Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Work on Luxury Yacht kubeconfig selection, multi-cluster client lifecycle, auth failure/recovery, selected/background clusters, cluster tabs, refresh subsystem rebuilds, and object catalog lifecycle
Cluster Auth Lifecycle
Use this when touching kubeconfig selection, cluster client setup, auth failure
overlays, retry/recovery, selected/background cluster state, cluster tabs,
refresh subsystem rebuilds, object catalog start/stop, or tests for cluster
add/remove behavior.
Read First
AGENTS.md
backend/AGENTS.md for backend lifecycle/client changes
frontend/AGENTS.md for frontend cluster state or UI changes
docs/architecture/multi-cluster.md
docs/architecture/data-freshness.md
docs/architecture/auth.md
docs/architecture/refresh-system.md
docs/architecture/catalog.md when object catalog lifecycle is involved
Backend Entry Points
backend/cluster_runtime_manager.go
backend/cluster_workspace_projection.go
backend/refresh_coordinator.go
backend/workspace_coordinator.go
backend/cluster_runtime_intent.go
backend/cluster_runtime_clients.go
backend/cluster_runtime_auth.go
backend/workspace_auth.go
backend/workspace_cluster_clients.go
backend/workspace_kubeconfigs.go
backend/workspace_state.go
backend/refresh_setup.go
backend/refresh_update.go
backend/refresh_subsystems.go
backend/refresh_recovery.go
backend/refresh_object_catalog.go
backend/internal/authstate
Frontend Entry Points
frontend/src/modules/kubernetes/config
frontend/src/modules/cluster
frontend/src/ui/layout/ClusterTabs.tsx
frontend/src/ui/overlays/AuthFailureOverlay.tsx
frontend/src/core/refresh
frontend/src/core/data-access
Checklist
Every cluster-data path carries clusterId; never infer the active
cluster in a backend/API/cache/action path.
Selected and background cluster sets stay distinct and refresh scopes are
rebuilt or cleared when cluster selection changes.
Refresh domains stay single-cluster. Cross-cluster displays fan out only
for clusters they display; retained inactive tabs stay passive.
Cluster add/remove updates aggregate refresh handlers and object catalog
services through the live update path, not only initial setup.
Watcher, auth, and transport callbacks publish non-blocking typed
ClusterRuntimeIntent values; Workspace is the single consumer and routes
accepted generations through the serialized selection mutation.
Auth-failed clusters do not block healthy clusters.
Retry/recovery rebuilds transport, refresh, object catalog, and frontend
diagnostics consistently.
Closing/removing a cluster cleans up streams, sessions, stale scopes, and
catalog state for that cluster.
Every frontend cluster-tab open/close affordance routes through
KubeconfigContext's unified selection transition (openKubeconfig,
closeKubeconfig, or setSelectedKubeconfigs); do not splice selected
clusters locally or call generated backend selection/close commands from
UI surfaces.
Tests cover at least one multi-cluster or auth-failure transition.
Validation
Use focused checks while iterating:
mise exec -- go test ./backend ./backend/internal/authstate
mise exec -- npm run typecheck --prefix frontend
mise exec -- npm run test --prefix frontend -- cluster kubeconfig auth refresh
