Triage Sentry Issue

Triage a Sentry Issue

Use this skill when an agent must decide how serious a Sentry issue is and where it fits in the queue.

Workflow

1. Gather issue metadata

Fetch the issue and record:

• title, level (error/fatal/warning), status, and age,
• event count and unique users (volume and reach),
• first seen / last seen (is it new, growing, or silent?),
• the primary culprit (file/function), and any assigned team/tag.

2. Assess impact

Build the triage picture:

• Volume: count + trend (rising/flat/falling) over the recent window.
• Reach: unique users affected and their share of active users.
• Severity: fatal/unhandled vs handled; does it affect a core flow?
• Freshness: new vs regressed (a regression changes the equation — see investigate-regression).

3. Assign priority

Propose a priority with the evidence:

• P0/P1: fatal, high volume or high-reach, core flow, or a regression.
• P2: recurring but contained error.
• P3: rare, cosmetic, or legacy path.

4. Recommend next action

• For high-priority: propose the investigation chain (investigate-regression → analyze-sentry-trace → debug-production-error).
• For low-priority: suggest grouping, ignoring, or scheduling — as a recommendation; mutating Sentry state (resolve, ignore, assign) is a write action requiring explicit user intent.

5. Report

Give: issue summary, impact metrics, priority with reasoning, and the recommended next step. Keep mutation of issue state out unless asked.