- Home
- Skills
- DevOps & Cloud
- Kubernetes Manifest and Helm Chart Writer and Reviewer
Works with the AI Agents you already use
Kubernetes Manifest and Helm Chart Writer and Reviewer
Write or review Kubernetes manifests, Helm charts and Kustomize overlays with restricted security, probes, HPA, PDB and NetworkPolicy. PRODUCTION READY / HARDEN FIRST / UNSAFE.
$5
Kubernetes Manifest and Helm Chart Writer and Reviewer
Example session with this skill installed
Here is the Dockerfile and deployment.yaml for our Node orders-api (port 3000). Review the deployment and turn it into a Helm chart we can use in prod. We use External Secrets and a shared Gateway.
- Read your context and instructions
- Compiled the kubernetes manifest
Kubernetes manifests: orders-api (review + helm)
Verdict: UNSAFE · DATABASE_URL with a password is inline in the Deployment, and the container runs as root.
Facts from source
- Port 3000 · no USER in Dockerfile (root) · writable: /app/logs, /tmp · health: none defined
- Image
orders-api:latest, 1 replica, no requests or limits
Findings
| # | Severity | Object | Finding | Fix |
|---|---|---|---|---|
| 1 | CRITICAL | Deployment | DATABASE_URL value inline in env | Move to Secret orders-api-db, use secretKeyRef; rotate the password |
| 2 | HIGH | Deployment | Runs as root, no seccomp, no capability drop | Add USER 10001 to Dockerfile; restricted securityContext |
| 3 | HIGH | Deployment | :latest image | Pin 1.8.0@sha256:<digest> (digest: open item) |
| 4 | HIGH | Deployment | No probes | startup + readiness /ready, liveness /healthz (app must expose them) |
| 5 | MEDIUM | Deployment | No requests; 1 replica | requests 100m / 256Mi (estimate), 2 replicas, HPA 2-6 |
| 6 | MEDIUM | (missing) | No PDB, no NetworkPolicy | PDB maxUnavailable: 1; default-deny + DNS + gateway allow |
| 7 | LOW | Ingress | ingress-nginx class | Plan move to HTTPRoute; controller is retired |
Chart (excerpt)
orders-api/
Chart.yaml values.yaml values.schema.json
templates/ _helpers.tpl deployment.yaml service.yaml httproute.yaml
hpa.yaml pdb.yaml networkpolicy.yaml serviceaccount.yaml NOTES.txt
# values.yaml
image:
repository: registry.example.com/orders-api
tag: "1.8.0"
digest: "" # set sha256 digest; overrides tag when present
resources:
requests: { cpu: 100m, memory: 256Mi } # estimate, tune from metrics
limits: { memory: 256Mi } # no CPU limit (avoids throttling)
existingSecret: orders-api-db # created by External Secrets
Validation
helm lint --strict: 1 chart(s) linted, 0 chart(s) failedkubeconform -strict -summary: run it after you pin the digest; kube-linter not installed (not run)
Open items
Image digest · add /healthz and /ready to the app · USER 10001 in Dockerfile
Apply (only after you say yes)
helm upgrade --install orders-api ./orders-api -n orders -f values-prod.yaml
Connects securely to your tools. The creator never sees your data.
About this skill
Most Kubernetes YAML that reaches production was copied from a tutorial: root containers, no probes, :latest images, secrets pasted into env vars, and no PodDisruptionBudget, so the first node drain takes the app down. Asking an AI from scratch usually gives you a Deployment and a Service and stops there.
This skill gives your coding agent a fixed procedure for writing and reviewing manifests:
- Reads your Dockerfile, compose file or description and lists the facts it found (port, user, writable paths, health endpoint)
- Writes a Deployment aligned with Pod Security Standards restricted: runAsNonRoot, seccomp RuntimeDefault, drop ALL capabilities, no privilege escalation, plus a read-only root filesystem with emptyDir mounts
- Adds startup, readiness and liveness probes, requests and a memory limit, and explains the CPU-limit tradeoff instead of guessing
- Adds Service, Gateway API HTTPRoute (or Ingress), HPA, PDB, topology spread, default-deny NetworkPolicy, ServiceAccount and the recommended app.kubernetes.io labels
- Never inlines secrets; points to External Secrets, Sealed Secrets, SOPS or the CSI driver
- Packages as plain YAML, a Helm chart with commented values.yaml and helpers, or Kustomize dev/staging/prod overlays
- Validates with helm lint, helm template and kubeconform, with optional kube-linter and Polaris
- Review mode returns a severity-ranked findings table and a PRODUCTION READY / HARDEN FIRST / UNSAFE verdict
For developers and platform engineers shipping a service to Kubernetes, or cleaning up charts before an audit.
Honesty note: the skill never applies anything to a cluster without your explicit yes, never fabricates image digests or tool output, and labels resource numbers as estimates. Facts were checked against kubernetes.io and helm.sh docs as of October 2026; it re-checks versions when your agent has web access.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Resource requests are estimates until you have real metrics. It does not create clusters, install Gateway API controllers or secret operators, or write CRDs. Offline validation relies on locally installed tools, and anything that changes a cluster runs only after you confirm.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 29 days ago
Nothing to install. It runs in your AI agent.
- Optional
- Helmkubeconformkubectlkube-linterorPolaris
kube-linter or Polaris adds extra checks. No cluster access is needed to write or review.