I created a prompt to be rerun at the end of each sprint but the results are inaccurate. How can I i

Question

I created a prompt to reuse at the end of each sprint to calculate epic movement, but the results always come back inaccurate or Rovo dies silently and never generates an error. I had tried to create an agent but that failed. What can I do to better improve the results from either my prompt or even get my agent working?

Dr Valeri Colon (Connect Centric) · Answer

@Escobar_ Andy welcome to the community. I’d separate the calculation from the AI. Rovo can vary on large analytical requests, and Atlassian tracks silent failures for bulky agent workloads under ROVO-615. Use Jira Automation/JQL to calculate the sprint metrics, then pass the verified numbers to Rovo for interpretation and narrative. Just note that Automation’s Lookup work items action currently returns up to 100 items, so larger datasets need to be split.

Sami Shaik · Answer

Hello @Escobar_ Andy This isn't a prompt problem, it's a known limitation. Rovo doesn't see all your issues. It samples a partial set (around 50 in agent contexts), so any calculation across a sprint will be wrong and different each run. The silent failure is a tracked Atlassian bug (ROVO-615): big requests hit an internal cap and die with no error. The fix: don't let Rovo do the math. Let Automation count, let Rovo write. * Jira Automation rule with trigger: Sprint completed * Lookup Issues action with JQL like project = XYZ AND sprint = {{sprint.id}} * Get exact numbers with smart values: {{lookupIssues.size}}, {{lookupIssues.Story Points.sum}} (run separate lookups for done vs. rolled-over) * Pass those computed numbers to your Rovo agent in the same rule. Its only job is writing the summary, not calculating it. Rovo can't get the totals wrong if it never produces them. Why your agent failed: agents can't run on their own. They must be invoked from an Automation rule (agent config, then Triggers). If you gave it JQL to search itself, you hit the ~50-issue ceiling.