| name | watch-deploys |
| description | Watch an in-flight deployment or Flux reconciliation until it settles or fails, reporting state transitions as they happen. Use for "watch the deploy", "monitor the rollout", "keep an eye on flux", or when a change was just pushed and its landing needs following. Scoped to the Artemis cluster. |
Skill: Watch Deploys
Delta-aware cluster state watcher. Use with /loop 30s /watch-deploys during active deployments.
Each tick shows only what changed since the last tick — not the full cluster state.
State Cache
Store the previous snapshot at /tmp/artemis-watch-deploys.json. On first run (no cache), show full current state as the baseline and write cache.
What to Track
flux get kustomizations -A
flux get helmreleases -A
kubectl get pods -A -o json
kubectl get nodes -o json
Use the ops tier's k8s-pods_list / k8s-resources_list tools if local kubectl is unavailable.
Delta Logic
Compare current snapshot to cached snapshot:
| Change type | What to show |
|---|
| Kustomization/HelmRelease: NotReady → Ready | RECOVERED: <name> (<ns>) |
| Kustomization/HelmRelease: Ready → NotReady | BROKE: <name> (<ns>) — <reason> |
| Kustomization/HelmRelease: newly Suspended | SUSPENDED: <name> (<ns>) |
| Pod restart count increased | RESTARTED: <pod> (<ns>) — restarts: N→M |
| Pod phase changed | PHASE CHANGE: <pod> (<ns>) — <old>→<new> |
| Node readiness changed | NODE: <name> — Ready→NotReady |
Output Format
Tick [20:14:32 ET] — Δ since 20:14:02
RECOVERED: immich (default) HelmRelease → Ready
BROKE: komf (media) HelmRelease → Failed: "container image not found"
RESTARTED: memini-xxx-yyy (cortex) restarts: 0→1
✓ Nodes: all ready
✓ No new Flux suspensions
If nothing changed: ✓ No changes since 20:14:02
Cadence
- Active deploy:
/loop 30s /watch-deploys
- Idle monitoring:
/loop 2m /watch-deploys
- Stop when all resources are back to Ready
