generating-chart

Generate Helm Chart

Generate or customize a Helm chart for Kubernetes workloads following battle-tested patterns from production services.

Philosophy

• Workload-appropriate defaults - Batch workers get exec probes and no service; web services get HTTP probes and ingress
• Security by default - Non-root, drop ALL capabilities, readOnlyRootFilesystem
• Config from environment - ConfigMap for non-sensitive, Secret (stringData) for sensitive values
• Checksum annotations - Auto-restart pods on config/secret changes
• Production extras are opt-in - KEDA, NetworkPolicy, PDB, ServiceMonitor only when requested

Workflow

1. Detect Existing Chart

Glob: chart/Chart.yaml, chart/values.yaml

• If chart/ exists -> customize mode (read existing, present findings, ask what to change)
• If chart/ doesn't exist -> generate mode (ask questions, run helm create chart, customize)

2. Ask Workload Type

Present via AskUserQuestion:

Option	Description
Batch worker (Recommended)	Poll-then-exit, replicaCount:0, exec probes, no service. For SQS/Kafka consumers.
Web service	HTTP endpoints, service, ingress, HTTP probes. For APIs and web apps.
CronJob	Scheduled batch, concurrencyPolicy:Forbid. For periodic tasks.

3. Ask Production Extras

Present via AskUserQuestion (multiSelect: true):

Option	Template	Description
KEDA ScaledObject	scaled-object.yaml	Queue-based autoscaling (SQS, Kafka)
NetworkPolicy	network-policy.yaml	Restrict egress to specific services
PodDisruptionBudget	pdb.yaml	Protect against voluntary disruptions
ServiceMonitor	service-monitor.yaml	Prometheus metrics scraping
None	--	Core templates only

4. Detect Config Patterns

Scan project files to auto-populate configmap/secrets keys and detect language/probe commands.

Full detection logic: references/config-detection.md

5. Load References

Read the following reference files based on selections:

Always	references/common-templates.md
Batch worker	references/batch-worker.md
Web service	references/web-service.md
CronJob	references/cronjob.md
Any extras selected	references/production-extras.md
Generate mode	references/helm-create-customizations.md

6. Generate Mode

Run helm create chart, then apply workload-specific customizations (batch worker, web service, or CronJob) and production extras.

Full customization checklists per workload type: references/generate-mode.md

7. Customize Mode

1. Read all files in chart/templates/ and chart/values.yaml
2. Present findings:
   • What the chart has (workload type, probes, config, security)
   • What's missing from best practices
   • What's non-standard
3. Ask which customizations to apply via AskUserQuestion
4. Apply selected changes

8. Report

Helm chart generated:

chart/
  Chart.yaml              - {name} v{version}
  values.yaml             - {workload_type} pattern
  templates/
    _helpers.tpl           - Standard helpers
    {workload_template}    - {description}
    configmap.yaml         - {n} config keys
    secrets.yaml           - stringData pattern
    NOTES.txt              - Operational notes
    {extras...}            - {descriptions}

Configuration:
  Workload: {type}
  Replicas: {count}
  Probes: {probe_type}
  Security: {security_summary}

Next steps:
  helm lint ./chart
  helm template {name} ./chart
  helm template {name} ./chart --set image.tag=test-123

To generate Makefile.deploy with build/push/deploy targets:
  /makesmith:deploy

9. Hand Off to makesmith

After chart generation, ask via AskUserQuestion:

Option	Description
Run /makesmith:deploy now	Generate Makefile.deploy with Helm deployment targets for this chart
Skip	I'll set up deployment separately

If the user chooses to run it, invoke the makesmith:generating-deploy skill. Do NOT duplicate Makefile.deploy generation logic in helmcraft — that is makesmith's responsibility.