# sentry-errors — Auto-Triage Subagent for Sentry

> Open-source Claude Code subagent that pulls recent Sentry errors via MCP, groups by component, and suggests fix priorities. Inspired by Boris Cherny.

## Install

Save the content below to `.claude/skills/` or append to your `CLAUDE.md`:

# sentry-errors — Auto-Triage Subagent for Sentry

## Quick Use

1. Install Sentry MCP first: see TokRepo workflow id 665 (Sentry MCP).
2. Save the file below to `.claude/agents/sentry-errors.md`.
3. Restart Claude Code (or `/agents reload`).
4. Each morning, say:
   > "Use sentry-errors to triage what came in overnight."

You get a grouped, prioritized digest of Sentry errors with suggested next actions.

> Inspired by Boris Cherny's described workflow on howborisusesclaudecode.com — he uses Sentry MCP plus a triage subagent to "fetch and triage error logs autonomously." This is a community-written equivalent.

---

## Intro

sentry-errors is a Claude Code subagent that pulls recent Sentry errors via the official Sentry MCP, groups them by component and root cause, and emits a prioritized triage list. It is built for the morning-standup pattern: instead of you scrolling Sentry, the subagent reads it for you and tells you what actually matters.

Boris Cherny describes this exact pattern as part of his Claude Code setup — he names error triage as one of the things subagents handle for him while he works on something else. This entry packages that pattern.

Works with: Claude Code 1.x + Sentry MCP. Setup: under 2 minutes once Sentry MCP is configured.

---

## How sentry-errors Works

Save to `.claude/agents/sentry-errors.md`:

```markdown
---
name: sentry-errors
description: Pull recent Sentry errors, group them by component and root cause, suggest priorities. Use for morning triage or after a deploy.
tools: mcp__sentry__*, Bash, Read
---

You are the sentry-errors subagent. Your single responsibility is to triage Sentry — not fix bugs, not write code, just turn a noisy event stream into an ordered action list.

## Workflow

1. Fetch errors from the last 24h via Sentry MCP (`list_issues` or `list_events` with `dateFrom = 24h ago`).
2. Group by `culprit` (file + function) and root-cause hash.
3. For each group, capture: count, first/last seen, affected users, any new release link.
4. Score priority:
   - P0: > 100 occurrences AND > 10 affected users AND seen first in the last release
   - P1: > 100 OR new in the last release
   - P2: everything else
5. For P0 + P1 groups, attach a one-line hypothesis and the most-recent commit that touched the culprit file (`git log -1 --format='%h %s' -- <file>`).
6. Emit the report below.

## Output format

sentry-errors triage — last 24h
================================
Total events: N
Distinct issues: N
P0: N | P1: N | P2: N

P0:
1. <culprit> — N events, N users, first seen <release>
   Hypothesis: <one-line>
   Last touch: <commit_hash> <commit_subject>

P1:
1. ...

P2 (collapsed): N issues, see Sentry for details.

## Boundaries

- Do not open issues, file PRs, or fix bugs.
- Do not silence/ignore errors in Sentry without explicit user instruction.
- If Sentry MCP is not installed, escalate: "Install Sentry MCP first."
```

### When to use

- Daily morning standup (before you start coding).
- Right after a production deploy (15-minute check).
- During an incident — fastest grouped view of what is actually breaking.

### When not to use

- For a single specific error you are already debugging — just open Sentry.
- For long-tail historical analytics (use Sentry's UI).

---

## Example session

```
You:    "Run sentry-errors on overnight events."
Claude: -> Sentry MCP -> fetch 24h
        -> groups 1,247 events into 28 issues
        -> 1 P0, 4 P1, 23 P2
        -> reports:
           P0: src/checkout/charge.ts:88 — 412 events, 87 users, first in v3.4.2
              Hypothesis: Stripe API key rotation broke production
              Last touch: a1b2c3d "Rotate Stripe key"
```

---

## FAQ

**Q: Does it require Sentry MCP?**
A: Yes — Sentry MCP is the data source. Install it first (TokRepo id 665).

**Q: Will it suppress or ignore events?**
A: No — read-only by design. Triage only, no destructive writes.

**Q: How is priority scored?**
A: Volume + user impact + release proximity. The rules are explicit in the Workflow section so you can tune them.

**Q: Can I scope to a specific Sentry project?**
A: Yes — edit the Workflow's `list_issues` call to pass `project=<slug>`.

**Q: Is this Boris Cherny's actual subagent?**
A: No — community-written equivalent based on his public description.

---

## Source & Thanks

> Inspired by Boris Cherny's "fetch and triage error logs autonomously" pattern on howborisusesclaudecode.com.

Citations:
- howborisusesclaudecode.com (Sentry section)
- Sentry MCP (TokRepo id 665): https://tokrepo.com/en/featured
- Pragmatic Engineer: https://newsletter.pragmaticengineer.com/p/building-claude-code-with-boris-cherny

<!-- ZH -->

## 快速使用

前置：先装 Sentry MCP（TokRepo id 665）。

1. 落 `.claude/agents/sentry-errors.md`
2. `/agents reload`
3. 早会前问 Claude："Use sentry-errors to triage what came in overnight."

输出按 P0/P1/P2 分级，每条带 hypothesis + 最近一次改动 culprit 文件的 commit。

## 边界

- 只读不写：不开 issue、不关 issue、不静默
- 不修 bug：触发只输出 triage 报告，修复留给人/code-fixer
- Sentry MCP 没装就报错让你先装


---
Source: https://tokrepo.com/en/workflows/sentry-errors-auto-triage-subagent-for-sentry-676b8063
Author: Skill Factory