Helm
Install applications onto an Nuphos-managed Kubernetes cluster using the helm CLI from inside the sandbox. Use this for chart-based deploys ("install Inngest", "deploy ingress-nginx", "set up cert-manager"). Includes an install script for the helm binary.
Get the complete skill folder, including its scripts and reference files.
When to use this skill
Use Helm when an AI agent needs a reusable SKILL.md workflow for this job: Install applications onto an Nuphos-managed Kubernetes cluster using the helm CLI from inside the sandbox. Use this for chart-based deploys ("install Inngest", "deploy ingress-nginx", "set up cert-manager"). Includes an install script for the helm binary.
When not to use it
Skip Helm when the task is outside the productivity category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.
How to install
Where will you use it?
mkdir -p ~/.claude/skills/helm # Save SKILL.md to ~/.claude/skills/helm/SKILL.md
- Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/helm/SKILL.md. For a shared project, commit the folder at .claude/skills/helm/ instead.
- Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
- For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.
Full install guide for Claude, Cursor, and Codex →
Try a starter prompt
After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.
Use the Helm skill for this task: Use Helm when this documented workflow matches the task. Confirm any required inputs or files before making changes.
What supports this recommendation?
70/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.
- Documentation
- 77/100
- Practical detail
- 68/100
- Source evidence
- 63/100
- Source signals
- 70/100
Model compatibility evidence
- claudeuntested
No model-specific signal or recorded compatibility test was found.
- chatgptuntested
No model-specific signal or recorded compatibility test was found.
- geminiuntested
No model-specific signal or recorded compatibility test was found.
- copilotuntested
No model-specific signal or recorded compatibility test was found.
- llamauntested
No model-specific signal or recorded compatibility test was found.
- perplexityuntested
No model-specific signal or recorded compatibility test was found.
- mistraluntested
No model-specific signal or recorded compatibility test was found.
- grokuntested
No model-specific signal or recorded compatibility test was found.
Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.
What this skill does
# helm
Use this skill when the user wants to **install or upgrade** something packaged as a Helm chart on one of their Nuphos clusters — e.g. "幫我部署 Inngest", "install ingress-nginx", "set up cert-manager". For inspection (`get pods`, `logs`, etc.) keep using `kubectl` — `helm` is for chart-driven deploys.
Helm reuses the same kubeconfig the `kubectl` skill writes to `~/.kube/config`, so the typical workflow is:
1. Pick the target cluster context (see the `kubectl` skill — its kubeconfig has a context for every cluster this session can reach). 2. Use `web_search` + `web_fetch` to confirm the chart's official repo / OCI URL and the recommended install instructions, then run `helm --kube-context <context> install ...`.
## Setup
```bash # Sanity check — pass the target context explicitly on every helm command. kubectl config get-contexts helm version --short helm --kube-context <context> list -A ```
### Fallback: helm or kubectl missing
If for some reason the binary isn't there (cold start without the runtime image), run the install script(s) — they're idempotent:
```bash bash skills/kubectl/scripts/install.sh bash skills/helm/scripts/install.sh ```
## Discovering the chart
Don't guess install commands. For any project the user names, do this first:
1. `web_search` for `"<name> helm chart"` or `"<name> install kubernetes"` to find the official source. 2. `web_fetch` the most authoritative result (project docs, GitHub README, ArtifactHub page) to read the exact `helm repo add` / `helm install` block. 3. Note any required dependencies (cert-manager, an ingress controller, a CRD pre-install step, persistent storage).
## Common operations
Every cluster-facing command needs `--kube-context` — the kubeconfig deliberately has no current context, so a command without it fails or, worse, lands on whatever a locally-loaded config left behind.
```bash ctx="gcp/my-project/prod" # from `kubectl config get-contexts`
# Add and update a classic chart repo (local-only, no cluster contact) helm repo add <name> <url> helm repo update
# Install / upgrade with custom values helm --kube-context "$ctx" upgrade --install <release> <chart> \ --namespace <ns> --create-namespace \ --version <chart-version> \ --set key=value \ --values values.yaml
# OCI-hosted charts (no `repo add` needed) helm --kube-context "$ctx" upgrade --install <release> oci://<registry>/<chart> --version <chart-version> -n <ns> --create-namespace
# Inspect helm --kube-context "$ctx" list -A helm --kube-context "$ctx" status <release> -n <ns> helm --kube-context "$ctx" get values <release> -n <ns> helm --kube-context "$ctx" get manifest <release> -n <ns> helm --kube-context "$ctx" history <release> -n <ns>
# Diff before applying (if helm-diff plugin is available) helm --kube-context "$ctx" diff upgrade <release> <chart> -n <ns> --values values.yaml
# Rollback / uninstall helm --kube-context "$ctx" rollback <release> <revision> -n <ns> helm --kube-context "$ctx" uninstall <release> -n <ns> ```
## Safety
- **Always confirm before installing.** Summarize: chart name & version, target namespace, key values (storage class, replicas, public exposure, default credentials). Wait for explicit user approval before running `helm install` / `helm upgrade`. The MongoDB-style "要我就以這個配置部署嗎?" pattern is good. - **Pin the chart version** with `--version <x.y.z>` so the deploy is reproducible. Don't blindly install latest. - **Default to a dedicated namespace** with `--namespace <release>` `--create-namespace`, not `default` and not `kube-system`. - **Quote secrets, don't echo them.** When generating passwords for a chart, write them via `--set-string foo.password=$(openssl rand -hex 16)` or a values file, not directly into the chat. - **Don't expose to the public internet by default.** If the chart has an ingress / LoadBalancer service, ask the user whether they want it exposed before enabling it. Prefer ClusterIP + port-forward for first install. - **Uninstall is destructive.** `helm uninstall` removes resources (and often PVCs depending on chart). Confirm before running and warn that data may be lost. - **No `--force` upgrades** unless the user asks — that recreates resources and can cause downtime.
Intended uses
- Use Helm when this documented workflow matches the task.
Implementation notes
- Review the source instructions and adapt inputs before running the workflow.
Try it, then tell us how it went
A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.