metrics

var JobsSpawnedTotal = promauto.NewCounter(
	prometheus.CounterOpts{
		Name: "agent_executor_jobs_spawned_total",
		Help: "Total number of jobs successfully spawned.",
	},
)

JobsSpawnedTotal counts successfully spawned jobs.

var ReconcilePassesTotal = promauto.NewCounterVec(
	prometheus.CounterOpts{
		Name: "executor_reconcile_passes_total",
		Help: "Total number of reconcile passes by outcome (ok, vault_unavailable, list_failed, aborted).",
	},
	[]string{"result"},
)

ReconcilePassesTotal counts reconcile pass outcomes, labelled by result. The failure reasons mirror ReconcileOnce's log events, and `ok` is counted on every pass that gets past its List — so an alert can fire on a backstop that never succeeds. That is the exact shape of the 2026-09 outage: the reconcile loop logged `reconcile_list_failed` once a minute and no other reconcile outcome was ever recorded, for months. A failure-only counter cannot express it — it saturates and then looks identical to a healthy quiet fleet, which is why the pass is counted on both sides of the outcome.

var ReconcileRedrivenTotal = promauto.NewCounter(
	prometheus.CounterOpts{
		Name: "executor_reconcile_redriven_total",
		Help: "Total number of tasks re-driven by the reconcile loop.",
	},
)

ReconcileRedrivenTotal counts tasks re-driven by the reconcile loop after a restart dropped their in-memory deferral state (spec 005).

var SkippedUnknownAssigneeTotal = promauto.NewCounterVec(
	prometheus.CounterOpts{
		Name: "agent_executor_skipped_unknown_assignee_total",
		Help: "Total number of tasks skipped because their assignee matches no agent Config CR, by assignee.",
	},
	[]string{"assignee"},
)

SkippedUnknownAssigneeTotal counts tasks skipped because their assignee matches no agent Config CR, labelled by the unknown assignee name. The bare TaskEventsTotal{result="skipped_unknown_assignee"} counter says a skip happened but not which assignee is failing; this label makes the signal actionable. Observed 2026-08-26 → 2026-09-03: 12 tasks stranded 8 days with only the bare counter moving, invisible as a routable alert.

var TaskEventsTotal = promauto.NewCounterVec(
	prometheus.CounterOpts{
		Name: "agent_executor_task_events_total",
		Help: "Total number of task events processed.",
	},
	[]string{"result"},
)

TaskEventsTotal counts task event processing outcomes.