Our Scrum Master wants to know if we can point OpenHands at our Jira "Bug" column and let it go to town. Can it autonomously triage, reproduce, and fix tickets, or is it too risky for an Agile sprint?
3 answers
We actually run a "Nightly Crew" of OpenHands agents. Every night at 2 AM, the agent pulls the three most detailed bug reports from our Jira. It attempts to reproduce the bug by writing a new test case. If it can reproduce it, it tries a fix. When the devs log in at 9 AM, they don't see a "fixed" bug; they see a Pull Request with a passing test and a suggested fix. This fits perfectly into the Agile and Scrum workflow because the human developer still does the "Code Review" during the morning standup. It’s increased our sprint velocity by roughly 25% because we aren't wasting the first two hours of every day on triage.
Does the agent understand "vague" bug reports like "The button doesn't work," or does it need perfectly documented steps to reproduce?
We use it to "Draft" our Sprint Retrospectives by summarizing all the code changes and blockers it encountered.
That's a clever use case, Angela. It turns the AI from just a "coder" into a true "team member."
It struggles with vague reports just like a human does! However, it’s great at "Web Research." If a report is vague, the agent will often browse the app's frontend using its built-in browser tool to try and find the error in the console logs. If it still can't find it, it posts a comment on the Jira ticket asking for more details. It’s a very proactive way to handle Quality Management without bugging the human team for every minor issue.