What problem does it solve? Engineering backlogs accumulate stale, contradictory, or already-solved issues, making it hard to decide what to work on next. This Skill reads an entire slice of the backlog (a milestone, label, or everything open), verifies each issue against the actual code, and produces a prioritized, dependency-aware plan instead of a raw issue list. ## Core Features & Use Cases - Whole-slice reading: Fetches every open issue in a milestone or label via wide and deep tiers, fanning out to parallel reader subagents when the slice is too large for one session. - Code-verified verdicts: Classifies each issue as current, needs rewrite, stale, or unverifiable based on the repository state, pinned to a base commit and tracker timestamp. - Tiered plan with dependency graph: Builds the dependency graph and critical path, tiers issues by importance (blocking, firing today, catching regressions), and proposes candidate parallel batches plus a hygiene plan. - Use Case: Before splitting an epic into parallel work waves, ask for a survey of the milestone to learn which issues are still real, which block others, and which batch to start with. ## Quick Start Survey all open issues in the current milestone, verify each against the code, and tell me what to take next.
