/byted-redis

name: byted-redis
description: Use this skill only for Volcengine Redis cloud resource operations, including querying, inspecting, creating, or managing Volcengine Redis instances and related cloud resources. Trigger it when the user mentions Volcengine Redis, Volcengine Redis instances, Volcengine Redis inventory/detail lookup, Volcengine Redis operational metadata, or Volcengine Redis MCP/OpenAPI access.
version: 1.1.1
license: Apache-2.0

Volcengine Redis MCP Server Skill

Purpose

Use this skill as the agent-facing workflow for Volcengine Redis cloud resource operations. The skill bundles MCP client scripts, startup scripts, credential guidance, and tool-selection instructions so agents can perform Volcengine Redis resource queries and management tasks.

This skill is scoped to Volcengine Redis only. Redis-related requests outside Volcengine, such as OSS Redis command execution, self-hosted Redis troubleshooting, application cache logic, Redis running in Kubernetes, or non-Volcengine cloud Redis products, should use other tools or skills.

Triggering Rules

Use this skill when the user asks for Volcengine Redis tasks such as:

• Redis instance inventory, detail lookup, status lookup, filtering by region, instance ID, name, tag, VPC, or subnet.
• Region, zone, VPC, subnet, EIP, spec, shard, node ID, endpoint, bandwidth, or topology information.
• Slow log, hot key, big key, key scan job, planned event, or operational metadata queries.
• Backup, PITR, cross-region backup, backup plan, or backup download URL queries.
• Parameter group, instance parameter, ACL command/category, account, or allowlist operations.
• Creating or modifying Volcengine Redis resources, including instances, accounts, allowlists, public endpoints, names, and parameters.

Do not use this skill for generic Redis command execution against a data endpoint, application-side cache debugging, Kubernetes pod metadata lookup, self-hosted Redis, other cloud providers' Redis products, or any non-Volcengine Redis resource unless the user explicitly says the target is Volcengine Redis.

Execution Workflow

1. Classify the request as read-only or mutating. 2. Check the required identifiers. Prefer explicit `region_id` and `instance_id` when available. If the user gives only a name or vague keyword, first call `describe_db_instances` with region and filter arguments, then use the returned instance ID for detail tools. 3. Verify credentials are available before starting MCP:

• `AUTHORIZATION` or `authorization`, or
• `VOLCENGINE_ACCESS_KEY` and `VOLCENGINE_SECRET_KEY`, with optional `VOLCENGINE_SESSION_TOKEN`.

4. If the region is not supplied, use `VOLCENGINE_REGION` when set. If neither the user nor the environment gives a region, ask for the region before making region-scoped calls. Prefer region-scoped discovery tools such as `describe_zones` or `describe_db_instances` once the region is known. 5. Use the bundled client in `scripts/mcp_client.py` for custom calls. For a smoke test, run `uv run scripts/call_redis_mcp_example.py` from this skill directory. 6. For mutating operations, summarize the target resource and intended change, then ask for user confirmation before calling the tool unless the user has already clearly authorized that exact mutation. 7. Return concise results with the resource identifiers, region, status, and the fields relevant to the user's question. Mention when output was filtered or truncated.

Credentials

The bundled stdio client supports two credential modes.

Environment AK/SK

Use these variables:

• `VOLCENGINE_ACCESS_KEY`
• `VOLCENGINE_SECRET_KEY`
• `VOLCENGINE_REGION`, for example `cn-beijing`

For temporary credentials, also use:

• `VOLCENGINE_SESSION_TOKEN`

Authorization Bearer Token

Use `AUTHORIZATION` or `authorization` with this format:

Authorization: Bearer BASE64_JSON_PAYLOAD

The decoded JSON payload should contain:

• `AccessKeyId`
• `SecretAccessKey`
• `SessionToken`
• `CurrentTime`
• `ExpiredTime`
• `Region`

`SessionToken` is required for STS credentials. Authorization credentials take precedence over environment AK/SK when both are present.

Never print full secrets, AK/SK values, session tokens, or authorization payloads. If a credential check is needed, report only whether each variable is present.

Tool Selection

Use read-only tools by default. The table below is a routing guide, not a closed capability list. The upstream MCP server is the source of truth for the currently available tools. If the server exposes new tools later, agents can call them through the same `RedisMCPClient.call_tool(name, arguments)` flow after checking the tool name, schema, and safety level.

| User intent | Preferred tool | Key arguments | | --- | --- | --- | | List regions | `describe_regions` | only when supported by the live server schema; otherwise provide `region_id` or use another region-scoped discovery tool | | List zones | `describe_zones` | `region_id` | | List VPCs or subnets | `describe_vpcs`, `describe_subnets` | `region_id`, optional VPC/subnet filters | | Find Redis instances | `describe_db_instances` | `region_id`, optional `instance_id`, `instance_name`, paging/filter fields | | Inspect one instance | `describe_db_instance_detail` | `region_id`, `instance_id` | | Query specs | `describe_db_instance_specs` | `region_id`, optional zone/spec filters | | Query shards or nodes | `describe_db_instance_shards`, `describe_node_ids` | `region_id`, `instance_id` | | Slow logs, hot keys, big keys | `describe_slow_logs`, `describe_hot_keys`, `describe_big_keys` | `region_id`, `instance_id`, time range when supported | | Backups and PITR | `describe_backups`, `describe_backup_plan`, `describe_pitr_time_window` | `region_id`, `instance_id` | | Parameters | `describe_db_instance_params`, `describe_parameter_groups`, `describe_parameter_group_detail` | `region_id`, `instance_id` or parameter group ID | | Accounts | `list_db_account` | `region_id