Triage a Rootly alert with AI
Check the alert against its underlying signal, current incidents and ownership information. Get an evidence-backed recommendation without acknowledging or changing the alert.
Triage a Rootly alert with AI when an alarm arrives beside a seemingly related outage. The alert-triage skill checks the payload, affected scope and available corroboration before deciding whether to recommend a link, a separate incident or further investigation. It records why the recommendation would change if new evidence arrives.
An alert labelled “checkout latency” might actually measure payment errors. A search outage in another region may have started at the same time. A defensible triage note flags the label mismatch, identifies any verified dependency and leaves an outdated team route open for confirmation.
What it does
- Check which signal crossed its threshold.
- Compare symptoms and regions with an existing incident.
- Look for corroborating metrics or permitted log samples.
- Flag a stale ownership route before anyone is paged.
- Explain the evidence needed to change the recommendation.
Install and use
- Install the skill and connect the official read-only Rootly MCP for authorised alert and incident records. Add permitted supporting charts or logs for the relevant interval.
- Ask, “Triage this checkout alert. Compare it with the open incident and explain whether the evidence supports attaching it.”
- Review the proposed disposition before an authorised responder acknowledges, links or escalates anything.
FAQ
Source
Made by Rootly. Original skill: alert-triage, published on GitHub under the Apache-2.0 licence.
View on GitHub