From Intility Developer Platform
Entry point for getting an app running on the Intility Developer Platform. Use when the user says things like "help me deploy", "get me started with Intility", "I want to run my app on the platform", "I'm new to Kubernetes and need to ship something", or "what do I do next?". Figures out where the user is in the journey (no cluster yet → cluster but no app → app deployed but no URL → ready to update) and routes to the right skill.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ext-devplatform-plugin:getting-startedThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
People who want to run an app on the Intility Developer Platform but don't have a deep Kubernetes background. The goal is to get them to a working, exposed app quickly, then help them learn as they go.
People who want to run an app on the Intility Developer Platform but don't have a deep Kubernetes background. The goal is to get them to a working, exposed app quickly, then help them learn as they go.
Keep the tone friendly and short. Don't lecture. When the user is about to learn something new, show them the command first, then a one-line explanation — not a paragraph.
This skill is a router. It detects what state the user is in and hands off to the right skill. It does not do the work itself.
Before anything else, confirm the CLIs the plugin depends on are present:
command -v indev >/dev/null 2>&1 && echo "indev: ok" || echo "indev: MISSING"
command -v oc >/dev/null 2>&1 && echo "oc: ok" || echo "oc: MISSING"
command -v docker >/dev/null 2>&1 && echo "docker: ok" || echo "docker: missing (optional)"
indev missing → stop. On macOS the quickest path is brew install intility/tap/indev; otherwise https://developers.intility.com has instructions. Nothing else works without this.oc missing → stop. macOS: brew install openshift-cli; otherwise same page as above.docker missing → fine to continue. Only prepare-app uses it, and only when an image still needs to be built or checked — it helps the user install Docker at that point.If everything is present, don't say anything — move on silently. Only narrate the check if something is missing.
Run these checks silently to figure out where the user is. Don't narrate every check.
indev account show
indev cluster list -o json 2>/dev/null
oc whoami 2>/dev/null
oc whoami --show-server 2>/dev/null
From the results, classify the situation:
| Situation | Signal |
|---|---|
Not logged in to indev | indev account show fails |
| No cluster yet | indev cluster list returns empty / no clusters and oc whoami fails |
| Has a cluster, not logged in | Cluster exists, oc whoami fails |
| Logged in, no app deployed | oc whoami works, but the user hasn't mentioned an app namespace |
| Has an app, needs a URL | User says it's deployed but not reachable |
| Wants to update | User mentions a new image / new version |
oc whoami working beats an empty indev cluster list. Clusters are sometimes provisioned by platform admins or a colleague, so they don't show up in this user's indev cluster list. If oc whoami succeeds, the user has a cluster — treat the situation as "has a cluster, logged in" and never route to create-cluster. Use oc whoami --show-server to tell them which cluster they're connected to.
If the situation is obvious from the user's message, just continue. Otherwise ask one short question:
Q: "What do you want to do?"
Options:
- "Get a cluster + deploy my first app"
- "Deploy an app on the cluster I already have"
- "Expose an app I already deployed"
- "Update an app I already deployed"
Only include options that match the detected state. If they have no cluster, don't offer "expose an app".
Based on the destination, invoke the right skill in this order. Do not do these yourself — call the skill.
| Goal | Skill order |
|---|---|
| First time, full journey | create-cluster → login → prepare-app → deploy-app → expose-app |
| Has cluster, new app | login (if needed) → prepare-app → deploy-app → expose-app |
| Just need a URL | expose-app |
| Just bumping a version | update-image |
| Coming back after a break — what's running? | status |
If the user has a cluster and apps already deployed but doesn't remember what's there, run status first so they (and you) have an accurate picture before deciding what to do next.
Between skills, give the user a one-line "what's next" so they know the journey isn't over. Example:
Cluster ready. Next: I'll get your app running on it.
After the app is exposed, give a short recap:
You're live at: http://<hostname>/
To ship a new version later, just ask: "update <app> to <new-image-tag>"
To add another app, just ask: "deploy another app" — it'll go in its own namespace on the same cluster.
prepare-app can walk them from "no Dockerfile, no registry" to a pushed image).See references/usage-examples.md for full conversational walkthroughs of common journeys (first-time setup, coming back after a break, adding a second app, missing prerequisites).
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
npx claudepluginhub intility/ext-devplatform-plugin --plugin ext-devplatform-plugin