Version: 1.2.0 Date: 2025-01-11 Status: 📋 READY TO EXECUTE
- Eliminate ~500+ lines of duplicated content
- Establish Single Source of Truth for each concept
- Reduce maintenance burden by 70%
- Prevent future contradictions through cross-references
- CLAUDE.md reduced from ~170 lines → ~80 lines
- AGENTS.md security rules removed (~31 instructions)
- Modularity philosophy consolidated (3 places → 1 authoritative + 2 links)
- All files have clear "Authoritative: Yes/No" markers
- ✅ CLAUDE.md - Remove duplications, add cross-references
- ✅ AGENTS.md - Remove security rules, keep only project patterns
- ✅ PROJECT_INTAKE.md - Replace modularity section with link
- ✅ SECURITY.md - Deduplicate internal repetitions
- ✅ WORKFLOW.md - Add authoritative markers
- ✅ Create DOCS_MAP.md
- ✅ Update CHANGELOG.md → v1.2.0
- ✅ Update README badges if needed
Current: 170+ lines, Navigator + Reference + Duplicated Rules
Target: 80-90 lines, Navigator ONLY with cross-references
🚀 Быстрый старт (lines 7-22)
- ✅ KEEP AS IS - good navigation
📦 Bash-команды (lines 24-44)
- ✅ KEEP AS IS - essential reference
🎨 Кодстайл (lines 46-65)
- ✅ KEEP AS IS - quick reference is OK
🔐 Безопасность (lines 67-80) ❌ REMOVE detailed rules ✅ REPLACE with:
## 🔐 Безопасность (КРИТИЧНО!)
**📖 ПОЛНАЯ ПОЛИТИКА:** SECURITY.md
**⚠️ ВСЕГДА перед кодированием:**
1. Прочитай SECURITY.md полностью
2. Следуй чеклисту для текущей стадии разработки (SECURITY.md → Stage X)
**Ключевые правила (подробно в SECURITY.md):**
- ✅ Валидация входных данных → SECURITY.md Stage 3
- ✅ Секреты в .env → SECURITY.md "Secrets Management"
- ✅ Параметризованные запросы → SECURITY.md "SQL Injection Prevention"
- ❌ Никогда не hardcode секреты → SECURITY.md "NEVER DO" section
- ❌ Никогда не доверяй user input → SECURITY.md "Zero Trust"
**Для AI-агентов:** Запускай `/security` для автоматического audit🏗️ Флоу работы (lines 82-109) ❌ REMOVE all detailed checklists ✅ REPLACE with:
## 🏗️ Флоу работы
**📖 ПОЛНАЯ ДОКУМЕНТАЦИЯ:** WORKFLOW.md
**Быстрая навигация:**
### Перед началом задачи
→ См. WORKFLOW.md → "Sprint Start" section
### Во время разработки
→ См. WORKFLOW.md → "Implementation" section
→ См. AGENTS.md → Project-specific patterns
### После изменений
→ `make typecheck && make lint`
### Sprint завершение
→ См. WORKFLOW.md → "Sprint Completion Checklist" (29 пунктов!)
→ ⚠️ КРИТИЧНО: Не пропусти Security Requirements!
### Git workflow
→ См. WORKFLOW.md → "Git Workflow" section
→ Commit messages: См. WORKFLOW.md → "Commit Templates"🔧 Настройка окружения (lines 111-135)
- ✅ KEEP AS IS - project-specific info
📋 Важные детали (lines 137-155)
- ✅ KEEP AS IS - [ЗАПОЛНИТЬ] sections
🐛 Отладка (lines 157-180)
- ✅ KEEP AS IS - helpful quick reference
🎯 Приоритеты (lines 182-189)
- ✅ KEEP AS IS - good summary
📝 Slash-команды (lines 191-210)
- ✅ KEEP AS IS - essential reference
- ✅ KEEP AS IS
🔄 Интеграции (lines 221-237)
- ✅ KEEP AS IS
📚 Полезные ссылки (lines 239-245)
- ✅ KEEP AS IS
🎓 Для новых разработчиков (lines 247-256)
- ✅ KEEP AS IS - good onboarding
Expected Result:
- Before: 170 lines
- After: ~85 lines
- Reduction: ~85 lines (50%)
- All references to WORKFLOW.md, SECURITY.md clear
Current: Contains ~31 security instructions duplicating SECURITY.md
Target: Project-specific patterns ONLY, security rules → links
🚫 NEVER DO → 🔐 Security (lines 143-154) ❌ REMOVE all 10 NEVER DO security rules ✅ REPLACE with:
### 🔐 Security (CRITICAL - READ SECURITY.md FIRST!)
**📖 ПОЛНАЯ ПОЛИТИКА БЕЗОПАСНОСТИ:** SECURITY.md
**ДО начала любой задачи с кодом:**
1. Прочитай SECURITY.md → Stage 1 (Planning)
2. Следуй чеклистам для текущей стадии
3. В случае сомнений → `/security` для audit
**Ключевые принципы (все детали в SECURITY.md):**
- 🔐 Secrets management → SECURITY.md "Environment Variables"
- 🔐 Input validation → SECURITY.md Stage 3
- 🔐 SQL injection prevention → SECURITY.md "Database Security"
- 🔐 XSS prevention → SECURITY.md "Output Sanitization"
**AGENTS.md содержит только project-specific security patterns!**
См. "Project Security Patterns" section ниже для специфичных правил этого проекта.✅ ALWAYS DO → 🔐 Security (lines 173-184) ❌ REMOVE all 11 ALWAYS DO security rules ✅ Already covered by section above
🔐 Security Review (lines 230-241) ❌ REMOVE 10-step Security Review checklist ✅ REPLACE with:
### 🔐 Security Review (Before Every Deploy)
**📖 ИСПОЛЬЗУЙ CHECKLIST ИЗ SECURITY.md:**
- См. SECURITY.md → Stage 5 (Pre-Deployment)
- См. SECURITY.md → "Security Sign-Off Template"
**Или используй автоматизацию:**
```bash
/security # Запустить AI-guided security audit
**ADD NEW SECTION (для project-specific patterns):**
```markdown
---
## 🔐 Project Security Patterns
> **Важно:** Этот раздел для СПЕЦИФИЧНЫХ для ЭТОГО проекта security patterns.
> Общие правила безопасности см. SECURITY.md
[ЗАПОЛНИТЬ по мере разработки проекта]
### Pattern 1: [Project-Specific Security Rule]
[Описание специфичного для проекта security pattern]
**Template:**
```markdown
### Pattern N: [Name]
**Context:** [When this applies]
**Rule:** [What to do]
**Reason:** [Why this is important for THIS project]
**Example:**
[Code example]
**Expected Result:**
- Remove: ~31 security instructions
- Add: Clear references to SECURITY.md
- Keep: Space for project-specific patterns
- Reduction: ~25-30 lines
---
### 3. Init/PROJECT_INTAKE.md - MODULARITY DEDUPLICATION
**Current:** Lines 536-576 contain 40-line detailed modularity explanation
**Target:** Replace with link to ARCHITECTURE.md
#### Changes:
**Section 25a. Модульная структура** (lines 536-576)
❌ REMOVE detailed 40-line explanation
✅ REPLACE with:
```markdown
### 25a. Модульная структура (ОБЯЗАТЕЛЬНО!)
**Почему модульная архитектура критична для работы с ИИ:**
📖 **ПОЛНАЯ ФИЛОСОФИЯ:** ARCHITECTURE.md → "Module Architecture" section
**Краткое объяснение:**
1. **Экономия токенов:** ИИ загружает только нужный модуль (100-200 строк), а не весь проект (1000+ строк)
2. **Простота:** Каждый модуль = отдельная задача = легко проверить
3. **Скорость:** Можно делать разные модули параллельно
**Принцип:** Приложение = набор маленьких LEGO-кубиков
📖 **Детали, примеры, диаграммы:** См. ARCHITECTURE.md → "Module Architecture"
**Ваш подход к модульности:**
- [ ] Да, хочу модульную структуру (рекомендовано!)
- [ ] Нет, хочу монолитную структуру (НЕ рекомендуется для работы с ИИ)
**Selected:** [ОТВЕТИТЬ: Модульная (recommended)]
**Как будем разрабатывать модули:**
[ОТВЕТИТЬ: Например, "По одному модулю за раз. См. ARCHITECTURE.md для типичной последовательности"]
Expected Result:
- Before: 40 lines
- After: ~20 lines
- Reduction: ~20 lines
- Full philosophy in ARCHITECTURE.md only
Current: "Input validation" mentioned 3 times
Target: Consolidate or clarify context
Line 209 (Stage 3 - Development):
- [ ] Input validation implementedContext: DEVELOPMENT checklist
Line 485 (Stage 4 - Testing):
- [ ] Input validation on all inputsContext: TESTING checklist
Line 493 (Stage 4 - Testing details):
- [ ] Input validation testedContext: Detailed testing steps
✅ KEEP ALL THREE - contexts differ:
- Line 209: "Implement validation" (dev)
- Line 485: "Check it's implemented" (test checklist)
- Line 493: "Actually test it" (test details)
❌ NO CHANGES NEEDED - these are different contexts, not duplications
Alternative (if we want to optimize):
- Keep lines 485 and 493 (testing)
- Line 209: Change to "See Stage 4 for validation checklist"
Recommendation: ✅ Keep as is - clarity > brevity for security
Current: No indication this is the authoritative workflow source
Target: Add header marker
File Header (after line 6) ADD:
> **📋 Authoritative Source:** This is the SINGLE SOURCE OF TRUTH for:
> - Sprint workflow and processes
> - Git workflow and commit templates
> - Sprint completion checklists
>
> Other files (CLAUDE.md, AGENTS.md) link here, don't duplicate.Expected Result:
- Clear indication of authoritative status
- Other files should reference this, not duplicate
Purpose: Map all documentation files and their relationships
Content:
# Documentation Map
> **Purpose:** Navigate the Claude Code Starter framework documentation
> **Version:** 1.2.0
> **Last Updated:** 2025-01-11
---
## 📚 Documentation Structure
### 🎯 Single Sources of Truth
These files are **authoritative** for their domains. Other files reference them.
| File | Domain | Lines | Role |
|------|--------|-------|------|
| **WORKFLOW.md** | Sprint processes, Git workflow | ~600 | Authoritative workflow |
| **SECURITY.md** | Security practices, checklists | ~500 | Authoritative security policy |
| **PROJECT_INTAKE.md** | Project requirements, User Personas | ~775 | Project definition template |
| **ARCHITECTURE.md** | Tech decisions, modularity philosophy | ~600 | Authoritative architecture |
| **BACKLOG.md** | Implementation status, roadmap | ~370 | Single source of truth for status |
| **AGENTS.md** | Project-specific AI patterns | ~465 | Project-specific patterns (NOT universal rules) |
### 🧭 Navigation Files
These files **link to** authoritative sources. They don't duplicate.
| File | Purpose | Target Audience |
|------|---------|----------------|
| **CLAUDE.md** | Auto-loads in Claude Code, navigation | AI agents (auto-loaded) |
| **README.md** | Project overview, getting started | Humans (first contact) |
| **README_RU.md** | Russian version | Russian-speaking users |
### 📖 Supporting Files
| File | Purpose |
|------|---------|
| **PLAN_TEMPLATE.md** | Template for planning tasks |
| **README-TEMPLATE.md** | Template for project README |
| **MIGRATION.md** | Guide for migrating legacy projects |
| **CHANGELOG.md** | Version history |
| **CONSISTENCY_AUDIT.md** | Audit report (v1.2.0) |
| **REFACTORING_PLAN.md** | This file |
---
## 🗺️ Concept Ownership Map
**When you need information about...**
| Concept | Authoritative File | Quick Reference |
|---------|-------------------|-----------------|
| Sprint completion checklist | WORKFLOW.md → "Sprint Completion" | CLAUDE.md → link |
| Security rules | SECURITY.md → stage checklists | CLAUDE.md → link |
| Git workflow | WORKFLOW.md → "Git Workflow" | CLAUDE.md → link |
| User Personas | PROJECT_INTAKE.md → "User Personas" | CLAUDE.md → mention |
| Modularity philosophy | ARCHITECTURE.md → "Module Architecture" | PROJECT_INTAKE.md → link |
| Project status | BACKLOG.md (single source of truth) | Always check first |
| Project-specific patterns | AGENTS.md (project-specific only!) | Filled during development |
---
## 🔄 Cross-Reference Rules
### For CLAUDE.md (Navigator)
- **DO:** Link to authoritative sources
- **DON'T:** Duplicate details from other files
- **Format:** "📖 См. WORKFLOW.md → Section Name"
### For AGENTS.md (Project Patterns)
- **DO:** Document project-specific patterns
- **DON'T:** Duplicate universal rules from SECURITY.md, WORKFLOW.md
- **Format:** Link to authoritative file for universal rules
### For Other Files
- **Always specify** if authoritative or reference
- **Link, don't duplicate** when referencing other concepts
- **Update DOCS_MAP.md** when adding new files
---
## 📝 Maintenance Guidelines
### When Adding New Concept
1. **Choose ONE authoritative file** for it
2. **Update this map** with the new concept
3. **Other files link** to the authoritative file
### When Updating Concept
1. **Update authoritative file** first
2. **Check DOCS_MAP.md** for files that reference it
3. **Update links** if section names changed
### Monthly Review
- Check for new duplications
- Verify links still valid
- Update this map if structure changed
---
## 🚫 Anti-Patterns to Avoid
❌ **DON'T duplicate detailed content** across files
❌ **DON'T create "summary" sections** that duplicate authoritative source
❌ **DON'T forget to update links** when moving/renaming sections
❌ **DON'T mix universal rules** (SECURITY.md) with project patterns (AGENTS.md)
✅ **DO link to authoritative sources**
✅ **DO keep CLAUDE.md as navigation only**
✅ **DO specify authoritative status** clearly
✅ **DO separate universal vs project-specific** content
---
*This map ensures consistency and prevents duplication across documentation*
*Last updated: 2025-01-11*
*Maintained by: Project Lead + AI Agents*After refactoring, test that framework still works:
Prompt: "Прочитай CLAUDE.md и объясни как начать новый sprint"
Expected: AI finds link to WORKFLOW.md → reads Sprint Start section → explains correctly
Prompt: "Какие security practices нужно соблюдать при разработке?"
Expected: AI points to SECURITY.md, mentions stages, explains key principles
Prompt: "Объясни почему модульная архитектура важна"
Expected: AI finds link in CLAUDE.md → reads ARCHITECTURE.md Module section → explains with diagrams
- Manually check all links work
- Verify section names match
- Ensure no broken references
- Total files with duplications: 6
- Duplicated lines: ~500+
- CLAUDE.md size: ~170 lines
- AGENTS.md security rules: ~31
- Modularity philosophy: 237 lines across 3 files
- Total files with duplications: 0
- Duplicated lines: ~50 (acceptable cross-references)
- CLAUDE.md size: ~85 lines
- AGENTS.md security rules: 0 (links only)
- Modularity philosophy: ~196 lines in 1 file + links
- Duplication reduction: 90%
- CLAUDE.md reduction: 50%
- Maintenance burden: -70%
- Contradiction risk: ELIMINATED
- Refactor Init/CLAUDE.md
- Refactor init_eng/CLAUDE.md (English version)
- Refactor Init/AGENTS.md
- Refactor init_eng/AGENTS.md
- Refactor Init/PROJECT_INTAKE.md (modularity section)
- Refactor init_eng/PROJECT_INTAKE.md
- Add authoritative marker to Init/WORKFLOW.md
- Add authoritative marker to init_eng/WORKFLOW.md
- Add authoritative marker to Init/SECURITY.md
- Add authoritative marker to init_eng/SECURITY.md
- Add authoritative marker to Init/ARCHITECTURE.md
- Add authoritative marker to init_eng/ARCHITECTURE.md
- Create DOCS_MAP.md
- Update CHANGELOG.md → v1.2.0
- Test AI agent with CLAUDE.md
- Test security references work
- Test modularity references work
- Manual check all cross-references
- Verify English versions match Russian
- Git add all changes
- Create comprehensive commit message
- Tag as v1.2.0
- Push to remote
git commit -m "$(cat <<'EOF'
refactor: Deduplicate framework meta-documentation (v1.2.0)
Eliminate ~500+ lines of duplicated content across 6 files.
## Changes
### CLAUDE.md (Russian & English)
- Reduced from ~170 → ~85 lines (50% reduction)
- Removed duplicated workflow/security rules
- Added cross-references to authoritative sources
- Now serves as navigator only
### AGENTS.md (Russian & English)
- Removed ~31 duplicated security instructions
- Added clear references to SECURITY.md
- Clarified purpose: project-specific patterns only
- Added template for project patterns
### PROJECT_INTAKE.md (Russian & English)
- Deduplicated modularity philosophy section
- Reduced from 40 → ~20 lines
- Added link to ARCHITECTURE.md for full philosophy
### Authoritative Markers
- Added to WORKFLOW.md, SECURITY.md, ARCHITECTURE.md
- Clarifies single source of truth status
### Documentation
- Created DOCS_MAP.md (navigation guide)
- Created CONSISTENCY_AUDIT.md (audit report)
- Created REFACTORING_PLAN.md (this file)
- Updated CHANGELOG.md → v1.2.0
## Impact
- Duplication reduced by 90%
- Maintenance burden reduced by 70%
- Contradiction risk eliminated
- Clear single source of truth for each concept
## Files Changed
- Init/CLAUDE.md, init_eng/CLAUDE.md
- Init/AGENTS.md, init_eng/AGENTS.md
- Init/PROJECT_INTAKE.md, init_eng/PROJECT_INTAKE.md
- Init/WORKFLOW.md, init_eng/WORKFLOW.md (+markers)
- Init/SECURITY.md, init_eng/SECURITY.md (+markers)
- Init/ARCHITECTURE.md, init_eng/ARCHITECTURE.md (+markers)
- DOCS_MAP.md (new)
- CONSISTENCY_AUDIT.md (new)
- REFACTORING_PLAN.md (new)
- CHANGELOG.md (v1.2.0)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
EOF
)"Status: 📋 READY TO EXECUTE Next Step: Get user approval → Begin Phase 1
🤖 Generated with Claude Code