Vks Explore

Published by vngcloud in vks-skills

No known issues1 installs

What this skill does

Use when the user wants to VIEW or CHECK the status of VKS (VNG Kubernetes Service) resources without changing anything — list clusters, inspect a cluster, see its node groups, list nodes, read events, or list available Kubernetes versions. Trigger phrases: what clusters do I have, list my clusters, show cluster X, is my cluster ready, what's the status of, why is my cluster/node group failing, show events, list node groups, how many nodes, what k8s versions are available, cluster nào đang chạy.

Add Vks Explore to your agent

Review the source and files first. When you are ready, copy the prompt instruction or use the CLI command supported by your environment.

Install with a prompt

Paste this into a compatible coding agent:

add this skill "vks-explore" from https://github.com/vngcloud/vks-skills

Install with the CLI

Run this command in a controlled environment after reviewing the repository:

npx skills add https://github.com/vngcloud/vks-skills --skill vks-explore

Skill instructions

Explore VKS (Read-Only)

Overview

Answer status questions about VKS in plain language. This skill is strictly read-only — it never creates, updates, or deletes anything, so it needs no confirmation and works even when the MCP server runs without --allow-write. Other VKS skills call into these same tools to show current state before a change.

Tools (all read-only)

Question	Tool
What clusters exist?	`cluster_list`
Details of one cluster?	`cluster_get` (cluster id)
What happened to a cluster? (debugging)	`cluster_get_events`
What node groups in a cluster?	`nodegroup_list`
Details of one node group?	`nodegroup_get`
What nodes in a node group?	`nodegroup_list_nodes`
What Kubernetes versions are available?	`cluster_versions_list`
Is auth working / which region & endpoints?	`get_access_token`

How to respond

Resolve the target. If the user names a cluster (not an id), call cluster_list and match by name. Exactly one match → use it. No match → say so and list the available clusters (offer /vks-create-cluster if they meant to create it). Multiple partial matches → list and ask. (Same procedure for node groups via nodegroup_list.)
Call the read tool(s) and translate the status. For "is it ready?" use cluster_get and report the status plainly using the status table in ../vks/references/resource-defaults.md (CREATING / ACTIVE / UPDATING / ERROR / DELETING / …); report any unlisted status verbatim.
Summarize, don't dump. Lead with the answer ("3 clusters, all ACTIVE"), then the supporting table. The tools already return markdown tables — keep them, but add a one-line takeaway.
Chain for "why is it broken?" Combine cluster_get (status) + cluster_get_events (recent events) and explain the likely cause. For a node-group problem, use nodegroup_get + nodegroup_list_nodes.

Examples

"What clusters do I have?" → cluster_list → "You have 2 clusters: prod (ACTIVE), staging (ACTIVE)." + table.
"Is my-cluster ready?" → resolve id via cluster_list → cluster_get → "Not yet — status is CREATING. I'll check again if you'd like."
"Why won't staging come up?" → cluster_get + cluster_get_events → explain the failing event.
"What k8s versions can I use?" → cluster_versions_list → highlight the recommended STABLE one.

Common mistakes

Asking for a cluster id when the user gave a name — resolve it via cluster_list.
Returning a raw table with no plain-language answer on top.
Suggesting or performing a change — that belongs to /vks-create-cluster or /vks-nodegroup. If the user wants to act, route them there.

Files included

SKILL.md