board-cleanup

name	board-cleanup
description	Scans Jira for stale issues, duplicates, and missing fields, then fixes what the user approves. Use when the user says the board is a mess, asks for a backlog grooming pass, wants to clean up Jira, or asks about stale/duplicate tickets before sprint planning.
allowed-tools	Read, WebFetch, Bash

Board Cleanup

Run a Jira hygiene pass. Act immediately — the user asked for cleanup, so skip the intent-detection step.

Parse arguments:

• --scope (default: all) — stale for stale-issue audit only, duplicates for dedup only, fields for missing-field audit only, all for everything

Step 1 — Scan for stale issues

If scope includes stale:

1. Pull all open issues (not Done, not Won't Do) across the team's project(s).
2. Flag issues with no update (comment, status change, field edit) in the last 10 business days.
3. For each stale issue: show key, title, status, assignee, last update date, and any linked issues.
4. Propose actions per issue: ping assignee for a status check, move to backlog, close as won't-do, or update due date.

Present the full stale-issue list before making any changes.

Step 2 — Scan for likely duplicates

If scope includes duplicates:

1. Search open issues for likely duplicates: similar titles, same component/label, opened within a few days of each other, or explicitly cross-referenced.
2. For each duplicate set: show both issues side-by-side — key, title, status, assignee, description summary.
3. Propose which issue to keep and which to close as a duplicate (linking to the kept issue).

Present all duplicate sets before closing anything.

Step 3 — Scan for missing required fields

1. Check all open issues for missing fields: story points/estimate, priority, assignee, sprint assignment (for issues that should be in the current sprint).
2. Present a table of issues with missing fields and what's missing.

Step 4 — Apply approved fixes

1. Walk through each finding from Steps 1–3.
2. Apply only the changes the user explicitly approves.
3. Report each change as it's made with a Jira link.

Connector failures

If Jira/Atlassian is unreachable, stop — this workflow requires Jira as the data source. Tell the user: "Jira isn't connected. Connect it, then rerun board cleanup."

Approval gates

• Never delete issues. Jira doesn't make this easy to undo, and neither does this skill. If the user asks, say the skill can't and point them to Jira directly.
• Never close an issue (Done or Won't Do) without explicit approval, even when staleness or duplication looks obvious. Flag and defer.
• Never auto-link or auto-close duplicates. Show side-by-side and wait for approval per pair.
• Side-by-side diffs for all changes. Show current value and proposed value; wait for approval per item.

Output

End with a summary: X issues updated, Y duplicates closed/linked, Z fields filled. Include Jira links to the affected issues.