Problem
Asynchronous standups needed ownership beyond asking three questions.
Constraints
Responses, schedules and actionable updates had to stay connected to Slack and JIRA.
What I did
SIYA was a curiosity project: a Slack agent for asynchronous daily standups.
The simple version asks three questions. The useful version owns the workflow: it knows the schedule, prompts the right people, tracks responses, summarizes team state, and syncs the actionable updates into JIRA.
System shape
A Lerna monorepo separated the Atlassian bot, the Slack bot, the Prisma data proxy, and the site. Cloudflare Workers handled the service surface. Durable Objects owned scheduling and live state. PlanetScale and Prisma backed persistence.
The design lesson
I modeled the standup as a repeated decision loop, not a form:
- who needs a prompt?
- which response maps to which issue?
- what should the team see?
- what belongs in the system of record?
That makes SIYA an early version of the workflow-to-agent pattern I still use: start with a ritual people already perform, isolate its decisions, give the system enough state to run the loop, and keep the result inside the tools the team already lives in.
Result
A Slack agent that scheduled prompts, tracked responses and synchronized updates into JIRA.
Tradeoffs
Workflow state adds complexity but makes the repeated standup process accountable and inspectable.