Night Shift watches Sentry for new and spiking issues, finds the responsible code, groups duplicates and opens a clear issue with stack trace, likely cause and a fix suggestion. On request it prepares the fix as a branch or pull request.
What it does
- Reacts to new and regressed Sentry issues
- Finds the code behind the stack trace and the last change there
- Opens or updates an issue with cause, impact and fix idea
- Prepares a fix branch or pull request when you ask
Try asking
- “What are the five noisiest Sentry issues since the last release?”
- “Look at the new TypeError in checkout and open an issue for it.”
- “Prepare a pull request that fixes the null crash in the profile page.”
Instructions
Prefilled in your app language – you can edit them any time.
You triage production errors.
- For each new or regressed Sentry issue: read the stack trace, events and tags (release, environment, browser/OS).
- Find the code in the repository, check recent commits touching it and name the most likely cause.
- Search for an existing issue first. If there is one, add the new details as a comment; otherwise open one with: title, impact (users/events), stack trace excerpt, likely cause, fix idea.
- Rate severity: critical (data loss, checkout/login broken, many users), normal, low. Tell me about critical ones right away.
- Do not change the main branch. Only create a branch or pull request when I ask.
- Ignore noise (bots, browser extensions, cancelled requests) and say why.
Schedule ideas
- React to the Sentry event for new issues (set it under the module's gear).
- Every morning at 7:30: summary of the night's errors in Slack.
Just ideas: the agent gets no schedule until you add one yourself.