Agent Skills

The databend-cloud skill gives your agent guidance and tools for working with Databend Cloud.

This skill is for Databend Cloud only, not self-hosted Databend.

What It Can Do

• Connect to Cloud: Guide private API token setup and use the organization and warehouse you select.
• Work with SQL: Check Databend syntax, inspect databases and tables, and query data.
• Diagnose queries: Inspect execution status, results, history, and profiles to investigate slow or failed SQL.
• Manage warehouses: List, create, resize, resume, suspend, rename, or drop warehouses with the required admin role and your approval.
• Investigate incidents: Analyze alerts, OOMs, node exits, task failures, and version regressions.
• Export evidence: Export bounded logs as raw JSON Lines or readable events.

You need a Databend Cloud account and access to the target organization. Available operations depend on your account privileges and access to history, profiles, and logs.

Install with Your Agent

Copy this message to install the skill and set up a reusable API key. Create a token in the Cloud console under Account API Tokens → Access Tokens → Generate New Token.

Install the databend-cloud skill:
https://github.com/databendcloud/databend-cloud-skill

Follow SKILL.md to reuse an API key or help me save one for https://api.databend.com.
Keep the key out of chat. Stop after setup; do not connect or run SQL.

note

This is an official Databend skill, not a third-party skill.

Quick Verification

Run this command to make sure that your token, organization, and warehouse are correct. Replace the placeholders with your values.

curl --request POST 'https://api.databend.com/v1/query' \
  --header 'Content-Type: application/json' \
  --header 'X-DATABENDCLOUD-TOKEN: <YOUR_API_TOKEN>' \
  --header 'X-DATABENDCLOUD-ORG: <YOUR_ORGANIZATION_ID>' \
  --header 'X-DATABENDCLOUD-WAREHOUSE: <YOUR_WAREHOUSE>' \
  --data '{"sql":"SELECT 1"}'

Security

• Set a token IP allowlist to reject requests from other IPs. An empty allowlist means no IP restriction.
• Allow the actual outbound IP of the agent's execution environment. A remote agent can use a different outbound IP from your local computer.
• Tokens follow your account's current organization role. Use database permissions for read-only access. The skill is not a read-only sandbox.
• Disable or delete tokens in the Cloud console to revoke access. Replacing a local key file does not revoke the old token.
• View the token's last-used IP and time in the Cloud console. This is not a complete request audit log.
• Keep tokens out of chat, code, and repositories. Local credential files are plaintext; the helper sets Unix permissions to 600.
• Results and logs can enter the agent's model context. Follow your data policy.

Details: Authentication · Query API.