scrum-event-backlog-refinement

shishi's avatarfrom shishi

Transform PBIs into ready status for AI execution. Use when refining backlog items, writing acceptance criteria, splitting stories, or ensuring Definition of Ready.

⭐3stars🔀1forks📁View on GitHub🕐Updated Jan 1, 2026

When & Why to Use This Skill

This Claude skill automates the Scrum backlog refinement process by transforming Product Backlog Items (PBIs) into an 'AI-Ready' state. It ensures that user stories meet the INVEST principles and include executable verification commands, allowing AI agents to execute tasks autonomously without human intervention. By bridging the gap between human requirements and AI execution, it streamlines the software development lifecycle and enhances sprint predictability.

Use Cases

  • Refining vague or high-level user stories into detailed, actionable PBIs that meet the 'Definition of Ready' for AI-driven development.
  • Splitting large epics or complex stories into smaller, independent units of work to improve flow and satisfy INVEST principles.
  • Generating executable verification commands and acceptance criteria to ensure that completed work can be automatically validated.
  • Auditing the backlog for anti-patterns and technical dependencies to prevent blockers before a sprint begins.
  • Maintaining a synchronized Scrum dashboard within a codebase to serve as the single source of truth for project progress.
namescrum-event-backlog-refinement
descriptionTransform PBIs into ready status for AI execution. Use when refining backlog items, writing acceptance criteria, splitting stories, or ensuring Definition of Ready.

You are an AI Backlog Refinement facilitator transforming PBIs into ready status where AI agents can execute them autonomously.

Single Source of Truth: scrum.ts in project root. Use scrum-dashboard skill for maintenance.

  • Splitting & Merging: See splitting.md for when to split large PBIs AND when to merge small ones back together
  • Anti-Patterns: See anti-patterns.md for common PBI mistakes to avoid

AI-Agentic Definition of Ready

Ready = AI can complete it without asking humans.

A PBI is ready when:

  1. AI can complete without human input
  2. User Story format (role, capability, benefit)
  3. Acceptance criteria have executable verification commands
  4. Dependencies are resolved
  5. INVEST principles are satisfied

INVEST Principles (AI-Agentic)

Principle AI-Agentic Interpretation
Independent Can reprioritize freely, AND no human dependencies
Negotiable Clear outcome, flexible implementation
Valuable User Story format makes value explicit
Estimable All information needed is available
Small Smallest unit delivering user value
Testable Has executable verification commands

Refinement Process

  1. Autonomous Refinement First - Explore codebase, propose acceptance criteria, identify dependencies
  2. If AI Can Fill All Gaps - Update status to ready
  3. If Story Is Too Big - Split into smaller stories (see splitting.md)
  4. If Story Lacks Value Alone - Merge with adjacent PBI (see splitting.md Anti-Patterns)
  5. If Needs Human Help - Keep as refining, document questions

Adaptation Check for Ready PBIs

Even ready PBIs must be inspected before Sprint Planning:

Check Question If Failed
Goal Alignment Still aligned with Product Goal? Re-evaluate priority
Codebase Changes Recent commits invalidate assumptions? Update acceptance criteria
Retrospective Insights Related improvements identified? Incorporate learnings
Verification Commands Still executable and meaningful? Update commands
Dependencies New blockers emerged? Document and resolve

If any check fails: Change status back to refining and address gaps.

Ron Jeffries' 3C Principle

  • Card: Story on card with estimates (intentionally brief)
  • Conversation: Details drawn out through PO discussion
  • Confirmation: Acceptance tests confirm correct implementation

Backlog Granularity

┌─────────────────┐
│  FINE-GRAINED   │  ← Ready for upcoming sprints (1-5 points)
├─────────────────┤
│    MEDIUM       │  ← Next 2-3 sprints, may need splitting
├─────────────────┤
│ COARSE-GRAINED  │  ← Future items, Just-in-Time refinement
└─────────────────┘

When items move up in priority, split to sprint-sized pieces. Don't refine everything upfront.

Collaboration

  • @scrum-team-product-owner: Product Goal alignment, value prioritization
  • @scrum-team-developer: Technical feasibility, effort estimation
  • @scrum-team-scrum-master: Definition of Ready enforcement