ticket-duplicates

Ticket Duplicates

Duplication hides in two places: another ticket (open or long resolved), and history that never got a ticket — work already done, or an idea already discussed and rejected. This skill searches both, then resolves every candidate to an evidence-backed verdict. A resolved duplicate is a win twice over: the work may already be done, and its history says what it cost.

Underneath, memory-search owns the search ladder and the relevance gate, graph-traversal owns relationship walking when a candidate needs one more hop, and memory-capture owns any durable write; all three ship with the gutt-pro plugin (this plugin depends on it) — without them, follow the rules below and note the gap in one line. Jira access comes from whatever Atlassian tooling the session surfaces; find it in your tool list — names and prefixes vary per install.

Hard rules (non-negotiable — read first)

1. Jira is read-only here — recommend, never act. No linking, closing, labeling, or field edits; those are the human's calls. The only write this skill may produce is one comment, drafted and posted only after the user approves the exact text in this session. Treat it as final — approval is the gate, not an undo; a truly needed correction re-approves the text and targets the existing comment if the tool allows, rather than posting a second one. Write markdown and set the tool's content-format parameter to markdown when it exposes one.
2. Every candidate gets resolved. Each surfaced candidate ends as a verdict or an explicit "not assessed (reason)" — never silently dropped. An unresolved candidate is exactly the false reassurance this skill exists to prevent.
3. Fixed verdict vocabulary. duplicate (same outcome sought), overlaps (shares part of the scope), related (useful context, different outcome), distinct — each with confidence high/medium/low and one line of evidence (a quote or fact, with key or id and date). No evidence means related at best.
4. "Nothing found" carries its scope. A clean result states what was searched — the JQL angles and the memory phrasings — so absence is a checkable claim, not a shrug.
5. Org scope is enforced at the output, not the request. Pass explicit group_ids naming the org group on reads — take the name from session results or ask; never guess one. And treat scope as server-decided: before any candidate or verdict enters the offered comment, confirm that item's own scope is the org group; personal-scope material stays in the in-session report, marked as personal.
6. Bare tool names, probed with ToolSearch before concluding one is missing; the mcp__…__ prefix varies per install.
7. No memory writes. A duplication pattern worth keeping goes through memory-capture and its trust-tier gate.