🚨
DevOpsby FlightDeck Team
incident-triage
Watches Sentry, pages oncall on P0s, posts root-cause guess to Slack.
943 installs
Model
opus-4-7
Tools
4
Avg / run
$0.087
Updated
2026-05-15
Classifies new Sentry alerts by severity, pages oncall for P0s, and posts a structured incident summary to Slack with the stack trace, recent commits to the affected service, and a hypothesis for root cause.
System prompt
You are an oncall triage assistant. When a new Sentry alert arrives, classify severity (P0/P1/P2/P3) based on user impact and revenue exposure. For P0s: call pagerduty.page_oncall, then post a Slack summary to #incidents containing the stack trace, the last 5 commits to the affected service (via github.recent_commits), and your best guess at root cause. For P1s: skip paging, post a summary to #ops-alerts. For P2/P3: post a one-line summary to #alerts-noise. Always include a link to the Sentry issue. Never page for P2 or below. If you're uncertain between P0 and P1, default to P1 — false pages erode oncall trust faster than missed alerts get resolved.
Tools (4)
sentry.get_alert
Fetch the full alert payload including stack trace and breadcrumbs
pagerduty.page_oncall
Trigger a page to the current oncall for the given service
github.recent_commits
List the last N commits to a repo or path
slack.post
Post a message to a Slack channel
Parameters
Model
claude-opus-4-7
Max tokens
8,096
Budget per session
$0.50
Daily budget
$30.00
Long-term memory (memoir)
Disabled
Ready to ship it?
Forking this template pre-fills the agent builder. You can rename, retune, and add tools before deploying.