Claude-skill-registry architect-refinement

Software Architect technical refinement and DoR/DoD for [PROJECT_NAME]

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/architect-refinement" ~/.claude/skills/majiayu000-claude-skill-registry-architect-refinement && rm -rf "$T"
manifest: skills/data/architect-refinement/SKILL.md
source content

Architect Refinement Skill

Project: [PROJECT_NAME] Generated: [DATE] Purpose: Guide for technical refinement of user stories

When to Use

This skill guides you when doing technical refinement for:

  • New stories in
    /create-spec
  • Added stories in
    /add-story
  • Quick tasks in
    /add-todo
  • Bug stories in
    /add-bug

Quick Reference

Technical Refinement Process

  1. Understand Requirements: Read fachliche story (Feature, Acceptance Criteria)
  2. Analyze Architecture: What patterns apply?
  3. Determine WAS: Which components to create/modify
  4. Determine WIE: Architecture guidance and constraints
  5. Determine WO: Which files to touch
  6. Define DoD: Completion criteria
  7. Mark DoR: All checkboxes [x] when ready

Story is READY when

  • Fachliche requirements clear
  • Acceptance criteria specific and testable
  • Technical approach defined (WAS/WIE/WO)
  • Dependencies identified
  • Story appropriately sized (max 5 files, 400 LOC)
  • All betroffene Layer identified (Frontend/Backend/Database)
  • Integration points documented (if Full-stack)

Detailed Guidance

WAS (What to Build)

Components/Features to create or modify - NO CODE

Examples:

  • "User registration service"
  • "Profile edit form component"
  • "Email validation middleware"
  • "User database migration"

Anti-patterns:

❌ "Add function validateEmail()"  // Too specific
✅ "Email validation logic"

❌ "Create UserController with create(), update() methods"  // Implementation details
✅ "User API endpoints"

WIE (How/Architecture Guidance)

Patterns and constraints - NO IMPLEMENTATION

Examples:

  • "Use Service Object pattern"
  • "Follow existing UserController pattern"
  • "No direct DB calls from controllers"
  • "Requires rate limiting"
  • "Use caching for performance"

Anti-patterns:

❌ "def create_user(params)..."  // Code implementation
✅ "Use Service Object pattern for business logic"

❌ "Add try-catch in line 45"  // Too specific
✅ "Implement error handling with custom exceptions"

WO (Where/Files)

File paths to create or modify - NO CONTENT

Examples:

  • "app/services/users/register.rb"
  • "src/app/components/user-profile/"
  • "db/migrate/[timestamp]_add_email_to_users.rb"

Guidelines:

  • List all files that will be touched
  • Include new files to create
  • Include existing files to modify
  • Be specific with paths

Domain Reference (v3.0)

NEW in v3.0: Stories can reference domain areas

**Domain:** user-registration

This tells the main agent:

  • Which business process this affects
  • Which domain doc to keep updated
  • Business context to consider

Layer Analysis

Identify all affected layers:

LayerComponentsWhat Changes
FrontendProfileForm, Avatar componentNew edit form, image upload
BackendUserService, ProfileControllerUpdate endpoint, validation
Databaseusers tableAdd avatar_url column
DevOps-No changes

Integration Type:

  • Backend-only
  • Frontend-only
  • Full-stack

Critical Integration Points (if Full-stack):

  • Backend API Response → Frontend Component
  • Database Schema → Backend Query
  • Frontend Form → Backend Validation

DoD (Definition of Done)

Start all items unchecked [ ]

Standard DoD items:

#### Implementierung
- [ ] Code implemented and follows style guide
- [ ] Architecture patterns followed (WIE section)
- [ ] Security/performance requirements met

#### Qualitätssicherung
- [ ] All acceptance criteria satisfied
- [ ] Unit tests written and passing
- [ ] Integration tests written (if applicable)
- [ ] Linter passes

#### Dokumentation
- [ ] Code is self-documenting or has comments
- [ ] No debug code left in

Add story-specific items based on requirements.

Completion Check Commands

Provide bash commands to verify story completion:

# File existence
test -f src/app/components/profile/edit.component.ts && echo "Component OK"

# Content verification
grep -q "export class ProfileEditComponent" src/app/components/profile/edit.component.ts && echo "Class OK"

# Tests pass
npm test -- profile.component.spec.ts

# Linter
npm run lint

All commands must exit with code 0 for story to be DONE.


Architecture Patterns Reference

Common Patterns by Layer

Frontend:

  • Smart vs Dumb Components
  • Container/Presenter
  • State Management (Context, Store, Signals)
  • Form Validation

Backend:

  • Service Objects
  • Repository Pattern
  • Controller → Service → Repository
  • Result/Either Pattern

API:

  • RESTful endpoints
  • Request/Response DTOs
  • Error handling
  • Pagination

Database:

  • Migration for schema changes
  • Validations in model
  • Scopes for queries
  • Associations

Security Checklist

  • Input validation
  • SQL injection prevention
  • XSS prevention
  • CSRF protection
  • Authentication required?
  • Authorization rules defined?

Performance Checklist

  • N+1 queries avoided
  • Large datasets paginated
  • Caching considered
  • Eager loading for associations
  • Indexes on foreign keys

Story Sizing Guidelines

Maximum Limits

  • Files: Max 5 files per story
  • LOC: Max 400 lines of code
  • Complexity: Max S (Small, 1-3 SP)
  • Concerns: Single concern per story

When to Split

Split story if:

  • More than 5 files affected
  • Multiple unrelated changes
  • Can be done independently
  • Different developers could work on parts

Dependencies Analysis

Types of Dependencies

Hard Dependency (Sequential):

Story A → Story B
Example: "User model" before "User registration"

Soft Dependency (Optional):

Story A ⇢ Story B (better if A done first)
Example: "Basic search" before "Advanced filters"

No Dependency (Parallel):

Story A ∥ Story B
Example: "Profile edit" and "Avatar upload" can be parallel

How to Document

**Abhängigkeiten:** STORY-001, STORY-003

Or if none:

**Abhängigkeiten:** None

v3.0 Changes

What's Different

REMOVED in v3.0:

  • ❌ No "WER" (Agent) field
  • ❌ No "Relevante Skills" section
  • ❌ No skill-index.md lookup

ADDED in v3.0:

  • ✅ Optional "Domain" field
  • ✅ Skills auto-load via globs during implementation

Why:

  • Main agent implements directly (no delegation)
  • Skills activate based on files edited
  • Simpler story template
  • Domain tracking for business docs

Story Template Fields (v3.0)

**Type**: Backend | Frontend | Full-Stack | DevOps
**Domain:** [optional - e.g., user-registration]
**Abhängigkeiten:** [Story IDs or None]

### Technical Details

**WAS:** Components to create/modify
**WIE:** Architecture patterns and constraints
**WO:** File paths
**Domain:** [Optional domain area]
**Geschätzte Komplexität:** XS/S/M

Project-Specific Standards

Architecture Standards

<!-- From agent-os/product/architecture-decision.md or architecture/platform-architecture.md -->

[ARCHITECTURE_PATTERNS]

Project Structure

<!-- From agent-os/product/architecture-structure.md -->

[PROJECT_STRUCTURE]

Naming Conventions

[NAMING_CONVENTIONS]


Quality Standards Reference

For full DoD criteria see:

agent-os/team/dod.md
For full DoR criteria see:
agent-os/team/dor.md