus a paid Cloud package for cluster diagnostics. Choose the skills that match your workflow; the paid package is not required to use the free skills.

Choose your skills

Installing any skill does not grant access to Cloud resources. Cloud operations still require valid credentials and the appropriate permissions.

Free public skills

The public RisingWave agent skills repository includes:
  • risingwave: guidance for connections, sources, materialized views, sinks, and streaming SQL.
  • risingwave-best-practices: guidance for schema design, streaming queries, CDC, and performance.
With Node.js and npm installed, run the following from your project directory and select your assistant when prompted:
To select just one skill:
These skills do not install the paid Cloud diagnostic package.

Free CLI skills

The RisingWave Cloud CLI bundles command-reference skills with the binary. Installing them does not contact the paid package service or require CLI authentication. Choose a supported target and run its command from your project directory:
These embedded skills describe CLI commands; they are different from the paid diagnostic package installed with --pro. Authenticate the CLI separately before using it to operate on Cloud resources. The paid package provides instructions and tools for discovering clusters and investigating metrics, logs, and SQL performance through Cloud APIs.
The paid package requires an organization with access to paid agent skills. Check AI Agent → Agent Skills in the Cloud console for availability. Installing skills does not grant additional permissions: API requests use the permissions of the user or service account configured in your CLI.

What’s included

Step 1: Prepare your CLI

Install the latest RisingWave Cloud CLI and check its version:
The diagnostic helper also requires Node.js 20 or later. Use a CLI release that supports --pro and includes the paid package verification key. If installation reports paid skill release trust root is not configured, update the CLI; do not disable signature verification.

Step 2: Authenticate

Browser sign-in does not authenticate the CLI. Choose either method below.
Service accounts work with CLI, agent, and automation workflows, regardless of how you sign in to the console.
  1. Open AI Agent → Credentials in the Cloud console.
  2. Create a service account, or select one you have permission to use. Creating service accounts and keys requires the appropriate organization permissions.
  3. Create an API key and save its secret, which is shown only once.
  4. Configure the CLI with that key:
Use a service account in the organization that has access to the package. Access to cluster metrics, logs, and SQL also depends on that service account’s permissions. See Role-based access control.
Keep API secrets and CLI configuration private. Do not paste them into assistant conversations, logs, or source control. Your shell may save commands containing secrets in its history; use your organization’s approved credential-handling practices.

Step 3: Install for your assistant

In AI Agent → Agent Skills, the Paid Agent Skills panel shows the available package version and installation commands. For a project-local installation, run the command from the project directory where you use your assistant. Choose your target:
Installs under .agents/skills/ in the current project.
The installer checks your package access, downloads the package, verifies its signature and file hashes, and writes the installed files and a lockfile. Keep the lockfile with the installation.

Step 4: Verify and use the skills

Verify the installed files with the same target and directory used for installation. For example, for Codex:
verify checks installed files against the lockfile. doctor checks the installed package and authenticated access to the package service; it is not a cluster health check. Open the project in your assistant and follow that client’s instructions for loading skills. Start with a bounded, read-only request, for example:
Use the RisingWave Cloud skills to investigate streaming latency for my cluster over the last 30 minutes. Start by discovering the cluster and region, and report your findings without changing resources.
The discovery skill resolves the cluster’s region; you do not need to select a region just to install the package. For manual CLI commands such as rwc cluster list, first select a region.

Update the paid package

Run the installation command again to fetch the current package, using the same target and project directory:
Keep your own notes or custom skills separate from the managed package files. If verification fails, do not edit the lockfile to suppress the mismatch; review local modifications before reinstalling.