Live Parity Skill

UNOFFICIAL/EXPERIMENTAL

Use this skill when running live Helm-vs-ConfigHub parity, GitOps/OCI evidence, kind clusters, cub-scout observations, or chart-specific live blockers.

The goal is to produce a live receipt with a precise outcome:

• pass: live evidence proves the named lane for the chart, version, base, and target profile.
• watch: the main path works, but a named controller, lifecycle, or target residue still needs review.
• blocked: the run cannot proceed until a named prerequisite, target shape, or tool gap is fixed.

Rules

• Run only one live lane at a time.
• Do not start a live run while tests/live-helm-confighub-parity-test is active.
• Never delete another lane's cluster.
• Use the current cub installer command family.
• Do not claim production support from render parity or from a single local cluster receipt.

Flow

Preflight first:

npm run live-parity:run -- --preflight \
  --recipe recipes/<repo>/<chart>/<version> \
  --base <base> \
  --target-profile <profile>

Run only after preflight passes:

npm run live-parity:run -- \
  --recipe recipes/<repo>/<chart>/<version> \
  --base <base> \
  --target-profile <profile>

Inspect progress:

cat runs/live-helm-confighub-compare/<slug>/stage.txt
kind get clusters
kubectl --kubeconfig <kubeconfig> get applications -n argocd
kubectl --kubeconfig <kubeconfig> get pods,jobs,pvc -A

Use cub-scout when controller state or live object state needs to be summarized:

KUBECONFIG=<kubeconfig> cub-scout gitops status --json
KUBECONFIG=<kubeconfig> cub-scout scan -n <namespace> --json

After a run, verify the committed evidence:

npm run live-parity:verify
npm run live-parity:rerun-plan:verify
git diff --check

Output

The expected output is a receipt under:

runs/live-helm-confighub-compare/<chart-base>/receipt.yaml

If the row does not pass, update the rerun plan and keep the named watch or blocked reason visible in the status surfaces.