---
title: "Cross-Platform AI Workflows - Integrace Více Nástrojů - AI Prompt"
description: "Naučte se navrhovat a optimalizovat pracovní postupy, které kombinují více AI nástrojů (ChatGPT, Claude, Copilot, Cursor, Perplexity) pro maximální efektivitu..."
image: "https://zacinamsai.cz/images/og-default.jpg"
---

Produktivita  Expert  Ostatní 

# Cross-Platform AI Workflows - Integrace Více Nástrojů

Naučte se navrhovat a optimalizovat pracovní postupy, které kombinují více AI nástrojů (ChatGPT, Claude, Copilot, Cursor, Perplexity) pro maximální efektivitu a kvalitu výstupů.

  ##   Prompt text 

 **Délka:** Dlouhý 

Jsi expert na navrhování cross-platform AI workflows a integrace různých AI nástrojů. Pomoz mi vytvořit optimální pracovní postup pro mou úlohu.  

Můj projekt a kontext:  
- Typ projektu: [TYP - např. "full-stack webová aplikace", "content marketing", "data analysis", "výzkumný projekt"]  
- Hlavní úloha: [ÚLOHA - např. "vývoj nové funkce", "psaní článků", "analýza dat", "tvorba dokumentace"]  
- Aktuální workflow: [SOUČASNÝ STAV - např. "používám jen ChatGPT", "manuální přepínání mezi nástroji", "neefektivní proces"]  
- Dostupné nástroje: [NÁSTROJE - např. "ChatGPT Plus, GitHub Copilot, Claude Pro", "free verze", "firemní licence"]  
- Časový budgeting: [ČAS - např. "2 hodiny denně na vývoj", "celý pracovní den", "side project 10h týdně"]  
- Bolestivé body: [PROBLÉMY - např. "ztrácím čas přepínáním", "duplikuji práci", "nekonzistentní výstupy"]  

Vytvoř mi:  

1. **Analýzu Současného Stavu**  
 - Kde ztrácím nejvíc času  
 - Které části procesu jsou inefficientní  
 - Co by AI mohlo zrychlit/zlepšit  
 - Kde potřebuji lidský oversight  

2. **Optimalizovaný Multi-Tool Workflow**  
 - Který nástroj použít pro kterou fázi  
 - Handoff body mezi nástroji (jak předávat kontext)  
 - Paralelní vs sekvenční tasking  
 - Context management strategie  
 - Templates a reusable prompts  

3. **Konkrétní Workflow Patterns**  
 Pro každou hlavní úlohu navrhni:  
 - **Tool Selection:** Který AI nástroj na kterou subtask  
 - **Sequencing:** V jakém pořadí taskov postupovat  
 - **Context Passing:** Jak předávat výstupy mezi nástroji  
 - **Quality Gates:** Kde kontrolovat kvalitu  
 - **Automation Opportunities:** Co automatizovat  

4. **Integration Points**  
 - Jak spojit Copilot/Cursor s Claude/ChatGPT  
 - Kdy použít Perplexity vs ChatGPT browsing  
 - Jak kombinovat code gen s dokumentací  
 - Best practices pro context sharing  
 - Tools a skripty pro automation  

5. **Workflow Templates**  

 **Template 1: Software Development**  
 &#96;&#96;&#96;  
 1. Research/Planning: Perplexity + Claude (research, architecture)  
 2. Code Generation: Cursor Composer + Copilot (implementation)  
 3. Code Review: Claude (code quality, security, best practices)  
 4. Documentation: ChatGPT (README, API docs, comments)  
 5. Testing: Cursor/Copilot (test generation, debugging)  
 &#96;&#96;&#96;  

 **Template 2: Content Creation**  
 &#96;&#96;&#96;  
 1. Topic Research: Perplexity (sources, data, insights)  
 2. Outline Creation: Claude (structure, flow, SEO)  
 3. Draft Writing: ChatGPT (long-form content)  
 4. Editing/Polish: Claude (style, consistency, quality)  
 5. SEO Optimization: ChatGPT (keywords, meta, structure)  
 &#96;&#96;&#96;  

 **Template 3: Data Analysis**  
 &#96;&#96;&#96;  
 1. Data Exploration: ChatGPT Code Interpreter (initial analysis)  
 2. Statistical Analysis: Claude (advanced statistics, interpretation)  
 3. Visualization: Python + Copilot (chart generation)  
 4. Insight Generation: Perplexity (context, benchmarks)  
 5. Report Writing: ChatGPT (executive summaries)  
 &#96;&#96;&#96;  

 Vytvoř podobné templates pro [TYP PROJEKTU].  

6. **Context Management Strategy**  
 - Jak organizovat konverzace  
 - Naming conventions pro projekty/chaty  
 - Context compression techniky  
 - Knowledge base building  
 - Dokumentace workflow  

7. **Automation &amp; Scripts**  
 - CLI tools pro rychlé přepínání  
 - Browser extensions pro workflow  
 - Keyboard shortcuts setup  
 - Template prompts/snippets  
 - Integration scripts  

8. **Měření Efektivity**  
 - KPI tracking (čas ušetřený, kvalita, produktivita)  
 - Bottleneck identification  
 - ROI calculation per nástroj  
 - Continuous improvement process  

9. **Common Pitfalls &amp; Solutions**  
 - Context loss při přepínání  
 - Inconsistent outputs  
 - Over-reliance na jeden nástroj  
 - Poor handoff management  
 - Inefficient context sharing  

10. **Advanced Patterns**  
 - Ensemble methods (kombinace outputs)  
 - Verification workflows (cross-checking)  
 - Iterative refinement loops  
 - Parallel processing  
 - Fallback strategies  

Zaměř se na:  
✅ Praktické, actionable kroky  
✅ Konkrétní nástroje a jejich role  
✅ Měřitelné improvement metriky  
✅ Real-world příklady  
✅ Quick wins + long-term optimization

  [  Interaktivní tvůrce ](https://zacinamsai.cz/prompts/cross-platform-ai-workflows-integrace-vice-nastroju/builder) 

 ##   Příklad použití 

### Vstup:

Typ projektu: Full-stack webová aplikace (Next.js, Node.js, PostgreSQL)  
Hlavní úloha: Vývoj nových features + dokumentace + testing  
Aktuální workflow: Používám jen GitHub Copilot v VS Code, všechno dělám tam  
Dostupné nástroje: GitHub Copilot Individual, Claude Free, ChatGPT Free, můžu investovat $30/měsíc  
Časový budgeting: 4 hodiny denně (vedlejší projekt)  
Bolestivé body: Copilot je dobrý na code gen, ale špatný na plánování architektury a dokumentaci. Často musím googlovat best practices. Testing píšu manuálně a zabere to věčnost.

### Výstup:

\# Cross-Platform AI Workflow Pro Full-Stack Development  
\## Optimalizace Vašeho Development Procesu 🚀  

---  

\## 1. Analýza Současného Stavu  

\### ⏰ Kde Ztrácíte Nejvíc Času  

**Current Time Distribution (odhadované):**  

&#124; Aktivita &#124; Čas/den &#124; % &#124; Inefficiency &#124;  
&#124;----------&#124;---------&#124;---&#124;--------------&#124;  
&#124; Architecture planning &#124; 30 min &#124; 12.5% &#124; ⚠️ Copilot není design tool &#124;  
&#124; Code writing &#124; 90 min &#124; 37.5% &#124; ✅ OK (Copilot silná stránka) &#124;  
&#124; Googling best practices &#124; 45 min &#124; 18.75% &#124; 🔴 Manuální research &#124;  
&#124; Writing tests &#124; 60 min &#124; 25% &#124; 🔴 Repetitivní práce &#124;  
&#124; Documentation &#124; 15 min &#124; 6.25% &#124; 🔴 Odkládáte to (nudné) &#124;  

**Total: 240 min/den, ~30% inefficient (72 min/den ztraceno)**  

\### 🎯 Kde AI Může Pomoci  

**High-Impact Opportunities:**  

1. **Architecture Planning (30 min → 10 min)** 💡  
 - Claude Pro: Zkušený architect, navrhne strukturu, API design  
 - Perplexity: Najde best practices, designové patterny  
 - **Savings: 20 min/den**  

2. **Research &amp; Best Practices (45 min → 10 min)** 📚  
 - Perplexity: Agreguje aktuální best practices s citacemi  
 - Claude: Vysvětlí kontext a trade-offs  
 - **Savings: 35 min/den**  

3. **Test Writing (60 min → 20 min)** ✅  
 - Copilot: Generuje unit tests  
 - Claude: Navrhne edge cases, integration test scenarios  
 - **Savings: 40 min/den**  

4. **Documentation (15 min → 15 min, ale BETTER)** 📝  
 - ChatGPT: Generuje README, API docs, code comments  
 - Claude: Review dokumentace, přidá chybějící části  
 - **Quality improvement, not time savings**  

**Total Potential Savings: 95 min/den (~40% time saved)**  

\### 🛠️ Tool Gap Analysis  

&#124; Need &#124; Current Tool &#124; Gap &#124; Better Solution &#124;  
&#124;------&#124;-------------&#124;-----&#124;-----------------&#124;  
&#124; Architecture design &#124; Copilot &#124; ❌ Copilot generuje kód, ne architekturu &#124; Claude (reasoning) &#124;  
&#124; Code generation &#124; Copilot &#124; ✅ Good &#124; Keep Copilot &#124;  
&#124; Best practices research &#124; Google &#124; ❌ Slow, fragmented &#124; Perplexity (aggregated answers) &#124;  
&#124; Test generation &#124; Copilot &#124; ⚠️ Basic tests only &#124; Copilot + Claude (edge cases) &#124;  
&#124; Documentation &#124; Manual &#124; ❌ Odkládáte to &#124; ChatGPT (batch generation) &#124;  
&#124; Code review &#124; Self &#124; ❌ Missing errors &#124; Claude (security, performance review) &#124;  

---  

\## 2. Optimalizovaný Multi-Tool Workflow  

\### 🎯 Recommended Tool Stack ($30 Budget)  

**Option A: Maximum Versatility ($30/mo)**  
- ✅ **GitHub Copilot Individual** ($10/mo) - Keep for code gen  
- ✅ **Claude Pro** ($20/mo) - Architecture, review, complex reasoning  
- ❌ **ChatGPT Free** - Use for documentation  
- ❌ **Perplexity Free** - Use for quick research  

**Option B: Maximum Coding Power ($30/mo)**  
- ✅ **Cursor Pro** ($20/mo) - Includes Copilot-like features + Composer  
- ✅ **Claude Free** - Architecture planning (100 msgs/day enough)  
- ✅ **Perplexity Free** - Research when needed  
- ✅ **$10 saved** or invest in ChatGPT Plus for docs  

**Recommendation: Option B (Cursor Pro + Claude Free)**  
- Cursor Composer replaces Copilot + VS Code  
- Multi-file editing = faster feature implementation  
- Claude Free enough for planning (you won't hit limits with 4h/day work)  
- Perplexity Free adequate for occasional research  

\### 📋 Workflow Blueprint  

**Phase-Based Tool Allocation:**  

&#96;&#96;&#96;  
Feature Development Lifecycle → Tool Assignment  

1. 📐 PLANNING (15 min)  
 └─ Claude: Architecture design, API contracts, data models  

2. 🔍 RESEARCH (10 min)  
 └─ Perplexity: Best practices, libraries, design patterns  

3. 💻 IMPLEMENTATION (90 min)  
 ├─ Cursor Composer: Multi-file feature implementation  
 ├─ Cursor Chat: Context-aware help during coding  
 └─ Cursor CMD+K: Quick inline edits  

4. ✅ TESTING (20 min)  
 ├─ Cursor: Generate basic unit tests  
 └─ Claude: Design edge case scenarios, integration tests  

5. 🔎 CODE REVIEW (10 min)  
 └─ Claude: Security audit, performance review, best practices check  

6. 📝 DOCUMENTATION (15 min)  
 ├─ ChatGPT Free: Generate README, API docs, inline comments  
 └─ Claude: Review docs for completeness, clarity  

Total: 160 min (vs 240 min current) = 33% faster  
&#96;&#96;&#96;  

---  

\## 3. Konkrétní Workflow Patterns  

\### Pattern 1: New Feature Implementation 🆕  

**Scenario:** User authentication with JWT tokens  

**Step-by-Step Multi-Tool Process:**  

\#### Step 1: Architecture Planning (Claude, 10 min)  

**Prompt Template:**  
&#96;&#96;&#96;  
Context: Next.js 14 app router + Node.js Express API + PostgreSQL  
Feature: JWT authentication system  

Design:  
1. Database schema for users table  
2. API endpoints (/auth/register, /auth/login, /auth/refresh)  
3. Middleware for protected routes  
4. Frontend auth context structure  
5. Security best practices  

Consider:  
- Token storage (httpOnly cookies vs localStorage)  
- Refresh token rotation  
- CSRF protection  
- Rate limiting  
- Password hashing (bcrypt)  
&#96;&#96;&#96;  

**Claude Output:** Architecture design document → Save to &#96;docs/auth-architecture.md&#96;  

\#### Step 2: Best Practices Research (Perplexity, 5 min)  

**Query:**  
&#96;&#96;&#96;  
JWT authentication best practices 2024:  
- Token expiration times  
- Refresh token strategies  
- Security headers  
- OWASP recommendations  
&#96;&#96;&#96;  

**Perplexity Output:** Curated best practices with sources → Add to architecture doc  

\#### Step 3: Database Schema (Cursor Composer, 10 min)  

**Composer Prompt:**  
&#96;&#96;&#96;  
Create PostgreSQL migration for users table based on:  
[paste Claude's schema design]  

Include:  
- UUID primary key  
- Email (unique, indexed)  
- Password hash  
- Created/updated timestamps  
- Refresh token storage  
&#96;&#96;&#96;  

**Files Created:**  
- &#96;migrations/001_create_users_table.sql&#96;  
- &#96;migrations/002_add_refresh_tokens_table.sql&#96;  

\#### Step 4: Backend Implementation (Cursor Composer, 40 min)  

**Composer Multi-File Prompt:**  
&#96;&#96;&#96;  
Implement JWT auth system:  

1. src/models/User.js - Sequelize model  
2. src/controllers/authController.js - Register, login, refresh, logout  
3. src/middleware/authMiddleware.js - Protected route middleware  
4. src/routes/auth.js - Auth routes  
5. src/utils/jwtHelper.js - Token generation/verification  

Use:  
- bcrypt for password hashing  
- jsonwebtoken library  
- httpOnly cookies for tokens  
- Express async error handling  
&#96;&#96;&#96;  

**Cursor generates all 5 files simultaneously using Composer Mode**  

\#### Step 5: Frontend Implementation (Cursor Composer, 30 min)  

**Composer Prompt:**  
&#96;&#96;&#96;  
Create Next.js 14 auth system:  

1. app/contexts/AuthContext.tsx - Auth provider  
2. app/hooks/useAuth.ts - Auth hook  
3. app/api/auth/[...nextauth]/route.ts - API routes  
4. app/components/LoginForm.tsx - Login UI  
5. app/components/ProtectedRoute.tsx - Route protection  
6. middleware.ts - Next.js middleware for auth  

Features:  
- Automatic token refresh  
- Loading states  
- Error handling  
- Redirect after login  
- Persist auth state  
&#96;&#96;&#96;  

**Cursor generates complete auth frontend**  

\#### Step 6: Testing (Cursor + Claude, 20 min)  

**Cursor for Unit Tests:**  
&#96;&#96;&#96;  
Generate Jest tests for:  
- authController.test.js (register, login, refresh)  
- authMiddleware.test.js (token validation)  
- jwtHelper.test.js (token generation/verification)  

Include:  
- Happy path scenarios  
- Error cases (invalid tokens, expired tokens)  
- Edge cases (missing headers, malformed requests)  
&#96;&#96;&#96;  

**Claude for Test Scenarios:**  
&#96;&#96;&#96;  
Review auth implementation and suggest:  
1. Integration test scenarios (full auth flow)  
2. Security test cases (XSS, CSRF, timing attacks)  
3. Edge cases I might have missed  
4. Performance test scenarios (token refresh under load)  
&#96;&#96;&#96;  

**Files Created:**  
- &#96;tests/unit/authController.test.js&#96;  
- &#96;tests/integration/authFlow.test.js&#96;  
- &#96;tests/security/authSecurity.test.js&#96;  

\#### Step 7: Security Review (Claude, 10 min)  

**Claude Security Audit Prompt:**  
&#96;&#96;&#96;  
Review this JWT auth implementation for security issues:  

[paste key files: authController.js, authMiddleware.js]  

Check for:  
- SQL injection vulnerabilities  
- XSS risks  
- CSRF protection gaps  
- Insecure token storage  
- Missing rate limiting  
- Weak password requirements  
- Timing attack vulnerabilities  
- Missing security headers  

Provide:  
- Security score (1-10)  
- Critical issues (must fix)  
- Recommended improvements  
- OWASP compliance check  
&#96;&#96;&#96;  

**Claude Output:** Security audit report → Fix critical issues  

\#### Step 8: Documentation (ChatGPT Free, 10 min)  

**ChatGPT Prompt:**  
&#96;&#96;&#96;  
Generate documentation:  

1. README.md section:  
 - How to set up auth  
 - Environment variables needed  
 - Usage examples (register, login, protected routes)  

2. API.md:  
 - Endpoint documentation  
 - Request/response examples  
 - Error codes  

3. SECURITY.md:  
 - Security considerations  
 - Best practices for deployment  
 - Token management guidelines  

Based on code:  
[paste authController.js and routes]  
&#96;&#96;&#96;  

**ChatGPT Output:** Complete documentation → Review with Claude for accuracy  

\#### Step 9: Integration &amp; Testing (Manual + Cursor, 20 min)  

1. **Manual Testing:**  
 - Test registration flow  
 - Test login flow  
 - Test token refresh  
 - Test protected routes  
 - Check error handling  

2. **Cursor Chat for Debugging:**  
 &#96;&#96;&#96;  
 Bug: Refresh token not rotating on /auth/refresh  

 Here's the current implementation:  
 [paste code]  

 Help me debug and fix.  
 &#96;&#96;&#96;  

---  

\### Pattern 2: Bug Fix &amp; Debugging 🐛  

**Scenario:** Payment processing failing intermittently  

**Multi-Tool Debugging Process:**  

\#### Step 1: Error Analysis (Cursor Chat, 5 min)  

**In Cursor, select error logs and prompt:**  
&#96;&#96;&#96;  
Analyze this payment error:  
[paste error stack trace]  

Help me:  
1. Identify root cause  
2. Suggest fix  
3. Prevent future occurrences  
&#96;&#96;&#96;  

\#### Step 2: Research Known Issues (Perplexity, 5 min)  

**Query:**  
&#96;&#96;&#96;  
Stripe payment webhook timeout errors Next.js:  
- Common causes  
- Solutions from GitHub issues  
- Best practices for webhook handling  
&#96;&#96;&#96;  

\#### Step 3: Solution Design (Claude, 5 min)  

**Claude Prompt:**  
&#96;&#96;&#96;  
Design solution for Stripe webhook timeouts:  

Current implementation:  
[paste webhook handler code]  

Requirements:  
- Handle long-running operations (&gt;30s)  
- Ensure idempotency  
- Implement retry logic  
- Add proper logging  

Suggest:  
- Architecture changes  
- Queue system (if needed)  
- Monitoring approach  
&#96;&#96;&#96;  

\#### Step 4: Implementation (Cursor, 20 min)  

**Cursor Composer:**  
&#96;&#96;&#96;  
Implement Claude's solution:  
1. Add job queue (BullMQ)  
2. Refactor webhook handler  
3. Add idempotency keys  
4. Implement retry logic  
5. Add logging/monitoring  
&#96;&#96;&#96;  

\#### Step 5: Testing (Cursor + Manual, 15 min)  

- Cursor generates tests  
- Manual testing with Stripe test mode  
- Verify idempotency  
- Test retry logic  

**Total: 50 min (vs 2-3 hours manual debugging)**  

---  

\### Pattern 3: Code Refactoring 🔄  

**Scenario:** Legacy component needs modernization  

**Refactoring Workflow:**  

\#### Step 1: Refactoring Plan (Claude, 10 min)  

**Prompt:**  
&#96;&#96;&#96;  
Analyze this React component and create refactoring plan:  

[paste legacy component code - 500 lines, class-based, mixed concerns]  

Goals:  
- Convert to functional component  
- Extract business logic to custom hooks  
- Improve performance (memoization)  
- Better error handling  
- TypeScript types  
- Test coverage  

Provide:  
- Step-by-step refactoring plan  
- Risk assessment  
- Estimated effort  
- Testing strategy  
&#96;&#96;&#96;  

**Claude Output:** Detailed refactoring roadmap  

\#### Step 2: Incremental Refactoring (Cursor Composer, 60 min)  

**Phase 1: Convert to Functional (Composer):**  
&#96;&#96;&#96;  
Convert UserProfile class component to functional:  
- Use hooks (useState, useEffect, useCallback)  
- Preserve all functionality  
- Maintain props interface  
&#96;&#96;&#96;  

**Phase 2: Extract Logic (Composer):**  
&#96;&#96;&#96;  
Extract business logic to custom hooks:  
1. useUserData - data fetching  
2. useUserPermissions - permission checking  
3. useUserActions - CRUD operations  

Move to hooks/ directory  
&#96;&#96;&#96;  

**Phase 3: Performance (Cursor CMD+K):**  
&#96;&#96;&#96;  
Optimize UserProfile performance:  
- Add React.memo  
- Memoize callbacks with useCallback  
- Memoize derived data with useMemo  
- Split into smaller sub-components  
&#96;&#96;&#96;  

**Phase 4: TypeScript (Cursor):**  
&#96;&#96;&#96;  
Add TypeScript types:  
- Props interface  
- State types  
- API response types  
- Custom hook return types  
&#96;&#96;&#96;  

\#### Step 3: Testing (Cursor + Claude, 30 min)  

**Cursor:** Generate React Testing Library tests  
**Claude:** Design complex interaction test scenarios  

\#### Step 4: Review (Claude, 10 min)  

**Claude Final Review:**  
&#96;&#96;&#96;  
Review refactored component:  
1. Code quality improvements  
2. Performance gains  
3. Maintainability score  
4. Missing test coverage  
5. Potential issues  
&#96;&#96;&#96;  

**Total: 110 min (vs 4-6 hours manual refactoring)**  

---  

\## 4. Integration Points &amp; Context Passing  

\### 🔗 Seamless Tool Transitions  

\#### Transition 1: Claude (Planning) → Cursor (Implementation)  

**Best Practice: Structured Output Format**  

**Claude Prompt (ends with):**  
&#96;&#96;&#96;  
Format output as:  
1. File structure (tree format)  
2. Each file with:  
 - File path  
 - Code implementation  
 - Dependencies needed  

This will be directly pasted into Cursor Composer.  
&#96;&#96;&#96;  

**Claude Output:**  
&#96;&#96;&#96;  
File Structure:  
src/  
├── models/  
│ └── User.js  
├── controllers/  
│ └── authController.js  
└── middleware/  
 └── authMiddleware.js  

--- File: src/models/User.js ---  
const { DataTypes } = require('sequelize');  
...  
[complete implementation]  

--- File: src/controllers/authController.js ---  
...  
&#96;&#96;&#96;  

**Cursor Composer Prompt:**  
&#96;&#96;&#96;  
Implement these files:  

[paste Claude's structured output]  

Ensure:  
- Correct imports  
- Proper error handling  
- TypeScript types  
&#96;&#96;&#96;  

**Result:** Seamless handoff, no context loss ✅  

\#### Transition 2: Perplexity (Research) → Claude (Design)  

**Perplexity Research Output → Notion/Docs:**  
Save Perplexity findings to shared doc (e.g., Notion page)  

**Claude Design Prompt:**  
&#96;&#96;&#96;  
Based on this research: [Notion link or paste key findings]  

Design [system/feature/architecture]  
&#96;&#96;&#96;  

**Benefit:** Claude has full research context ✅  

\#### Transition 3: Cursor (Code) → Claude (Review)  

**Export Code for Review:**  

**Option A: Direct Paste (Small Changes)**  
&#96;&#96;&#96;  
Review this implementation:  

File: authController.js  
[paste code, max ~500 lines]  

Check:  
- Security issues  
- Performance problems  
- Best practices violations  
&#96;&#96;&#96;  

**Option B: Summary for Large Changes**  
&#96;&#96;&#96;  
I implemented JWT auth system:  

Key Files:  
1. authController.js - Handles login/register/refresh  
2. authMiddleware.js - Validates tokens  
3. jwtHelper.js - Token generation/verification  

Architecture decisions:  
- httpOnly cookies for token storage  
- Refresh token rotation every 7 days  
- bcrypt with 12 rounds for passwords  

Please review:  
- Is architecture secure?  
- Any critical vulnerabilities?  
- Performance concerns?  
- Better approaches?  

[paste critical code sections only, ~200 lines]  
&#96;&#96;&#96;  

**Result:** Effective review without overwhelming Claude's context ✅  

\#### Transition 4: Claude (Review Findings) → Cursor (Fixes)  

**Claude Review Output (structured):**  
&#96;&#96;&#96;  
Security Issues:  
1. 🔴 CRITICAL: SQL injection in line 45  
 Fix: Use parameterized queries  

2. ⚠️ WARNING: Missing rate limiting  
 Fix: Add express-rate-limit middleware  

Code to fix:  
--- authController.js:45 ---  
BEFORE: db.query(&#96;SELECT * FROM users WHERE email = '${email}'&#96;)  
AFTER: db.query('SELECT * FROM users WHERE email = ?', [email])  
&#96;&#96;&#96;  

**Cursor Prompt:**  
&#96;&#96;&#96;  
Fix these security issues:  

[paste Claude's structured findings]  

Apply fixes to:  
- authController.js  
- Add rate limiting middleware  
&#96;&#96;&#96;  

**Result:** Direct, actionable fixes implemented ✅  

---  

\### 📋 Context Management Best Practices  

\#### 1. Project Documentation Hub (Notion/Obsidian)  

**Structure:**  
&#96;&#96;&#96;  
Project Root/  
├── Architecture/  
│ ├── System Design.md (Claude outputs)  
│ ├── API Contracts.md  
│ └── Database Schema.md  
├── Research/  
│ ├── Best Practices.md (Perplexity findings)  
│ ├── Library Comparisons.md  
│ └── Design Patterns.md  
├── Implementation/  
│ ├── Feature Specs.md  
│ └── Code Review Notes.md (Claude reviews)  
└── Decisions/  
 └── ADRs/ (Architecture Decision Records)  
&#96;&#96;&#96;  

**Workflow:**  
1. Claude outputs → Save to Architecture/  
2. Perplexity research → Save to Research/  
3. Claude reviews → Save to Implementation/  
4. Reference docs in all tool prompts  

**Benefits:**  
- ✅ Persistent context across sessions  
- ✅ No information loss  
- ✅ Easy reference in prompts  
- ✅ Team knowledge sharing  

\#### 2. Conversation Naming Convention  

**Format:** &#96;[Tool]-[Project]-[Feature]-[Date]&#96;  

**Examples:**  
- &#96;Claude-NextApp-Auth-2024-01-15&#96;  
- &#96;Cursor-NextApp-PaymentAPI-2024-01-16&#96;  
- &#96;Perplexity-NextApp-StripeIntegration-2024-01-17&#96;  

**Benefits:**  
- ✅ Easy to find past conversations  
- ✅ Understand context at a glance  
- ✅ Track decision timeline  

\#### 3. Cross-Tool Context Compression  

**When context is too large for single prompt:**  

**Technique: Summary → Details Pattern**  

**Step 1:** Get Claude to summarize  
&#96;&#96;&#96;  
Summarize this architecture document in bullet points:  
- Key decisions  
- Critical components  
- Dependencies  
- Constraints  

[paste full doc]  
&#96;&#96;&#96;  

**Step 2:** Use summary in other tools  
&#96;&#96;&#96;  
Cursor Composer: Implement auth based on this design:  

Summary:  
- JWT tokens in httpOnly cookies  
- Refresh token rotation  
- PostgreSQL for user storage  
- bcrypt password hashing  

[add specific implementation details only where needed]  
&#96;&#96;&#96;  

**Result:** Efficient context passing without token waste ✅  

\#### 4. Template Library  

**Create Reusable Prompt Templates:**  

**Template: Feature Planning (Claude)**  
&#96;&#96;&#96;markdown  
\# Feature Planning Template  

Context:  
- Project: [PROJECT_NAME]  
- Tech Stack: [STACK]  
- Feature: [FEATURE_DESCRIPTION]  

Design:  
1. Architecture overview  
2. Database schema changes  
3. API endpoints needed  
4. Frontend components  
5. Testing strategy  
6. Security considerations  
7. Performance implications  

Output Format:  
- File structure tree  
- Code snippets for each file  
- Dependencies to install  
- Configuration changes  
&#96;&#96;&#96;  

**Template: Code Review (Claude)**  
&#96;&#96;&#96;markdown  
\# Code Review Template  

Context: [FEATURE_NAME] implementation  

Files Changed:  
1. [FILE_PATH] - [PURPOSE]  
2. [FILE_PATH] - [PURPOSE]  

Review Focus:  
- ⚠️ Security vulnerabilities  
- ⚡ Performance bottlenecks  
- 📏 Code quality &amp; best practices  
- 🔄 Maintainability concerns  
- ✅ Test coverage gaps  

Provide:  
- Priority ranking (🔴 Critical → 🟡 Nice-to-have)  
- Specific line numbers  
- Fix suggestions with code  
&#96;&#96;&#96;  

**Template: Research Query (Perplexity)**  
&#96;&#96;&#96;markdown  
\# Research Query Template  

Topic: [TOPIC]  

Questions:  
1. What are current best practices for [TOPIC] in 2024?  
2. What are common pitfalls and how to avoid them?  
3. What are the top recommended libraries/tools?  
4. What do industry experts recommend?  
5. Are there any security/performance considerations?  

Format:  
- Bullet points  
- Include sources (with credibility assessment)  
- Comparison tables where applicable  
&#96;&#96;&#96;  

**Store templates in:** &#96;~/ai-workflows/templates/&#96;  

**Benefits:**  
- ✅ Consistent prompt quality  
- ✅ Faster workflow execution  
- ✅ Easier onboarding for team  
- ✅ Reproducible results  

---  

\## 5. Automation &amp; Scripts  

\### 🤖 CLI Tools for Workflow Automation  

\#### Script 1: Multi-Tool Context Exporter  

**Purpose:** Export Cursor project context for Claude/ChatGPT  

**File:** &#96;~/scripts/export-context.sh&#96;  

&#96;&#96;&#96;bash  
\#!/bin/bash  
\# Export project context for AI tools  

PROJECT_DIR=$1  
OUTPUT_FILE="${PROJECT_DIR}/ai-context.md"  

echo "# Project Context Export" &gt; $OUTPUT_FILE  
echo "Generated: $(date)" &gt;&gt; $OUTPUT_FILE  
echo "" &gt;&gt; $OUTPUT_FILE  

\# File structure  
echo "## File Structure" &gt;&gt; $OUTPUT_FILE  
tree -L 3 -I 'node_modules&#124;dist&#124;build' $PROJECT_DIR &gt;&gt; $OUTPUT_FILE  
echo "" &gt;&gt; $OUTPUT_FILE  

\# Key files content  
echo "## Key Files" &gt;&gt; $OUTPUT_FILE  
for file in package.json README.md tsconfig.json; do  
 if [ -f "${PROJECT_DIR}/${file}" ]; then  
 echo "### ${file}" &gt;&gt; $OUTPUT_FILE  
 echo "\&#96;\&#96;\&#96;" &gt;&gt; $OUTPUT_FILE  
 cat "${PROJECT_DIR}/${file}" &gt;&gt; $OUTPUT_FILE  
 echo "\&#96;\&#96;\&#96;" &gt;&gt; $OUTPUT_FILE  
 echo "" &gt;&gt; $OUTPUT_FILE  
 fi  
done  

\# Recent git changes  
echo "## Recent Changes" &gt;&gt; $OUTPUT_FILE  
cd $PROJECT_DIR  
git log --oneline -n 10 &gt;&gt; $OUTPUT_FILE  

echo "✅ Context exported to ${OUTPUT_FILE}"  
echo "📋 Ready to paste into Claude/ChatGPT"  
&#96;&#96;&#96;  

**Usage:**  
&#96;&#96;&#96;bash  
./export-context.sh ~/projects/my-nextjs-app  
\# Outputs: ~/projects/my-nextjs-app/ai-context.md  
\# Paste into Claude for architecture discussions  
&#96;&#96;&#96;  

\#### Script 2: Claude Review Automation  

**Purpose:** Send code changes to Claude for automated review  

**File:** &#96;~/scripts/claude-review.sh&#96;  

&#96;&#96;&#96;bash  
\#!/bin/bash  
\# Automated code review with Claude API  

\# Get changed files since last commit  
CHANGED_FILES=$(git diff --name-only HEAD~1 HEAD)  

\# Build review prompt  
PROMPT="Review these changes:  

Files Changed:  
${CHANGED_FILES}  

Changes:  
"  

\# Append file diffs  
for file in $CHANGED_FILES; do  
 PROMPT="${PROMPT}  
--- ${file} ---  
$(git diff HEAD~1 HEAD -- $file)  
"  
done  

PROMPT="${PROMPT}  

Check for:  
- Security issues  
- Performance problems  
- Best practices violations  
- Missing tests"  

\# Call Claude API (requires anthropic CLI or API key)  
echo "$PROMPT" &#124; claude-api review  

\# Or copy to clipboard for manual review  
echo "$PROMPT" &#124; pbcopy # macOS  
echo "📋 Review prompt copied to clipboard. Paste into Claude."  
&#96;&#96;&#96;  

**Usage:**  
&#96;&#96;&#96;bash  
git commit -m "Add auth feature"  
./claude-review.sh # Auto-generates review prompt  
&#96;&#96;&#96;  

\#### Script 3: Workflow Session Manager  

**Purpose:** Track AI tool usage and switch contexts efficiently  

**File:** &#96;~/scripts/ai-session.sh&#96;  

&#96;&#96;&#96;bash  
\#!/bin/bash  
\# AI Workflow Session Manager  

ACTION=$1  
TOOL=$2  

case $ACTION in  
 start)  
 # Create new session  
 SESSION_ID="$(date +%Y%m%d-%H%M%S)-${TOOL}"  
 mkdir -p ~/ai-sessions/$SESSION_ID  
 echo "📝 Started $TOOL session: $SESSION_ID"  
 echo "$SESSION_ID" &gt; ~/.current-ai-session  

 # Open tool  
 case $TOOL in  
 cursor) open -a "Cursor" ;;  
 claude) open "https://claude.ai" ;;  
 chatgpt) open "https://chat.openai.com" ;;  
 perplexity) open "https://perplexity.ai" ;;  
 esac  
 ;;  

 log)  
 # Log current context  
 SESSION_ID=$(cat ~/.current-ai-session)  
 MESSAGE=$3  
 echo "[$(date +%H:%M:%S)] $MESSAGE" &gt;&gt; ~/ai-sessions/$SESSION_ID/log.txt  
 echo "✅ Logged to session $SESSION_ID"  
 ;;  

 export)  
 # Export session notes  
 SESSION_ID=$(cat ~/.current-ai-session)  
 cat ~/ai-sessions/$SESSION_ID/log.txt  
 ;;  

 switch)  
 # Switch to different tool  
 ./ai-session.sh start $TOOL  
 ;;  
esac  
&#96;&#96;&#96;  

**Usage:**  
&#96;&#96;&#96;bash  
\# Start Cursor session  
./ai-session.sh start cursor  

\# Log what you're doing  
./ai-session.sh log "Implementing auth with Cursor Composer"  

\# Switch to Claude for review  
./ai-session.sh switch claude  
./ai-session.sh log "Reviewing auth implementation"  

\# Export session timeline  
./ai-session.sh export  
&#96;&#96;&#96;  

**Output:**  
&#96;&#96;&#96;  
[14:30:15] Implementing auth with Cursor Composer  
[14:55:22] Reviewing auth implementation  
[15:10:45] Fixed security issues found in review  
&#96;&#96;&#96;  

\#### Script 4: Template Prompt Generator  

**Purpose:** Quick access to prompt templates  

**File:** &#96;~/scripts/ai-prompt.sh&#96;  

&#96;&#96;&#96;bash  
\#!/bin/bash  
\# Generate prompts from templates  

TEMPLATE=$1  

TEMPLATES_DIR=~/ai-workflows/templates  

case $TEMPLATE in  
 plan)  
 cat $TEMPLATES_DIR/feature-planning.md &#124; pbcopy  
 echo "📋 Feature planning template copied"  
 ;;  
 review)  
 cat $TEMPLATES_DIR/code-review.md &#124; pbcopy  
 echo "📋 Code review template copied"  
 ;;  
 research)  
 cat $TEMPLATES_DIR/research-query.md &#124; pbcopy  
 echo "📋 Research template copied"  
 ;;  
 bug)  
 cat $TEMPLATES_DIR/bug-debugging.md &#124; pbcopy  
 echo "📋 Debugging template copied"  
 ;;  
 list)  
 echo "Available templates:"  
 ls -1 $TEMPLATES_DIR &#124; sed 's/.md//'  
 ;;  
esac  
&#96;&#96;&#96;  

**Usage:**  
&#96;&#96;&#96;bash  
./ai-prompt.sh plan # Copy planning template  
\# Paste into Claude, fill in [PLACEHOLDERS]  

./ai-prompt.sh review # Copy review template  
\# Paste into Claude with code  

./ai-prompt.sh list # See all templates  
&#96;&#96;&#96;  

---  

\### ⌨️ Keyboard Shortcuts Setup  

\#### Alfred/Raycast Workflows (macOS)  

**Workflow 1: Quick AI Tool Switcher**  

**Trigger:** &#96;ai [tool]&#96;  

**Actions:**  
- &#96;ai cursor&#96; → Opens Cursor + runs session script  
- &#96;ai claude&#96; → Opens Claude + copies context  
- &#96;ai chat&#96; → Opens ChatGPT  
- &#96;ai plex&#96; → Opens Perplexity  

**Implementation (Alfred):**  
&#96;&#96;&#96;bash  
\# Alfred Script Filter  
query=$1  
case $query in  
 cursor) open -a "Cursor" &amp;&amp; ./ai-session.sh start cursor ;;  
 claude) open "https://claude.ai" &amp;&amp; ./export-context.sh . ;;  
 chat) open "https://chat.openai.com" ;;  
 plex) open "https://perplexity.ai" ;;  
esac  
&#96;&#96;&#96;  

**Workflow 2: Context Exporter**  

**Trigger:** &#96;context&#96;  

**Action:** Exports current project context to clipboard  

&#96;&#96;&#96;bash  
./export-context.sh $(pwd)  
cat ai-context.md &#124; pbcopy  
echo "📋 Context copied. Ready for Claude/ChatGPT."  
&#96;&#96;&#96;  

**Workflow 3: Template Inserter**  

**Trigger:** &#96;prompt [template-name]&#96;  

**Actions:**  
- &#96;prompt plan&#96; → Copies planning template  
- &#96;prompt review&#96; → Copies review template  
- &#96;prompt research&#96; → Copies research template  

---  

\### 🔌 Browser Extensions for Workflow  

\#### Extension 1: AI Context Manager  

**Purpose:** Save webpage content directly to AI tools  

**Features:**  
- Right-click → "Send to Claude"  
- Right-click → "Research with Perplexity"  
- Save code snippets with context  
- Export conversation history  

**DIY with Bookmarklets:**  

&#96;&#96;&#96;javascript  
// Send selection to Claude  
javascript:(function(){  
 const text = window.getSelection().toString();  
 const url = window.location.href;  
 const context = &#96;From: ${url}  

${text}&#96;;  
 navigator.clipboard.writeText(context);  
 alert('📋 Context copied! Paste into Claude.');  
})();  
&#96;&#96;&#96;  

**Save as bookmark, click when selecting text**  

\#### Extension 2: Code Snippet Manager  

**Purpose:** Build reusable prompt library  

**Recommended:** Espanso (text expander)  

**Config:** &#96;~/Library/Application Support/espanso/match/ai.yml&#96;  

&#96;&#96;&#96;yaml  
matches:  
 # Quick Claude prompts  
 - trigger: ":plan"  
 replace: "Design architecture for [FEATURE]:  

Requirements:  
-   

Tech Stack:  
-   

Considerations:  
- "  

 - trigger: ":review"  
 replace: "Review this code:  

[PASTE CODE]  

Check for:  
- Security  
- Performance  
- Best practices  
- Test coverage"  

 - trigger: ":research"  
 replace: "Research best practices for [TOPIC] in 2024:  
- Current approaches  
- Common pitfalls  
- Recommended tools  
- Expert opinions"  

 # Quick file exports  
 - trigger: ":export"  
 replace: "{{output}}"  
 vars:  
 - name: output  
 type: shell  
 params:  
 cmd: "cat ai-context.md"  
&#96;&#96;&#96;  

**Usage:**  
- Type &#96;:plan&#96; → Auto-expands to planning template  
- Type &#96;:review&#96; → Auto-expands to review template  
- Type &#96;:export&#96; → Inserts project context  

---  

\## 6. Měření Efektivity  

\### 📊 KPI Tracking Framework  

\#### Metrics to Track  

&#124; Metric &#124; Before AI &#124; Target &#124; Measurement Method &#124;  
&#124;--------&#124;-----------&#124;--------&#124;-------------------&#124;  
&#124; Feature development time &#124; 8 hours &#124; 5 hours &#124; Time tracking per feature &#124;  
&#124; Bug fix time &#124; 2 hours &#124; 45 min &#124; Ticket resolution time &#124;  
&#124; Code review time &#124; 1 hour &#124; 15 min &#124; PR review duration &#124;  
&#124; Documentation time &#124; 30 min &#124; 10 min &#124; Docs commit timestamps &#124;  
&#124; Testing coverage &#124; 60% &#124; 85% &#124; Jest coverage reports &#124;  
&#124; Production bugs &#124; 5/month &#124; 2/month &#124; Error monitoring &#124;  
&#124; Developer satisfaction &#124; 6/10 &#124; 9/10 &#124; Weekly self-assessment &#124;  

\#### Tracking Template (Weekly)  

**File:** &#96;~/ai-workflows/metrics/week-[DATE].md&#96;  

&#96;&#96;&#96;markdown  
\# AI Workflow Metrics - Week of [DATE]  

\## Time Savings  

&#124; Task &#124; Traditional Time &#124; AI-Assisted Time &#124; Savings &#124; Tool Used &#124;  
&#124;------&#124;-----------------&#124;------------------&#124;---------&#124;-----------&#124;  
&#124; Auth feature &#124; 8h &#124; 5h &#124; 3h (37%) &#124; Cursor + Claude &#124;  
&#124; Payment bug &#124; 2h &#124; 45min &#124; 1h 15min (62%) &#124; Cursor + Perplexity &#124;  
&#124; Refactor UserProfile &#124; 6h &#124; 2h &#124; 4h (67%) &#124; Cursor Composer &#124;  
&#124; API docs &#124; 45min &#124; 10min &#124; 35min (78%) &#124; ChatGPT &#124;  

**Total Savings:** 8h 50min over 20h work week = 44% time saved  

\## Quality Improvements  

- ✅ Zero security issues (Claude review)  
- ✅ Test coverage 87% (up from 65%)  
- ✅ Zero production bugs  
- ✅ All PRs reviewed in &lt;30min  

\## Tool Usage  

- **Cursor:** 15h (code implementation)  
- **Claude:** 3h (planning, review)  
- **ChatGPT:** 1h (documentation)  
- **Perplexity:** 1h (research)  

\## ROI Calculation  

**Cost:** $30/mo  
**Time Saved:** ~35h/month (8.75h/week * 4)  
**Hourly Rate:** $50 (example)  
**Value:** $1,750/month  
**ROI:** 5,733% 🚀  

\## Learnings This Week  

- Cursor Composer is amazing for multi-file refactoring  
- Claude's architecture planning saves hours of debugging later  
- Perplexity is faster than Google for technical research  
- ChatGPT docs are good enough for first draft  

\## Next Week Goals  

- Try ensemble method (Claude + ChatGPT architecture)  
- Build automated code review pipeline  
- Create more prompt templates  
&#96;&#96;&#96;  

\#### Monthly Dashboard  

**Track Long-Term Trends:**  

&#96;&#96;&#96;markdown  
\# Monthly AI Workflow Performance  

\## Time Savings Trend  

&#124; Month &#124; Hours Saved &#124; Efficiency Gain &#124; Top Tool &#124;  
&#124;-------&#124;-------------&#124;-----------------&#124;----------&#124;  
&#124; Jan 2024 &#124; 20h &#124; 25% &#124; Cursor &#124;  
&#124; Feb 2024 &#124; 35h &#124; 44% &#124; Cursor + Claude &#124;  
&#124; Mar 2024 &#124; 45h &#124; 56% &#124; Full workflow &#124;  

\## Quality Metrics  

&#124; Month &#124; Test Coverage &#124; Production Bugs &#124; Code Review Time &#124;  
&#124;-------&#124;--------------&#124;-----------------&#124;------------------&#124;  
&#124; Jan &#124; 65% &#124; 5 &#124; 60min &#124;  
&#124; Feb &#124; 78% &#124; 3 &#124; 30min &#124;  
&#124; Mar &#124; 87% &#124; 1 &#124; 15min &#124;  

\## ROI  

- **Total Investment:** $90 (3 months @ $30/mo)  
- **Total Time Saved:** 100 hours  
- **Total Value:** $5,000 (@ $50/hour)  
- **Lifetime ROI:** 5,455%  
&#96;&#96;&#96;  

---  

\## 7. Common Pitfalls &amp; Solutions  

\### ❌ Pitfall 1: Context Loss When Switching Tools  

**Problem:** Lose track of decisions when moving from Claude → Cursor → ChatGPT  

**Symptoms:**  
- Implementing different architecture than planned  
- Forgetting edge cases discussed in research  
- Inconsistent coding style  

**Solutions:**  

✅ **Solution 1: Central Documentation Hub**  
- Save ALL AI outputs to Notion/Obsidian  
- Reference docs in every prompt  
- Example: "Based on architecture doc: [link]"  

✅ **Solution 2: Session Logging**  
- Use &#96;ai-session.sh&#96; script to log each transition  
- Export session timeline at end of feature  
- Review timeline before moving to next task  

✅ **Solution 3: Explicit Context Handoff**  

**Bad Transition:**  
&#96;&#96;&#96;  
Claude: [designs auth system]  
→ Switch to Cursor  
Cursor: "Implement JWT auth" ❌ Missing context  
&#96;&#96;&#96;  

**Good Transition:**  
&#96;&#96;&#96;  
Claude: [designs auth system, saves to docs/auth-design.md]  
→ Switch to Cursor  
Cursor: "Implement auth based on: [paste auth-design.md content]" ✅  
&#96;&#96;&#96;  

---  

\### ❌ Pitfall 2: Over-Reliance on Single Tool  

**Problem:** Using only Cursor for everything (planning, coding, review, docs)  

**Why It's Bad:**  
- Cursor is optimized for code generation, not architecture planning  
- Claude is better at security review  
- ChatGPT is faster for documentation  
- Missing tool-specific strengths  

**Solutions:**  

✅ **Solution: Task-Specific Tool Selection**  

**Decision Tree:**  

&#96;&#96;&#96;  
Task Type → Best Tool  

Planning/Architecture → Claude (reasoning strength)  
Research → Perplexity (cited sources)  
Code Implementation → Cursor (multi-file, context)  
Code Generation (small) → Copilot (speed)  
Security Review → Claude (thoroughness)  
Documentation → ChatGPT (speed, quality)  
Debugging → Cursor Chat (project context)  
&#96;&#96;&#96;  

**Track Your Usage:**  
&#96;&#96;&#96;markdown  
\# This Week's Tool Distribution  

❌ BAD: Cursor 95%, Claude 5%, Others 0%  
✅ GOOD: Cursor 60%, Claude 25%, ChatGPT 10%, Perplexity 5%  
&#96;&#96;&#96;  

---  

\### ❌ Pitfall 3: Inconsistent Outputs (Different Tools, Different Styles)  

**Problem:** Claude suggests one architecture, Cursor implements another  

**Example:**  
&#96;&#96;&#96;  
Claude: "Use repository pattern with dependency injection"  
Cursor: [generates direct database queries in controllers] ❌  
&#96;&#96;&#96;  

**Solutions:**  

✅ **Solution 1: Explicit Style Guide in Every Prompt**  

**Create:** &#96;CODING_STANDARDS.md&#96;  

&#96;&#96;&#96;markdown  
\# Project Coding Standards  

\## Architecture Patterns  
- Repository pattern for data access  
- Dependency injection for services  
- API routes in /api directory  
- Services in /services directory  

\## Code Style  
- TypeScript strict mode  
- ESLint + Prettier  
- Functional components (React)  
- Async/await (no .then())  

\## Testing  
- Jest + React Testing Library  
- 80%+ coverage target  
- Test files next to source files  
&#96;&#96;&#96;  

**Then in EVERY Cursor prompt:**  
&#96;&#96;&#96;  
Implement [feature] following project standards:  

[paste CODING_STANDARDS.md]  

Ensure:  
- Repository pattern  
- TypeScript strict  
- 80%+ test coverage  
&#96;&#96;&#96;  

✅ **Solution 2: Ensemble Verification**  

**Process:**  
1. Claude designs architecture  
2. Cursor implements  
3. **Claude reviews implementation** against original design  
4. Fix inconsistencies  

**Claude Review Prompt:**  
&#96;&#96;&#96;  
Compare this implementation against the original design:  

Original Design (from earlier conversation):  
[paste Claude's architecture doc]  

Actual Implementation:  
[paste Cursor's code]  

Check:  
- Does implementation match design?  
- Any deviations?  
- Are deviations improvements or mistakes?  
&#96;&#96;&#96;  

---  

\### ❌ Pitfall 4: Poor Handoff Management (Information Loss)  

**Problem:** Context lost between tools  

**Example:**  
&#96;&#96;&#96;  
Perplexity: "Best practice is to use httpOnly cookies for JWT"  
→ Switch to Cursor  
Cursor: [implements localStorage JWT] ❌ Lost research finding  
&#96;&#96;&#96;  

**Solutions:**  

✅ **Solution: Structured Handoff Format**  

**Perplexity Research Output (structured):**  
&#96;&#96;&#96;markdown  
\# JWT Authentication Research  

\## Key Findings  

\### Security Best Practices  
1. ✅ Use httpOnly cookies (prevents XSS)  
2. ✅ Implement refresh token rotation  
3. ✅ Set short access token expiry (15min)  
4. ❌ NEVER use localStorage (XSS vulnerable)  

\### Implementation Requirements  
- Cookie: httpOnly, Secure, SameSite=Strict  
- Access token: 15min expiry  
- Refresh token: 7 day expiry, rotate on use  

[Save to: docs/auth-research.md]  
&#96;&#96;&#96;  

**Cursor Prompt (references research):**  
&#96;&#96;&#96;  
Implement JWT auth following research findings:  

[paste auth-research.md]  

Ensure:  
- httpOnly cookies (NOT localStorage)  
- 15min access token expiry  
- Refresh token rotation  
&#96;&#96;&#96;  

---  

\### ❌ Pitfall 5: Inefficient Context Sharing (Token Waste)  

**Problem:** Pasting entire codebase into Claude/ChatGPT  

**Example:**  
&#96;&#96;&#96;  
Prompt: "Review my app"  
[pastes 50 files, 10,000 lines] ❌  
- Hits token limit  
- Expensive  
- Unfocused review  
&#96;&#96;&#96;  

**Solutions:**  

✅ **Solution: Targeted Context Sharing**  

**Bad Review Request:**  
&#96;&#96;&#96;  
Review my entire app:  
[50 files pasted]  
&#96;&#96;&#96;  

**Good Review Request:**  
&#96;&#96;&#96;  
Review JWT auth security:  

Key Files (300 lines total):  
1. authController.js - Login/register logic  
2. authMiddleware.js - Token validation  
3. jwtHelper.js - Token generation  

Focus on:  
- SQL injection risks  
- XSS vulnerabilities  
- Token security  

[paste only these 3 files, ~300 lines]  
&#96;&#96;&#96;  

**Result:**  
- ❌ Bad: 10,000 tokens, generic review  
- ✅ Good: 500 tokens, focused, actionable review  

---  

\## 8. Advanced Patterns  

\### 🎯 Pattern 1: Ensemble Methods (Combining Outputs)  

**When to Use:** Critical decisions, security reviews, complex architecture  

**Process:**  

\#### Step 1: Get Multiple Perspectives  

**Prompt Same Question to 2-3 Tools:**  

**Claude:**  
&#96;&#96;&#96;  
Design authentication system for Next.js app:  
- Requirements: [details]  
- Security: Enterprise-grade  
- Scale: 100k users  
&#96;&#96;&#96;  

**ChatGPT:**  
&#96;&#96;&#96;  
Design authentication system for Next.js app:  
- Requirements: [same details]  
- Security: Enterprise-grade  
- Scale: 100k users  
&#96;&#96;&#96;  

**Perplexity:**  
&#96;&#96;&#96;  
What are best practices for Next.js authentication at scale?  
- Enterprise security requirements  
- 100k users  
- Include sources  
&#96;&#96;&#96;  

\#### Step 2: Compare &amp; Synthesize  

**Create Comparison Matrix:**  

&#124; Aspect &#124; Claude Recommendation &#124; ChatGPT Recommendation &#124; Perplexity Sources &#124;  
&#124;--------&#124;---------------------&#124;----------------------&#124;-------------------&#124;  
&#124; Token Storage &#124; httpOnly cookies &#124; httpOnly cookies &#124; ✅ Industry consensus &#124;  
&#124; Session Management &#124; Redis &#124; Database &#124; ✅ Redis for scale &#124;  
&#124; Refresh Strategy &#124; Rotation &#124; Rotation &#124; ✅ OWASP recommended &#124;  
&#124; **Difference** &#124; Recommends Redis &#124; Suggests DB first &#124; - &#124;  

**Decision:**  
- Agree on: httpOnly cookies, token rotation ✅  
- Investigate: Redis vs Database for sessions  
- Final choice: Redis (better for 100k scale, backed by sources)  

\#### Step 3: Validation Review  

**Send Final Design Back to Claude:**  
&#96;&#96;&#96;  
I synthesized these three recommendations:  
[paste comparison matrix]  

My final decision:  
- httpOnly cookies ✅  
- Redis sessions ✅  
- Token rotation ✅  

Does this design have any critical flaws I missed?  
&#96;&#96;&#96;  

**Result:** High-confidence architecture with multiple validations ✅  

---  

\### 🎯 Pattern 2: Verification Workflows (Cross-Checking)  

**When to Use:** Security-critical code, production deployments, refactoring  

**Multi-Stage Verification:**  

\#### Stage 1: Implementation (Cursor)  

&#96;&#96;&#96;  
Implement payment processing:  
- Stripe integration  
- Webhook handling  
- Idempotency  
- Error handling  
&#96;&#96;&#96;  

**Output:** Payment processing code  

\#### Stage 2: Security Review (Claude)  

&#96;&#96;&#96;  
Security audit this payment code:  

[paste payment processing code]  

Check for:  
- API key exposure  
- Webhook signature verification  
- Idempotency key handling  
- Error message leakage  
- SQL injection  
- XSS risks  
&#96;&#96;&#96;  

**Output:** Security findings list  

\#### Stage 3: Cross-Verification (ChatGPT)  

&#96;&#96;&#96;  
Second opinion on these security findings:  

Original Code:  
[paste code]  

Claude's Findings:  
[paste findings]  

Questions:  
1. Are these findings valid?  
2. Any critical issues Claude missed?  
3. Any false positives?  
&#96;&#96;&#96;  

**Output:** Validated security findings  

\#### Stage 4: Fix &amp; Re-Review (Cursor + Claude)  

&#96;&#96;&#96;  
Cursor: Implement fixes from validated findings  
Claude: Review fixes to confirm issues resolved  
&#96;&#96;&#96;  

**Result:** Defense-in-depth security validation ✅  

---  

\### 🎯 Pattern 3: Iterative Refinement Loops  

**When to Use:** Complex features, performance optimization, UX improvements  

**Loop Structure:**  

\#### Iteration 1: MVP Implementation  

**Cursor:**  
&#96;&#96;&#96;  
Implement basic search feature:  
- Full-text search  
- Display results  
- Basic relevance ranking  
&#96;&#96;&#96;  

**Output:** Basic search working ✅  

\#### Iteration 2: Performance Review  

**Claude:**  
&#96;&#96;&#96;  
Review search performance:  

[paste implementation]  

Optimize for:  
- 10k+ records  
- &lt;100ms response time  
- Efficient queries  
&#96;&#96;&#96;  

**Claude Suggestions:**  
- Add database index on search column  
- Implement query debouncing  
- Add pagination  
- Use database FTS instead of LIKE queries  

\#### Iteration 3: Apply Optimizations  

**Cursor:**  
&#96;&#96;&#96;  
Apply performance optimizations:  

[paste Claude's suggestions]  

Implement:  
- PostgreSQL full-text search indexes  
- Debouncing (300ms)  
- Pagination (50 results/page)  
&#96;&#96;&#96;  

**Output:** Fast search ✅  

\#### Iteration 4: UX Review  

**ChatGPT:**  
&#96;&#96;&#96;  
Review search UX:  

Current features:  
- Search input with debouncing  
- Paginated results  
- Relevance ranking  

Suggest improvements for:  
- User feedback (loading states)  
- Error handling  
- Empty states  
- Keyboard navigation  
&#96;&#96;&#96;  

**ChatGPT Suggestions:**  
- Loading spinner  
- "No results" message  
- Highlight search terms in results  
- Arrow key navigation  

\#### Iteration 5: Final Polish  

**Cursor:**  
&#96;&#96;&#96;  
Add UX improvements:  
[paste ChatGPT's suggestions]  
&#96;&#96;&#96;  

**Output:** Polished, fast, user-friendly search ✅  

**Result:** Iterative refinement from MVP → production-ready feature  

---  

\### 🎯 Pattern 4: Parallel Processing (Simultaneous Tasks)  

**When to Use:** Independent features, batch operations, research tasks  

**Scenario:** Implement 3 independent features simultaneously  

**Parallel Track 1: Authentication**  
&#96;&#96;&#96;  
Cursor Session 1: Implement JWT auth  
Claude Session 1: Review auth security  
Timeline: Hours 0-2  
&#96;&#96;&#96;  

**Parallel Track 2: Payment Processing**  
&#96;&#96;&#96;  
Cursor Session 2: Integrate Stripe  
Perplexity Session 1: Research Stripe best practices  
Timeline: Hours 0-2 (simultaneous)  
&#96;&#96;&#96;  

**Parallel Track 3: Documentation**  
&#96;&#96;&#96;  
ChatGPT Session 1: Generate API docs  
Timeline: Hours 0-2 (simultaneous)  
&#96;&#96;&#96;  

**Convergence Point (Hour 3):**  
&#96;&#96;&#96;  
Claude: Review all 3 features together  
- Integration points  
- Consistency check  
- Combined testing strategy  
&#96;&#96;&#96;  

**Result:** 3 features delivered in 3 hours instead of 9 hours sequential ✅  

---  

\### 🎯 Pattern 5: Fallback Strategies (When Tools Fail)  

**When to Use:** AI hallucinations, incorrect outputs, context limitations  

**Multi-Level Fallback Chain:**  

\#### Level 1: Primary Tool (Cursor)  

&#96;&#96;&#96;  
Task: Implement WebSocket real-time chat  
Tool: Cursor Composer  
&#96;&#96;&#96;  

**If Cursor generates broken code or hallucinates libraries:**  

\#### Level 2: Alternative Tool (Claude)  

&#96;&#96;&#96;  
Claude: Cursor generated this WebSocket code but it's not working:  

[paste Cursor's code]  

Error:  
[paste error message]  

Please debug and provide correct implementation.  
&#96;&#96;&#96;  

**If Claude can't solve it:**  

\#### Level 3: Research + Manual (Perplexity + Docs)  

&#96;&#96;&#96;  
Perplexity: How to implement WebSocket chat in Next.js 14?  
- Include working code examples  
- Common errors and solutions  
- Best libraries to use  
&#96;&#96;&#96;  

**Then manually implement based on research**  

\#### Level 4: Community + Docs  

&#96;&#96;&#96;  
- Check official Next.js docs  
- Search GitHub issues  
- Stack Overflow  
- Ask in Discord/Slack communities  
&#96;&#96;&#96;  

**Result:** Always have a path forward when AI fails ✅  

---  

\## 9. 30-Day Implementation Roadmap  

\### Week 1: Foundation Setup  

**Day 1-2: Tool Setup**  
- ✅ Install Cursor Pro ($20/mo)  
- ✅ Set up Claude Free account  
- ✅ Set up ChatGPT Free account  
- ✅ Set up Perplexity Free account  
- ✅ Install Alfred/Raycast  
- ✅ Install Espanso for text expansion  

**Day 3-4: Workflow Scripts**  
- ✅ Create &#96;export-context.sh&#96;  
- ✅ Create &#96;ai-session.sh&#96;  
- ✅ Create &#96;ai-prompt.sh&#96;  
- ✅ Set up Alfred workflows  
- ✅ Configure Espanso snippets  

**Day 5-7: Template Library**  
- ✅ Create feature planning template  
- ✅ Create code review template  
- ✅ Create research template  
- ✅ Create debugging template  
- ✅ Create &#96;CODING_STANDARDS.md&#96;  

**Metrics:** Track time spent on setup tasks (baseline)  

---  

\### Week 2: Basic Workflow Practice  

**Day 8-10: Single-Tool Mastery**  
- Practice Cursor Composer (implement 2-3 small features)  
- Practice Claude planning (design 2-3 features)  
- Practice ChatGPT docs (generate docs for existing features)  

**Day 11-14: Two-Tool Workflows**  
- Claude (plan) → Cursor (implement)  
- Perplexity (research) → Cursor (implement)  
- Cursor (code) → Claude (review)  

**Metrics:**  
- Track time per task (compare to historical baseline)  
- Note friction points  
- Document what works well  

---  

\### Week 3: Multi-Tool Integration  

**Day 15-17: Three-Tool Workflows**  
- Perplexity (research) → Claude (design) → Cursor (implement)  
- Cursor (implement) → Claude (review) → ChatGPT (document)  

**Day 18-21: Full Workflow Practice**  
- Implement complete feature using all tools  
- Track handoffs between tools  
- Measure end-to-end time savings  

**Metrics:**  
- Time savings vs baseline  
- Quality improvements (bugs, test coverage)  
- Subjective satisfaction rating  

---  

\### Week 4: Optimization &amp; Automation  

**Day 22-24: Workflow Refinement**  
- Identify bottlenecks  
- Optimize slow transitions  
- Add more automation scripts  
- Refine templates based on usage  

**Day 25-28: Advanced Patterns**  
- Try ensemble method on critical decision  
- Practice verification workflow on security code  
- Use iterative refinement on complex feature  

**Day 29-30: Measurement &amp; Iteration**  
- Calculate Week 4 metrics  
- Compare to Week 1 baseline  
- Document learnings  
- Plan next month improvements  

**Expected Results:**  
- 30-50% time savings by Week 4  
- Higher code quality (fewer bugs)  
- Better documentation  
- Higher confidence in outputs  

---  

\## 10. Quick Reference Cheat Sheet  

\### 🎯 Decision Tree: Which Tool When?  

&#96;&#96;&#96;  
Task Type → Recommended Tool  

Planning/Architecture → Claude (reasoning strength)  
Research &amp; Best Practices → Perplexity (cited sources)  
Code Implementation → Cursor (multi-file context)  
Quick Code Snippets → Copilot (speed)  
Security Review → Claude (thoroughness)  
Documentation → ChatGPT (speed)  
Debugging → Cursor Chat (project context)  
Refactoring → Cursor Composer (multi-file)  
Test Generation → Cursor + Claude (coverage + edge cases)  
Performance Optimization → Claude (analysis) + Cursor (implementation)  
&#96;&#96;&#96;  

\### ⚡ Quick Commands  

&#96;&#96;&#96;bash  
\# Export project context  
./export-context.sh ~/projects/my-app  

\# Start new AI session  
./ai-session.sh start cursor  

\# Copy template prompt  
./ai-prompt.sh plan  

\# Switch tools  
./ai-session.sh switch claude  
&#96;&#96;&#96;  

\### 📋 Prompt Templates  

**Feature Planning (Claude):**  
&#96;&#96;&#96;  
Design [FEATURE] for [PROJECT]:  

Requirements:  
- [REQ 1]  
- [REQ 2]  

Tech Stack: [STACK]  

Provide:  
- Architecture overview  
- File structure  
- Implementation code  
- Dependencies  
- Testing strategy  
&#96;&#96;&#96;  

**Code Review (Claude):**  
&#96;&#96;&#96;  
Review [FILES]:  

[PASTE CODE]  

Check:  
- Security vulnerabilities  
- Performance issues  
- Best practices  
- Test coverage gaps  
&#96;&#96;&#96;  

**Research (Perplexity):**  
&#96;&#96;&#96;  
Best practices for [TOPIC] in 2024:  
- Current approaches  
- Common pitfalls  
- Recommended tools  
- Expert sources  
&#96;&#96;&#96;  

---  

\## Závěr &amp; Next Steps  

\### ✅ Immediate Actions (This Week)  

1. **Set Up Tool Stack:**  
 - Install Cursor Pro ($20/mo)  
 - Create Claude Free account  
 - Set up ChatGPT Free + Perplexity Free  

2. **Create Workflow Scripts:**  
 - Download &#96;export-context.sh&#96;, &#96;ai-session.sh&#96;, &#96;ai-prompt.sh&#96;  
 - Set up Alfred/Raycast workflows  
 - Configure Espanso text snippets  

3. **First Feature With New Workflow:**  
 - Pick small feature (2-4 hour task)  
 - Follow Pattern 1 (Feature Implementation)  
 - Track time and results  

\### 📊 Success Metrics to Track  

**Weekly Dashboard:**  
&#96;&#96;&#96;markdown  
&#124; Metric &#124; Baseline &#124; Week 1 &#124; Week 4 &#124; Target &#124;  
&#124;--------&#124;----------&#124;--------&#124;--------&#124;--------&#124;  
&#124; Feature Dev Time &#124; 8h &#124; 7h &#124; 5h &#124; 5h &#124;  
&#124; Bug Fix Time &#124; 2h &#124; 1.5h &#124; 45min &#124; 45min &#124;  
&#124; Code Review Time &#124; 1h &#124; 45min &#124; 15min &#124; 15min &#124;  
&#124; Test Coverage &#124; 60% &#124; 70% &#124; 85% &#124; 85% &#124;  
&#124; Production Bugs &#124; 5/mo &#124; 4/mo &#124; 2/mo &#124; 2/mo &#124;  
&#124; Satisfaction &#124; 6/10 &#124; 7/10 &#124; 9/10 &#124; 9/10 &#124;  
&#96;&#96;&#96;  

\### 🎯 30-Day Goals  

- **Week 1:** Tool setup + basic workflows  
- **Week 2:** Two-tool workflow mastery  
- **Week 3:** Multi-tool integration practice  
- **Week 4:** 40%+ time savings, high-quality outputs  

\### 🚀 Long-Term Vision  

**3 Months:**  
- Fully automated workflow  
- 50%+ productivity gain  
- Custom integrations and scripts  
- Team workflow templates  

**6 Months:**  
- Advanced automation (CI/CD integration)  
- Ensemble methods for critical code  
- Comprehensive prompt library  
- Measurable ROI documentation  

---  

\## 🎁 Bonus: Tool Comparison Matrix  

&#124; Feature &#124; Cursor &#124; Claude &#124; ChatGPT &#124; Perplexity &#124; Copilot &#124;  
&#124;---------&#124;--------&#124;--------&#124;---------&#124;------------&#124;---------&#124;  
&#124; **Code Generation** &#124; ⭐⭐⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐ &#124; ⭐⭐⭐⭐⭐ &#124;  
&#124; **Multi-File Editing** &#124; ⭐⭐⭐⭐⭐ &#124; ⭐ &#124; ⭐ &#124; ⭐ &#124; ⭐⭐ &#124;  
&#124; **Architecture Design** &#124; ⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124; ⭐⭐⭐⭐ &#124; ⭐⭐ &#124; ⭐ &#124;  
&#124; **Security Review** &#124; ⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐ &#124; ⭐⭐ &#124;  
&#124; **Research** &#124; ⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124; ⭐ &#124;  
&#124; **Documentation** &#124; ⭐⭐ &#124; ⭐⭐⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124; ⭐⭐ &#124; ⭐⭐⭐ &#124;  
&#124; **Speed** &#124; ⭐⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124;  
&#124; **Context Window** &#124; ⭐⭐⭐⭐ &#124; ⭐⭐⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐⭐ &#124; ⭐⭐⭐ &#124;  
&#124; **Cost** &#124; $20/mo &#124; $20/mo &#124; Free &#124; Free &#124; $10/mo &#124;  

**Best Combinations:**  

1. **Budget: $30/mo**  
 - Cursor Pro ($20) + Claude Free + Perplexity Free  
 - Best for: Full-stack development  

2. **Budget: $40/mo**  
 - Cursor Pro ($20) + Claude Pro ($20)  
 - Best for: Complex architecture + security-critical apps  

3. **Budget: $50/mo**  
 - Cursor Pro ($20) + Claude Pro ($20) + ChatGPT Plus ($20)  
 - Best for: Maximum productivity across all tasks  

**Your Recommended Stack: Cursor Pro + Claude Free ($20/mo total)**  
- Covers 90% of your needs  
- Leaves $10 in budget for occasional ChatGPT Plus month  
- Claude Free's 100 msgs/day enough for 4h/day work  

---  

**Good luck with your new AI-powered workflow! 🚀 Remember: The goal isn't to use the most expensive tools, but to use the RIGHT tool for each task.** 💡

### Kde použít tento prompt?

Najděte vhodné AI nástroje pro použití tohoto promptu a maximalizujte jeho efektivitu.

 [ 🔧 Vhodné nástroje ](https://zacinamsai.cz/tools?category=produktivita) [ 📖 Návody ](https://zacinamsai.cz/articles?search=Ostatn%C3%AD) [ 📚 Slovník ](https://zacinamsai.cz/glossary?search=Produktivita) 

### Podobné prompty

 [  

  Expert  

####  AI Tool Limitations &amp; Workarounds - Překonávání Omezení 

Praktický průvodce známými omezeními AI nástrojů (hallucinations, context limits, reasoning failures) a efektivními workaroundy pro produktivní práci i při těchto omezeních.

  Produktivita  Zobrazit  

 ](https://zacinamsai.cz/prompts/ai-tool-limitations-workarounds-prekonavani-omezeni) [  

  Začátečník  

####  Analýza konkurence přes Google Search 

Tento prompt pro Tactiq efektivně využije data ze SerpApi k analýze výsledků vyhledávání pro zvolené klíčové slovo a připraví strategický souhrn.

  Produktivita  Zobrazit  

 ](https://zacinamsai.cz/prompts/tactiq-analyza-konkurence-pres-google-search) [  

  Expert  

####  Asynchronous communication framework 

Implementuj async-first komunikační strategie pro remote work a deep work protection.

  Produktivita  Zobrazit  

 ](https://zacinamsai.cz/prompts/async-communication-framework) [  

  Expert  

####  Automatický plán projektu v Motion 

Komplexní instrukce pro AI nástroj Motion, který má autonomně naplánovat a rozvrhnout složitý pracovní projekt na základě priorit, termínů a vašich preferencí.

  Produktivita  Zobrazit  

 ](https://zacinamsai.cz/prompts/motion-automaticky-plan-projektu-v-motion) 

### Objevte další AI prompty

 Prozkoumejte naši sbírku Produktivita promptů a najděte ty, které vám pomohou dosáhnout lepších výsledků.

 [ Další Produktivita prompty ](https://zacinamsai.cz/prompts?category=productivity) [ Všechny prompty ](https://zacinamsai.cz/prompts)

```json
{
        "@context": "https://schema.org",
        "@type": "WebSite",
        "name": "Začínám s AI",
        "alternateName": "zacinamsai.cz",
        "url": "https://zacinamsai.cz",
        "description": "Průvodce světem umělé inteligence - Katalog AI nástrojů, články, prompty a slovník pojmů",
        "inLanguage": "cs",
        "potentialAction": {
            "@type": "SearchAction",
            "target": {
                "@type": "EntryPoint",
                "urlTemplate": "https://zacinamsai.cz/search?q={search_term_string}"
            },
            "query-input": {
                "@type": "PropertyValueSpecification",
                "valueRequired": true,
                "valueName": "search_term_string"
            }
        },
        "publisher": {
            "@type": "Organization",
            "name": "Začínám s AI",
            "url": "https://zacinamsai.cz",
            "logo": {
                "@type": "ImageObject",
                "url": "https://zacinamsai.cz/images/zacinamsai_logo.png"
            }
        }
    }
```
