Skip to content

hoavdc/CodexKit / skills/codexkit-backlog-refiner

codexkit-backlog-refiner

Verified

Refine Product Backlog Items to meet INVEST criteria. Write User Stories with Acceptance Criteria in Given/When/Then format, estimate with Story Points, and flag dependencies. Use before sprint planning when backlog items need grooming. Do not use to prioritize the backlog — that is the Product Owner's decision.

FileKindSize
SKILL.mdskill md4.2 KB
agents/openai.yamlconfig171 B
---
name: codexkit-backlog-refiner
description: Refine Product Backlog Items to meet INVEST criteria. Write User Stories with Acceptance Criteria in Given/When/Then format, estimate with Story Points, and flag dependencies. Use before sprint planning when backlog items need grooming. Do not use to prioritize the backlog — that is the Product Owner's decision.
version: 1.0.0
category: review
---

# Backlog Refiner

## Purpose

Transform raw backlog items into sprint-ready User Stories that meet INVEST criteria, with clear Acceptance Criteria and reliable estimates.

## When to use

- mid-sprint refinement sessions (preparing items for the next sprint)
- when backlog items are vague, oversized, or lack acceptance criteria
- when the team needs help decomposing epics into estimable stories
- when onboarding a new Product Owner who needs story-writing structure

## When not to use

- prioritizing backlog order (Product Owner responsibility)
- strategic roadmap discussions or feature discovery
- items already refined and ready for sprint planning

## Inputs

- raw backlog items (titles, rough descriptions, business context)
- product goal or current product increment theme
- team's story point calibration reference (what does 1 point mean?)
- known technical constraints or dependencies
- Definition of Done for the product

## Procedure

1. **Review each item** against INVEST checklist:
   - I — Independent: not blocked by other stories
   - N — Negotiable: details can be discussed with PO
   - V — Valuable: delivers clear value to user or business
   - E — Estimable: team can estimate the effort
   - S — Small: completable within one sprint
   - T — Testable: has verifiable acceptance criteria
2. **Write User Story** using 3C template:
   - Card: "As a [persona], I want [action] so that [benefit]"
   - Conversation: context, questions for clarification
   - Confirmation: Acceptance Criteria in Given/When/Then
3. **Estimate effort** using Planning Poker reference scale:
   - 1 pt ≈ 4h | 2 pt ≈ 1 day | 3 pt ≈ 1.5 days | 5 pt ≈ 3 days
   - 8 pt = needs breakdown | 13+ pt = Epic, must split
4. **Flag items** that are too large — recommend split strategies.
5. **Identify dependencies** on other teams, services, or external parties.
6. **Document risks** or unknowns that could block the story.

## Output

- refined User Stories with INVEST checklist scores
- Acceptance Criteria (3–5 per story, Given/When/Then format)
- Story Point estimates with rationale
- split recommendations for oversized items
- dependency and risk flags

## Definition of done

- every story passes all 6 INVEST criteria
- every story has at least 3 Acceptance Criteria in Given/When/Then
- no story exceeds 8 story points without a split recommendation
- dependencies are explicitly documented

## Examples

- "Refine these 8 raw feature requests into sprint-ready user stories."
- "This epic 'User Authentication' is too big — help me split it into stories."
- "Write acceptance criteria for the 'Password Reset' story."

## Quality Criteria

- [ ] Feedback is specific and references exact locations in the reviewed material
- [ ] Each critique includes a concrete improvement suggestion
- [ ] Severity is categorized (critical / important / nice-to-have)
- [ ] Positive aspects are acknowledged alongside areas for improvement

## Verification (4C)

| Check | Question |
|-------|----------|
| **Correctness** | Is the feedback technically accurate and properly contextualized? |
| **Completeness** | Were all major sections of the reviewed material addressed? |
| **Context-fit** | Is the review granularity appropriate for the material's maturity level? |
| **Consequence** | If the author implemented all feedback literally, what could go wrong? |

## Edge Cases

- **Material is too early-stage for detailed review** — Provide structural feedback only. Note that content review is deferred until it matures.
- **Reviewer lacks domain expertise** — Focus on structure, clarity, and consistency. Flag domain-specific claims as 'Needs SME verification'.
- **Author is defensive or resistant to feedback** — Lead with what works well. Frame changes as questions rather than mandates.

## Changelog

- v1.0.0 — Initial release
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall