What problem does it solve? Open issue backlogs accumulate duplicates, subsets, and small related edits filed weeks apart on the same skill's eval snapshot, and each one pays its own expensive evaluation run. This Skill finds those piles by computing per-slot queue depth and merges N-way where the issues are genuinely one piece of work, buying back evaluation runs. ## Core Features & Use Cases - Queue computation: Runs slots.py against GitHub project board, open issue, and open PR data to rank eval slots by queue depth and flag coverage gaps. - Four merge verdicts: Classifies groups as duplicate, absorb (strict subset), batch (one-sitting edits), or split-the-lanes, with evidence gathered by opening the cited files. - Approved application only: Proposes partitions first, then folds content, retitles targets, moves assignees and labels, and closes losers with reasoning after lead approval. - Use Case: A repo lead asks to "merge the backlog" before promoting work; the Skill reads every deep queue, proposes a partition of fifteen issues into four groups, and reports how many evaluation runs the merges buy back. ## Quick Start Ask the assistant to merge the backlog by computing the eval-slot queues and proposing which open issues should be combined.
