UNOFFICIAL/EXPERIMENTAL
These skills are reusable operating knowledge for humans and agents working on helm-expt. They are not product proof tables, and they do not make Pilot or any other external demo runner part of helm-expt.
Use them when a chart needs more than a static render check:
| Skill | Use when |
|---|---|
| Live Parity | Running Helm-vs-ConfigHub live parity, GitOps/OCI evidence, kind clusters, or cub-scout observations. |
| Large App Evidence Funnel | A chart has many Units and a single pass/fail status hides where the operation is stuck. |
| Target Facts And Lifecycle | A base needs Secrets, CRDs, APIService readiness, topology, storage, or generated runtime state before apply. |
| Hook And Secret Lifecycle | A chart uses Helm hooks, webhook certificates, service-account token Secrets, or generated Secret material. |
| Serious Chart Playbooks | Working on kube-prometheus-stack, Consul, cert-manager, External Secrets, Loki, or other hard charts. |
Common Rules
- Keep live lanes serial.
- Use the current
cub installercommand family. - Do not turn
watchorblockedintopassunless the receipt contains the evidence. - Target shape means Kubernetes topology and prerequisites, not ConfigHub worker architecture.
- ConfigHub OCI/GitOps delivery remains workerless from the target cluster's point of view: Argo or Flux pulls desired state from the published artifact.
- Scoped checks are normal while working. Run full
npm run verifyonly for a release gate or when explicitly requested.
Regenerate or verify the skill set:
npm run skills:verify