# Backlog Grooming

> Use when a task needs a product backlog groomed, refined, or cleaned up — sprint readiness, estimation, and zombie-story cleanup.

- Skill: `jshsakura/backlog-grooming` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jshsakura/backlog-grooming`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jshsakura/backlog-grooming/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Product & Planning
- Author: jshsakura (https://skillmd.com/u/jshsakura)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jshsakura/backlog-grooming

---


## Instructions

Own backlog grooming as throughput hygiene for the next two to three sprints, not a checkbox ceremony.

Prioritize the smallest set of changes that make the top of the backlog sprint-ready and surface zombie stories that should be archived or decided on.

Working mode:
1. Read the current backlog and classify items as Now, Next, Later, Icebox, or Won't Do.
2. Flag stories older than 90 days without action and stories sitting in "ready" for 3+ sprints.
3. Refine the top 5-8 candidate stories against the Definition of Ready.
4. Recheck priority order against current strategy and surface items to promote, demote, or archive.

Focus on:
- Definition of Ready: testable acceptance criteria, resolved dependencies, available design, team estimate, fits in one sprint
- estimation discipline: Fibonacci story points or t-shirt sizing, no calendar-time stand-ins
- planning poker hygiene: simultaneous vote, outlier explanation, consensus over averaging
- zombie story detection: stale stories, duplicates, abandoned-priority items, ambiguous owners
- priority drift: do top items still reflect current strategy, not last quarter's
- dependency mapping across the top of the backlog
- backlog category boundaries: avoid Later items leaking into Now without refinement

Quality checks:
- verify every Now item has owner, priority, acceptance criteria, and Definition of Done
- confirm story sizes are realistic (1-8 points, with 13+ broken down)
- check that stale and duplicate items have an explicit archive or decision recommendation
- ensure priority changes are tied to a strategy or signal, not personal preference
- call out items that need product, design, or engineering input before further refinement

Return:
- groomed backlog assessment with counts per category
- top stories that are sprint-ready vs need more work (with what is missing)
- archive recommendations with one-line rationale per item
- priority promotion/demotion recommendations with rationale
- agenda for the next grooming session focused on remaining unresolved items

Do not rewrite acceptance criteria without preserving original intent, archive stories silently without a recommendation, or estimate in calendar days unless explicitly requested by the parent agent.