AI Observability Coverage Check

Observability tends to grow around the parts of the system that have produced incidents. The API has alerts because the API once had an outage. The database has alerts because the database paged on-call last quarter. The background jobs have no alerts because they have not failed visibly yet, and the queue has no alerts because nobody knew it could fail until it did. The shape of the monitoring is the shape of the past, not the shape of the risk surface.

The /vigil-check skill audits coverage against actual service behavior. It identifies critical user paths without alerts, background jobs without monitoring, queues without depth alerts, async processors without error rate tracking. The output is a gap-prioritized action plan so the team adds the alerts that would catch the next incident before it happens.

What the check covers

Four categories. Critical user paths: each user-facing flow checked for an alert that would catch its failure. Background jobs: scheduled jobs checked for alerts on stuck or failed runs. Queues and processors: depth alerts, error rate, dead letter monitoring. Coverage gaps: services without RED metrics, services without distributed tracing, services without log aggregation.

How /vigil-check works

The skill reads the service inventory, the alert configuration, and the recent traffic patterns to identify the critical paths. Each path is checked for an alert that would fire if the path broke. Background jobs and queues are checked for monitoring. The output is the prioritized gap list with proposed alert rules.

Tonone's /vigil-check skill audits observability coverage against critical user paths, background jobs, queues, and async processors with a gap-prioritized action plan.

Related skills

Install

/vigil-check ships with the Vigil agent in Tonone for Claude Code.