🧯Skill + MCP

Sentry issue triage

Groups open Sentry issues by root cause, assigns severity and owner, outputs a standup-ready table.

Surfaces: skill · mcp
Complexity: intermediate
Trigger: natural
Est. tokens: 6,200
Requires MCP: sentry

What It Does

Pulls the top unresolved Sentry issues, maps each stack frame to a file in the repo, and groups issues by inferred root cause. Assigns severity and an owner via CODEOWNERS or git blame, then outputs a Markdown table ready for standup.

When It Triggers

▸"Triage Sentry"
▸"Which bugs should we fix first?"
▸"Look at open Sentry issues"

Prerequisites

▸Sentry MCP connected
▸Repo checked out locally (for CODEOWNERS + blame)

SKILL.md

markdown

---
name: sentry-issue-triage
description: Triages open Sentry issues, groups by root cause, and suggests fixes. Uses Sentry MCP. Use when user says "triage Sentry", "look at open issues", or "which bugs should I fix first".
---

# Sentry Issue Triage

## Requires
- Sentry MCP connected

## Steps
1. Call `sentry_listIssues` with status=unresolved, sorted by event count
2. For the top 10 issues:
   - Fetch the latest stack trace via `sentry_getIssue`
   - Map the stack frame to a file in the repo
   - Classify: regression, new bug, known flake, env config
3. Group by inferred root cause (same top frame ⇒ same group)
4. For each group, propose:
   - Severity (P0/P1/P2)
   - Owner (via CODEOWNERS or git blame)
   - A one-line fix hypothesis
5. Output: Markdown table ready for standup

Variants

▸Run on a schedule via /loop to produce a fresh triage every morning.
▸Pipe the output into agent-orchestrates-skills to auto-open GitHub issues for each group.