Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Work on logs, shell exec, debug containers, port-forward, node drain/maintenance, long-running operations, permissions, lifecycle, and cleanup tests
Operations Workflows
Use this when touching logs, shell exec, debug containers, port-forward, node
drain/maintenance, session lifecycle, operation cancellation, permission-gated
actions, or related tests.
Core Contracts
Read:
AGENTS.md
backend/AGENTS.md
frontend/AGENTS.md
docs/workflows/logs/overview.md
docs/workflows/shell-debug.md
docs/architecture/permissions.md
docs/architecture/multi-cluster.md
docs/architecture/auth.md when cluster auth or recovery is involved
docs/workflows/operation-lifecycle.md when touching live operation
registry, cleanup, or status behavior
Backend Entry Points
backend/operations_coordinator.go
backend/runtime_operations.go
backend/refresh/containerlogsstream
backend/resources/pods/logs.go
backend/resources/nodes/logs.go
backend/app_log_service_commands.go
backend/shell_sessions.go
backend/resources/pods/debug.go
backend/portforward*.go
backend/nodemaintenance
backend/refresh/snapshot/node_maintenance.go
backend/refresh/snapshot/service.go
cluster lifecycle cleanup orchestrated by WorkspaceCoordinator and
ApplicationLifecycle in backend/cluster_runtime_clients.go,
backend/workspace_kubeconfigs.go, and backend/application_lifecycle.go
Requests and events carry clusterId and full target identity.
Runtime operation entries that target Kubernetes objects carry
clusterId, group, version, kind, and concrete namespace/name
where applicable.
Permission checks and capability reasons are visible in the UI.
Streams, sessions, and long-running operations clean up on close,
disconnect, cluster removal, auth failure, and app shutdown.
Cancellation/stop paths are idempotent.
Frontend cluster-tab close delegates to KubeconfigContext's unified
selection transition; backend selection cleanup handles removed-cluster
runtime operations instead of per-workflow cleanup in UI surfaces.
Shell backlog, port-forward details, and drain history remain owned by
their workflow stores; the runtime registry only owns global presence and
cleanup.
Sessions status renders shell sessions and port forwards only; active
drains may appear in cluster-close warnings but not as Sessions panel
detail rows.
object-maintenance keeps aggregate and node-specific scopes active
concurrently so node drain indicators and an open drain modal do not reset
each other.
object-maintenance remains uncached and singleflight-bypassed in the
backend snapshot service because it represents live app-managed drain
state.
Drain progress and history render in DrainNodeModal; keep the active or
most recent drain attempt visible after a drain starts or completes.
Frontend state resets on cluster/namespace/object changes.
Logs preserve transport-specific behavior documented in the logs docs.
Tests cover lifecycle, permission-denied, and cleanup behavior.
Validation
Use focused checks while iterating:
mise exec -- go test ./backend ./backend/resources/pods ./backend/resources/nodes ./backend/nodemaintenance
mise exec -- go test ./backend/refresh/snapshot
mise exec -- npm run typecheck --prefix frontend
mise exec -- npm run test --prefix frontend -- Logs Shell port-forward drain orchestrator
