Brownfield Onboarder

Brownfield Onboarder - Intelligent CLAUDE.md Merger

Purpose: Intelligently merge existing CLAUDE.md backups into SpecWeave's structure without bloating the main CLAUDE.md file.

When to Use: After installing SpecWeave into an existing project that had CLAUDE.md

Philosophy: Keep CLAUDE.md as a concise guide, distribute detailed content to appropriate SpecWeave folders.

Modes: Supports both Quick Start (incremental) and Comprehensive (upfront) approaches 🆕

Two-Mode Support 🆕

The brownfield-onboarder works differently based on the chosen documentation path:

Quick Start Mode (Incremental)

Philosophy: Merge only essential context, defer detailed docs to per-increment

What to merge immediately:

• ✅ Core architecture overview (high-level)
• ✅ Tech stack and infrastructure
• ✅ Critical patterns (auth, payment, security)
• ✅ Team conventions and workflows
• ✅ Project summary and domain context

What to defer (document per increment):

• ⏸️ Detailed business rules (extract when modifying that code)
• ⏸️ Module-specific documentation (extract when working on that module)
• ⏸️ API-level documentation (extract when touching those APIs)
• ⏸️ Code examples (extract as needed)

Result: Minimal upfront merge (30-60 minutes), detailed docs grow incrementally

Comprehensive Mode (Upfront)

Philosophy: Merge everything upfront for complete context

What to merge:

• ✅ All architecture documentation
• ✅ All business rules
• ✅ All module-specific docs
• ✅ All API documentation
• ✅ All conventions and patterns
• ✅ All code examples

Result: Complete merge (1-3 hours), full context available immediately

Mode Selection

Auto-detection:

// Settings auto-detected
const mode = config.brownfield?.mode || 'auto';

if (mode === 'auto') {
  // Use complexity from brownfield-analyzer
  const complexity = await readComplexityAssessment();
  mode = complexity.recommendedPath === 'Quick Start' ? 'incremental' : 'comprehensive';
}

User can override:

# Force Quick Start mode
brownfield-onboarder --mode quick-start

# Force Comprehensive mode
brownfield-onboarder --mode comprehensive

The Problem

When installing SpecWeave into an existing project:

1. Project already has CLAUDE.md with valuable project-specific context
2. SpecWeave installs its own CLAUDE.md as the development guide
3. Old CLAUDE.md is backed up to .claude/backups/CLAUDE-backup-{timestamp}.md
4. Need to intelligently merge project-specific content WITHOUT bloating SpecWeave's CLAUDE.md

The Solution: Smart Distribution

Instead of bloating CLAUDE.md, distribute content to appropriate folders:

Project-specific content → SpecWeave folders:

# Internal Documentation (strategic, team-only)
Architecture details    → .specweave/docs/internal/architecture/existing-system.md
Technology stack        → .specweave/docs/internal/architecture/tech-stack.md
Business rules          → .specweave/docs/internal/strategy/business-rules.md
Team workflows          → .specweave/docs/internal/processes/team-workflows.md
Deployment process      → .specweave/docs/internal/processes/deployment.md
Domain knowledge        → .specweave/increments/{####-name}/docs/domain/{domain}.md

# Public Documentation (user-facing, can be published)
Project conventions     → .specweave/docs/public/guides/project-conventions.md
API conventions         → .specweave/docs/public/guides/api-conventions.md
Code style              → .specweave/docs/public/guides/code-style.md

Only add to CLAUDE.md: High-level project summary (1-2 paragraphs max)

Activation

Trigger: User runs specweave merge-docs or asks "merge my old CLAUDE.md"

Auto-detection:

1. Check if .claude/backups/CLAUDE-backup-*.md exists
2. If multiple backups, use most recent
3. If no backups, inform user and exit gracefully

Analysis Process

Step 1: Parse Backup CLAUDE.md

Extract sections:

interface ParsedCLAUDEmd {
  projectName: string;
  projectDescription: string;
  techStack: TechStack;
  architecture: ArchitectureSection[];
  conventions: Convention[];
  workflows: Workflow[];
  domainKnowledge: DomainSection[];
  teamGuidelines: TeamGuideline[];
  deploymentProcess: DeploymentSection[];
  apiDesign: APISection[];
  businessRules: BusinessRule[];
  codeExamples: CodeExample[];
  customInstructions: Instruction[];
}

Section Detection Keywords:

• Tech Stack: "technology", "framework", "database", "infrastructure", "stack", "tools"
• Architecture: "architecture", "system design", "components", "services", "microservices"
• Conventions: "naming convention", "code style", "pattern", "standard", "guideline"
• Workflows: "workflow", "process", "pipeline", "deployment flow", "release process"
• Domain: Domain-specific terms (e.g., "patient", "booking", "payment", "order")
• Business Rules: "business rule", "validation", "policy", "constraint", "requirement"
• API Design: "API", "endpoint", "REST", "GraphQL", "authentication", "authorization"
• Deployment: "deploy", "CI/CD", "environment", "production", "staging"

Step 2: Classify Content (Mode-Aware) 🆕

For each section, determine:

1. Generic or Project-Specific?

   • Generic: Common programming advice, general best practices
   • Project-specific: Domain knowledge, team conventions, project architecture

2. Overlap with SpecWeave CLAUDE.md?

   • Compare section with SpecWeave's CLAUDE.md
   • If >80% similar, skip (already covered)
   • If <80% similar, extract unique content

3. Essential or Detailed? 🆕

   • Essential: Core architecture, critical patterns, tech stack, team workflows
   • Detailed: Module-specific rules, detailed APIs, code examples

4. Mode-Based Decision 🆕

   • Quick Start Mode: Merge essential only, defer detailed
   • Comprehensive Mode: Merge everything

5. Target Destination

   • Determine best SpecWeave folder for this content
   • See "Content Distribution Rules" below

Classification Matrix 🆕:

Example (Quick Start):

Analyzing CLAUDE.md backup (Quick Start mode)...

Found sections:
  ✅ Core Architecture (merge now)
  ✅ Tech Stack (merge now)
  ✅ Auth Pattern (merge now - critical)
  ⏸️ Payment Business Rules (defer - extract when working on payments)
  ⏸️ User Module API (defer - extract when modifying user code)
  ⏸️ Code Examples (defer - extract as needed)

Merging 3 sections immediately, deferring 3 for incremental extraction.

Step 3: Content Distribution Rules

Rule 1: Domain Knowledge → Specifications

Indicators: Business concepts, entities, domain terminology

Example:

# Old CLAUDE.md
## Domain Model

Our platform manages **patient appointments** with **healthcare providers**.
Key entities:
- Patient (demographics, insurance, medical history)
- Provider (specialties, availability, credentials)
- Appointment (time slot, status, notes)
- Clinic (location, services, staff)

Business rules:
- Appointments must be 15-60 minutes
- Patients can cancel up to 24 hours before
- Providers can override cancellation policy

Destination: specifications/modules/appointments/domain-model.md

CLAUDE.md addition: None (link from CLAUDE.md to specifications)

Rule 2: Architecture → .specweave/docs/architecture/

Indicators: System design, component descriptions, data flow

Example:

# Old CLAUDE.md
## System Architecture

We use a microservices architecture:
- API Gateway (Kong) - routing, authentication
- Booking Service (Node.js) - appointment management
- Notification Service (Python) - email/SMS
- Payment Service (Node.js) - Stripe integration
- Database (PostgreSQL) - shared across services

Destination: .specweave/docs/internal/architecture/existing-system.md

CLAUDE.md addition:

## Project-Specific Architecture

See [Existing System Architecture](.specweave/docs/internal/architecture/existing-system.md) for complete microservices architecture.

Rule 3: Conventions → .specweave/docs/guides/

Indicators: Naming conventions, code style, patterns

Example:

# Old CLAUDE.md
## Naming Conventions

- API endpoints: `/api/v1/{resource}/{action}` (kebab-case)
- Database tables: `{domain}_{entity}` (snake_case)
- TypeScript interfaces: `I{Name}` prefix (PascalCase)
- React components: `{Name}Component.tsx` suffix

Destination: .specweave/docs/public/guides/project-conventions.md

CLAUDE.md addition: None (standard conventions, no need to clutter CLAUDE.md)

Rule 4: Workflows → .specweave/docs/guides/

Indicators: Deployment process, CI/CD, release workflow

Example:

# Old CLAUDE.md
## Deployment Process

1. Create feature branch from `main`
2. Implement feature with tests
3. Create PR (requires 2 approvals)
4. Merge → auto-deploy to staging
5. Manual approval → deploy to production
6. Rollback via GitHub Actions if needed

Destination: .specweave/docs/internal/processes/deployment.md

CLAUDE.md addition:

## Deployment

See [Deployment Guide](.specweave/docs/internal/processes/deployment.md).

Rule 5: Business Rules → specifications/modules/

Indicators: Validation rules, policies, constraints

Example:

# Old CLAUDE.md
## Business Rules

### Appointment Booking
- Patients can book up to 3 months in advance
- Maximum 5 active appointments per patient
- Same-day appointments require $50 deposit
- Insurance verification required before booking

Destination: .specweave/docs/internal/strategy/appointments/business-rules.md

CLAUDE.md addition: None (specifications are source of truth)

Rule 6: Tech Stack → .specweave/docs/architecture/

Indicators: Technologies, frameworks, tools

Example:

# Old CLAUDE.md
## Tech Stack

- Frontend: Next.js 14, React, Tailwind CSS
- Backend: Node.js 20, Express, TypeScript
- Database: PostgreSQL 16, Prisma ORM
- Cache: Redis
- Queue: BullMQ
- Infrastructure: Hetzner Cloud, Terraform
- Monitoring: Grafana, Prometheus

Destination: .specweave/docs/internal/architecture/tech-stack.md

CLAUDE.md addition:

## Tech Stack

Next.js 14 + Node.js 20 + PostgreSQL 16 + Hetzner Cloud

See [Tech Stack Details](.specweave/docs/internal/architecture/tech-stack.md).

Rule 7: API Design → .specweave/docs/guides/

Indicators: API conventions, authentication, error handling

Example:

# Old CLAUDE.md
## API Design

All APIs follow REST conventions:
- Authentication: JWT in Authorization header
- Errors: Standard structure { error, message, details }
- Pagination: page, limit query params
- Filtering: field[operator]=value
- Versioning: /api/v1, /api/v2

Destination: .specweave/docs/public/guides/api-conventions.md

CLAUDE.md addition: None (guide covers it)

Rule 8: Code Examples → Discard or Minimal

Indicators: Code snippets, example implementations

Decision:

• If generic (standard pattern): Discard (SpecWeave CLAUDE.md already has examples)
• If project-specific (custom pattern): Extract to guide

Example:

# Old CLAUDE.md - Generic React pattern
function UserList() {
  const [users, setUsers] = useState([]);
  // ... standard React code
}

Action: Discard (generic React, not project-specific)

Example:

# Old CLAUDE.md - Custom authentication pattern
// Our custom auth hook (wraps Supabase)
function useCustomAuth() {
  const { session } = useSupabase();
  const { roles } = useRoleProvider();
  return { user: session?.user, hasRole: (role) => roles.includes(role) };
}

Action: Extract to .specweave/docs/public/guides/authentication.md (project-specific pattern)

Step 4: Update CLAUDE.md (Minimal)

ONLY add high-level project summary to SpecWeave's CLAUDE.md:

---

## Project-Specific Context

**Project**: Healthcare Appointment Booking Platform
**Domain**: Healthcare, Patient Management, Provider Scheduling

### Quick Links
- [Domain Model](.specweave/increments/####-name/docs/domain/appointments/domain-model.md)
- [Existing System Architecture](.specweave/docs/internal/architecture/existing-system.md)
- [Tech Stack](.specweave/docs/internal/architecture/tech-stack.md)
- [Business Rules](.specweave/docs/internal/strategy/appointments/business-rules.md)
- [Deployment Guide](.specweave/docs/internal/processes/deployment.md)
- [Project Conventions](.specweave/docs/public/guides/project-conventions.md)

**Note**: All project-specific details are in linked documents. This keeps CLAUDE.md concise.

---

Total addition: ~15 lines max

Intelligence Rules

Avoid Bloat

Never add to CLAUDE.md:

• Generic programming advice (SpecWeave CLAUDE.md already has it)
• Detailed code examples (put in guides)
• Long architecture descriptions (put in architecture docs)
• Business rule details (put in specifications)
• API documentation (put in guides)

Only add to CLAUDE.md:

• 1-2 sentence project description
• Domain/industry context
• Links to detailed docs

Avoid Duplicates

Before creating files, check if similar content exists:

// Check if domain model already exists
if (exists("specifications/modules/appointments/domain-model.md")) {
  // Compare content
  existingContent = read("specifications/modules/appointments/domain-model.md");
  newContent = extractDomainModel(backupCLAUDEmd);

  if (similarity(existingContent, newContent) > 0.8) {
    // Skip, already documented
    skip();
  } else {
    // Merge unique content
    mergedContent = merge(existingContent, newContent);
    write("specifications/modules/appointments/domain-model.md", mergedContent);
  }
}

Preserve Accuracy

When extracting content:

• Don't paraphrase technical details
• Preserve exact terminology
• Keep code examples verbatim
• Maintain formatting (tables, lists, code blocks)

User Confirmation

Before writing files, show user:

I found the following project-specific content in your backup CLAUDE.md:

📦 Domain Model (Healthcare Appointments)
   → .specweave/increments/####-name/docs/domain/appointments/domain-model.md

🏗️ Microservices Architecture
   → .specweave/docs/internal/architecture/existing-system.md

🛠️ Tech Stack (Next.js + Node.js + PostgreSQL)
   → .specweave/docs/internal/architecture/tech-stack.md

📋 Business Rules (Booking policies)
   → .specweave/docs/internal/strategy/appointments/business-rules.md

🔧 Project Conventions (Naming, code style)
   → .specweave/docs/public/guides/project-conventions.md

🚀 Deployment Process (CI/CD workflow)
   → .specweave/docs/internal/processes/deployment.md

📝 CLAUDE.md Update
   → Add 12-line project summary with links

Total files to create: 6
Total lines added to CLAUDE.md: 12

Proceed with merge? (y/n)

Output: Merge Report

After merge, generate mode-specific report:

Quick Start Mode Report 🆕

# CLAUDE.md Merge Report - Quick Start Mode

**Date**: 2025-10-26
**Backup File**: .claude/backups/CLAUDE-backup-20251026-143022.md
**Merge Status**: ✅ Complete (Essential content only)
**Mode**: Quick Start (Incremental Documentation)

---

## Files Created (Essential Only)

1. ✅ `.specweave/docs/internal/architecture/core-architecture.md` (120 lines)
2. ✅ `.specweave/docs/internal/architecture/tech-stack.md` (80 lines)
3. ✅ `.specweave/docs/internal/architecture/critical-patterns.md` (100 lines)
4. ✅ `.specweave/docs/public/guides/project-conventions.md` (90 lines)
5. ✅ `.specweave/docs/internal/processes/deployment.md` (70 lines)

**Total**: 5 files, 460 lines (essential content)

---

## CLAUDE.md Updated

**Added**: 10 lines (project summary + links)

**Location**: Lines 850-860

---

## Content Distribution (Quick Start)

| Content Type | Lines | Status | Destination |
|--------------|-------|--------|-------------|
| Core Architecture | 120 | ✅ Merged | .specweave/docs/internal/architecture/ |
| Tech Stack | 80 | ✅ Merged | .specweave/docs/internal/architecture/ |
| Critical Patterns | 100 | ✅ Merged | .specweave/docs/internal/architecture/ |
| Conventions | 90 | ✅ Merged | .specweave/docs/public/guides/ |
| Deployment | 70 | ✅ Merged | .specweave/docs/internal/processes/ |
| **CLAUDE.md** | **10** | ✅ **Updated** | **Root** |
| **Subtotal Merged** | **470** | | |
| | | | |
| Domain Model (detailed) | 450 | ⏸️ Deferred | Extract when working on appointments |
| Business Rules (detailed) | 280 | ⏸️ Deferred | Extract when working on payments |
| User Module API | 150 | ⏸️ Deferred | Extract when modifying user code |
| Code Examples | 200 | ⏸️ Deferred | Extract as needed per increment |
| **Subtotal Deferred** | **1,080** | | **Document incrementally** |

**Result**: 470 lines merged now, 1,080 lines to extract per increment

**Benefit**: Start in 30-60 minutes, not 1-3 hours

---

## Deferred Content (Extract Per Increment)

The following content remains in the backup and will be extracted when you work on related features:

### 📦 Domain Documentation
- `appointments/domain-model.md` (450 lines)
  → Extract when creating increment for appointments feature

### 📋 Business Rules
- `payments/business-rules.md` (280 lines)
  → Extract when creating increment for payment modifications

### 🔌 API Documentation
- `users/api-endpoints.md` (150 lines)
  → Extract when creating increment for user service changes

### 💻 Code Examples
- Various code snippets (200 lines)
  → Extract as needed

**How to extract later**:
```bash
# When starting increment for appointments
/inc "Refactor appointment booking"

# In spec.md, reference:
# "See backup: .claude/backups/CLAUDE-backup-*.md (appointments section)"

# Or ask:
# "Extract appointment documentation from CLAUDE.md backup"

Skipped Content

• Generic React patterns (25 lines) - Already covered in SpecWeave
• Standard git workflow (15 lines) - Common knowledge
• TypeScript basics (40 lines) - Not project-specific

Total skipped: 80 lines (generic content)

Next Steps

1. ✅ Review merged essential docs (30 min)
2. ✅ Start first increment (immediate)
3. ⏸️ Extract detailed docs as you work on features

Time saved: ~2 hours (vs comprehensive upfront)

### Comprehensive Mode Report

```markdown
# CLAUDE.md Merge Report - Comprehensive Mode

**Date**: 2025-10-26
**Backup File**: .claude/backups/CLAUDE-backup-20251026-143022.md
**Merge Status**: ✅ Complete (All content)
**Mode**: Comprehensive (Upfront Documentation)

---

## Files Created

1. ✅ `.specweave/increments/####-name/docs/domain/appointments/domain-model.md` (450 lines)
2. ✅ `.specweave/docs/internal/architecture/existing-system.md` (320 lines)
3. ✅ `.specweave/docs/internal/architecture/tech-stack.md` (180 lines)
4. ✅ `.specweave/docs/internal/strategy/appointments/business-rules.md` (280 lines)
5. ✅ `.specweave/docs/public/guides/project-conventions.md` (200 lines)
6. ✅ `.specweave/docs/internal/processes/deployment.md` (150 lines)

**Total**: 6 files, 1,580 lines

---

## CLAUDE.md Updated

**Added**: 12 lines (project summary + links)

**Location**: Lines 850-862 (after "Living Documentation Principles")

---

## Content Distribution

| Content Type | Lines | Destination |
|--------------|-------|-------------|
| Domain Model | 450 | .specweave/increments/####-name/docs/domain/ |
| Architecture | 320 | .specweave/docs/internal/architecture/ |
| Tech Stack | 180 | .specweave/docs/internal/architecture/ |
| Business Rules | 280 | .specweave/docs/internal/strategy/ |
| Conventions | 200 | .specweave/docs/public/guides/ |
| Deployment | 150 | .specweave/docs/internal/processes/ |
| **CLAUDE.md** | **12** | **Root** |

**Result**: 99.2% of content distributed to appropriate folders, not bloating CLAUDE.md

---

## Skipped Content

- Generic React patterns (25 lines) - Already covered in SpecWeave
- Standard git workflow (15 lines) - Common knowledge
- TypeScript basics (40 lines) - Not project-specific

**Total skipped**: 80 lines (generic content)

---

## Next Steps

1. ✅ Review generated files in `.specweave/docs/internal/` and `.specweave/docs/public/`
2. ✅ SpecWeave uses auto-detection
3. ✅ Run `npm run docs:dev` to preview documentation
4. ✅ Create features from specifications: `specweave plan-feature {name}`

---

## Backup Preserved

Original backup remains at:
`.claude/backups/CLAUDE-backup-20251026-143022.md`

You can reference it anytime.

---

**Merge Complete! 🎉**

Your project-specific knowledge is now integrated into SpecWeave's structure.

Example Usage

User Request

User: "I installed SpecWeave and it backed up my CLAUDE.md. Can you merge the important parts?"

Your Response

I'll analyze your backup CLAUDE.md and intelligently merge project-specific content.

🔍 Analyzing: .claude/backups/CLAUDE-backup-20251026-143022.md

Found:
- Project: Healthcare Appointment Booking Platform
- Domain model: Patients, Providers, Appointments, Clinics
- Architecture: Microservices (4 services)
- Tech stack: Next.js + Node.js + PostgreSQL
- Business rules: Booking policies, cancellation rules
- Conventions: Naming, API design
- Deployment: CI/CD workflow

📦 Content Distribution Plan:

1. Domain Model → specifications/modules/appointments/domain-model.md
2. Architecture → .specweave/docs/architecture/existing-system.md
3. Tech Stack → .specweave/docs/architecture/tech-stack.md
4. Business Rules → specifications/modules/appointments/business-rules.md
5. Conventions → .specweave/docs/guides/project-conventions.md
6. Deployment → .specweave/docs/guides/deployment.md

CLAUDE.md: Add 12-line summary with links (no bloat!)

Total: 6 files, ~1,580 lines
CLAUDE.md additions: 12 lines (99.2% distributed to folders)

Proceed with merge? (y/n)

After User Approval

✅ Merge complete!

Created:
1. .specweave/increments/####-name/docs/domain/appointments/domain-model.md
2. .specweave/docs/internal/architecture/existing-system.md
3. .specweave/docs/internal/architecture/tech-stack.md
4. .specweave/docs/internal/strategy/appointments/business-rules.md
5. .specweave/docs/public/guides/project-conventions.md
6. .specweave/docs/internal/processes/deployment.md

Updated:
- CLAUDE.md (added 12-line project summary)

Report saved: ai-logs/reports/CLAUDE-MERGE-REPORT-20251026.md

Your project knowledge is now integrated! 🎉

Next: Review files and run `npm run docs:dev` to preview.

Tool Usage

Scan for backups:

const backups = glob(".claude/backups/CLAUDE-backup-*.md");
const latestBackup = backups.sort().reverse()[0];  // Most recent

Parse content:

const content = read(latestBackup);
const sections = parseSections(content);  // Split by headers
const classified = classifyContent(sections);  // Domain, Architecture, etc.

Avoid duplicates:

if (exists(targetPath)) {
  const existing = read(targetPath);
  if (similarity(existing, newContent) > 0.8) {
    skip();  // Already documented
  }
}

Related Documentation

• BROWNFIELD-INTEGRATION-STRATEGY.md
• brownfield-analyzer skill
• CLAUDE.md

Project-Specific Learnings

Before starting work, check for project-specific learnings:

# Check if skill memory exists for this skill
cat .specweave/skill-memories/brownfield-onboarder.md 2>/dev/null || echo "No project learnings yet"

Project learnings are automatically captured by the reflection system when corrections or patterns are identified during development. These learnings help you understand project-specific conventions and past decisions.