Inspiration

While working as a software engineering intern, I was introduced to Agile ceremonies like sprint planning and daily standups. Over time, I noticed a recurring pattern: teams often discovered sprint risks too late — during standups or retrospectives — when the damage was already done.

Standups were noisy. Alerts existed. But clarity was missing.

I kept wondering: why doesn’t Jira tell teams when a sprint is drifting, why it’s happening, and who needs to clarify it — before escalation is required?

That question led to Sprint Stoplight.


What it does

Sprint Stoplight continuously monitors live Jira sprint signals — progress vs time, scope change, workload imbalance, and stale issues — and computes a real-time sprint risk score.

Instead of auto-posting alerts, it focuses on guided intervention:

  • Detects sprint-level risk early
  • Identifies responsibility gaps
  • Drafts role-specific clarification questions
  • Requires human approval before anything is posted in Jira

The goal is not automation — it’s better conversations at the right time.


How I built it

Sprint Stoplight is built on Atlassian Forge using a Custom UI project page.

The backend computes deterministic sprint metrics directly from Jira data and applies team-configurable policy thresholds. When risk crosses defined limits, the system recommends creating a single “Standup Hub” issue to centralize discussion.

A Rovo agent is used to draft bounded, no-blame clarification questions — but humans stay firmly in control.


Challenges

The hardest part was balancing signal and noise.

It’s easy to generate alerts. It’s much harder to ensure:

  • Metrics are explainable
  • Policies actually affect outcomes
  • Humans stay in the loop
  • The UI feels operational, not AI-generated

This project taught me how real teams interact with tooling under pressure — and why restraint is often more powerful than automation.


What I learned

  • Sprint risk is a system-level problem, not an issue-level one
  • Good tooling should guide decisions, not replace them
  • Explainability and governance matter more than raw intelligence
Share this project:

Updates