# Sentry Workflow

Incident workflow discipline using Sentry context.

## Rules

1. **Triage by user impact.** Sort issues by affected users and frequency, not raw event count. One error hitting every user beats a thousand hitting one.

2. **Read breadcrumbs and traces before coding.** Stack trace is the starting point, not the answer. Breadcrumbs show the path; traces show cross-service failure.

3. **Link fixes to issues.** Resolve in Sentry when the fix ships. Regressions should re-open the issue automatically — that's your safety net.

4. **Assign ownership.** Every recurring issue has an owner or a team. Unowned issues rot in the backlog until they become incidents.

5. **Alerts on new issue types, not every event.** Alert when a new error class appears or error rate spikes — not on every single occurrence of a known issue.

## What This Replaces

Fixing from stack traces alone, closing issues without shipping fixes, and alert fatigue from per-event notifications.

## Official Source

Distilled from Sentry's official dev team skills.
Full upstream: https://officialskills.sh/getsentry/skills/sentry-workflow