Backlogs lie. Every issue claims to be urgent, and sprint planning becomes a gut-feel argument. Sprint Triage Agent turns triage into arithmetic: it reads each Jira issue — type, priority, story points, labels, linked Confluence pages, age — and scores it 0–100 with a visible breakdown (impact 30% / urgency 25% / effort 15% inverted / risk 15% / staleness 15%) and a plain verdict: DO NOW (≥70), NEXT (≥50), BACKLOG (≥30), or STALE. Every score comes with human-readable reasons, so the team can argue with the math instead of each other.
Ask it in Rovo chat — "Rank my sprint backlog" — and it returns the ranked order with scores and reasons, suggests labels and an assignee (learned from who actually finished similar Done issues), and applies labels and priority only after you confirm. The same scoring is exposed as MCP tools (score-issue, score-backlog, suggest-assignee, get-issue), so an external coding agent in Claude Code or Cursor can ask what to work on next and get the same ranked answer with reasons. Writes are Rovo-chat-only by design: apply-triage is never exposed via MCP, so nothing changes without a human saying yes.
Built with Forge (rovo:agent + rovo:mcp, Node.js, @forge/api). Demo data is a real seeded Jira project: 14 issues across bugs, stories, tasks and one epic, with priorities Highest–Lowest and story points 1–13. Local scoring on the seed data ranks TRIAGE-6 (73) and TRIAGE-1 (72) as DO NOW.