Requirements Breakdown Prompts

Turn 50-point stories into 3-8 point stories in 60 seconds.

The Problem

Your user story takes 2-3 sprints instead of 1. Why?

It bundles 5+ features into 1 story.

The Solution

Copy-paste prompts for AI-assisted story breakdown. See docs/ for Agile fundamentals to use with AI.

Who This Is For

Primary audience: Product Owners without technical background

These prompts translate technical complexity into simple terms so you can create realistic stories without needing to understand code.

Can also be used by:

• Software Engineers (breaking down their own work)
• QA Engineers (understanding test scope)
• Technical Product Owners (with codebase knowledge)

Why use AI for story breakdown?

Team benefits:

• Save time: Break down stories in 60 seconds instead of hours
• Reduce effort: AI scans code and extracts context automatically
• Improve collaboration: Clear stories = engineers know what to build, QA knows what to test, PO tracks progress

End user benefits:

• Faster delivery: Small stories ship in days, not weeks
• Better quality: Each story is independently testable
• More predictable: Teams deliver on time with realistic estimates

Two Workflows

Basic Flow (No Codebase Access)

Quick breakdown without technical context:

1. Open the Break Down Large Story prompt
2. Copy → Paste into ChatGPT/Claude with your story
3. Done. You have 3-8 point stories with acceptance criteria

Use when: Business-only features, UI changes, or no codebase access.

Ideal Flow (With Codebase Scanning)

For Product Owners with codebase access:

Collaborate with Claude Code in terminal to scan your codebase first:

1. Open terminal in your codebase with Claude Code
2. Use Scan Codebase for Context to get non-technical overview
   • AI scans actual code files (no guessing!)
   • Extracts key components, services, dependencies
   • Translates technical details into simple terms
   • Outputs actionable breakdown points
3. Copy context from AI's findings
4. Use Break Down Large Story with technical context
5. Get realistic breakdown based on real architecture

Why codebase scanning matters:

• ✅ Eliminates AI hallucination - AI reads actual code, doesn't guess
• ✅ Discovers hidden dependencies (3rd party APIs, shared services)
• ✅ Accurate complexity from actual code, not assumptions
• ✅ Engineers trust estimates (based on code they wrote)
• ✅ Fewer "we didn't know it touches X" surprises mid-sprint

Use when: Technical features, service integrations, database changes.

Real Example

Before: 1 story, 50 points, took 6 weeks

After: 10 stories, 48 points total, delivered in 6 weeks (on time)

See the complete breakdown →

Available Prompts

• Break Down Large Story → Story is too large or bundles multiple features
• Rewrite Correct Format → Story is unclear
• Scan Codebase for Context → Understand existing implementation (use with Claude Code)

Contributing

Have a recurring problem that needs a prompt solution? We want to hear about it!

Suggest a new prompt:

1. File an issue using the "Prompt Idea" template
2. Describe your problem and how often you face it
3. If it's recurring (not a one-time issue), we'll create a prompt

Improve existing prompts: Submit a Pull Request with before/after examples

See CONTRIBUTING.md for full guidelines.

Why frequency matters? One-time problems don't need prompts. We focus on recurring pain points.

License

MIT