dusk bot
routinessentry error digest

dusk botroutines

a morning digest of unresolved sentry errors

give the morning engineer an issue to open for each error that kept happening overnight. this routine reads unresolved sentry issues with production events in the last day.

by the dusk bot team · checked · how we made this

you get one line per issue in dusk chat, with the latest occurrence you can inspect. repeated events stay together.

pixel art: an orange pixel bot, a yellow pixel bot and an indigo pixel bot beside the sentry mark on a white tile, a newspaper and an alarm clock and a calendar, on a sandy path in front of a lily pond, one saying "seen again overnight"

the message

paste this into a chat with your bot

every day at 8am, list unresolved sentry issues with production events in the past 24 hours. use only project id 427136 in the organization millbank-tools.

follow the issue pages. for each group, read its events in that same window and environment. take the newest event time you can verify. in this chat list the five most recently seen groups, newest first. include the issue id, error title, latest event time with zone, and owner if available. say how many other groups qualified.

never resolve or assign an issue. leave raw event bodies out of chat. stay quiet if the whole search succeeds with no matches. if any read fails, name the missing coverage.
when
daily at 8am in your time zone, looking back 24 hours
needs
sentry read access to your organization, the project’s issues and their events. replace the organization slug and project id with yours.

what comes back

an example, not a real run

production · past 24 hours · 08:00 utc
web-418 · timeout saving delivery address · latest 07:42 utc · owner nell finch
web-405 · missing receipt template · latest 03:16 utc · no owner listed
no other groups qualified.

make it yours

  • pick the handoff hour. schedule the digest just before someone takes the queue. keep the time zone beside each event.
  • separate test traffic. use the exact environment name your project uses for live traffic.
  • choose the cutoff. for a weekend handoff, widen the event window and label it in the digest.

when it goes wrong

  • an old error looks new. the latest event tells you when it happened again. open the issue before calling it a new defect.
  • a familiar error disappears. compare the project id, environment and time window with a known event. a mistyped filter can return an empty list.1

what the handoff leaves out

this digest answers which errors recurred and still need a look. a burst that someone resolved overnight falls outside the search. if you need an incident timeline, include resolved issues in a separate review.

a morning routine is the wrong pick for waking the on-call engineer during an outage. keep your incident alerts for that job.

questions

does the first issue have the biggest impact?

the order tells you which error happened most recently. an event at 07:59 can outrank an older outage. use the issue details to judge severity.

why does the same id appear on consecutive mornings?

the error happened in both windows and still qualifies. keeping the id makes it easier to follow the investigation across handoffs.

sources

  1. composio: sentry actions, checked 20 november 2026
  2. dusk bot docs: routines
  3. dusk bot docs: connected apps

drafted with ai help from the dusk bot docs and the sources listed on this page, then checked against them. if something here is wrong or out of date, write to [email protected].