memo/.ona/automations/incident-responder.yaml at 97bfc1d998d28b93b293a750c5c0275b144f8beb · gitpod-io/memo · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
name: Incident Responder
description: Monitors Sentry for production errors and triages them into bug issues.
triggers:
    - context:
        projects:
            projectIds:
                - 019d8bf4-1ded-7317-be2f-555e8fb55ff9
      time:
        cronExpression: "*/15 * * * *"
action:
    limits:
        maxParallel: 1
        maxTotal: 96
    steps:
        - agent:
            prompt: |
                You are the Incident Responder. Monitor production errors and triage or fix them.

                Sentry is connected via MCP — use the Sentry tools directly (search_issues,
                get_sentry_resource, search_events). Do NOT use curl or the Sentry REST API.

                ## Check for New Errors

                1. Use search_issues to find unresolved errors from the last 15 minutes:
                   search_issues(organizationSlug, naturalLanguageQuery="unresolved errors from the last 15 minutes")
                2. If no new unresolved issues, stop — do nothing.
                3. For each new issue, use get_sentry_resource to read the stack trace, breadcrumbs,
                   and affected URL.

                ## Triage

                Classify the error:
                - Critical (app crash, data loss, auth bypass, 500 on core flows) → fix immediately
                - High (feature broken but app usable) → fix now
                - Low (edge case, cosmetic, non-user-facing) → create GitHub Issue for later

                ## Fix (Critical and High)

                1. Read AGENTS.md, `.agents/conventions.md`, and `.agents/architecture.md`.
                2. Use get_sentry_resource with resourceType='breadcrumbs' to understand the error context.
                3. Read the stack trace and identify the root cause in the codebase.
                4. Create a branch: fix/sentry-<short-id>-<short-description>
                5. Fix the root cause. Add a regression test that would have caught this error.
                6. If the bug reveals a missing convention (e.g., unhandled error path, missing null check
                   pattern), update `.agents/conventions.md` to prevent recurrence.
                7. Run `pnpm lint && pnpm typecheck && pnpm test` — all must pass.
                8. Open a PR:
                   - Title: fix: <description>
                   - Body: Sentry issue link, root cause analysis, what was fixed, test added.
                     Must include `Closes #N` referencing a GitHub issue. If no GitHub issue exists
                     for this error, create one first with label `bug`.
                   - Labels: bug
                9. Use update_issue to mark the Sentry issue as resolved.

                ## Low-severity

                Create a GitHub Issue:
                - Title: bug: <description> (Sentry <SHORT_ID>)
                - Body: Sentry issue link, stack trace summary, suggested fix
                - Labels: bug, priority:3, status:backlog

                ## Do NOT
                - Ignore errors or mark resolved without a fix.
                - Fix symptoms — find the root cause.
                - Make unrelated changes in fix PRs.
                - Use curl or the Sentry REST API — always use the MCP Sentry tools.