Save premium requests on your AI coding IDEs. This simple prompt instructions helps your AI work better and use less premium requests and give the agent new instructions while its currently working.

TaskSync is an autonomous agent protocol that creates persistent agent. Instead of typing lots of messages back and forth, you write tasks in a tasks.txt file. Your AI agent continuously monitors this file, executes tasks autonomously, and maintains persistent operation until manually terminated.

Tasks monitoring - continuously checks your tasks.txt file every 60 seconds to 5 minutes for new tasks.

Dual file system - AI uses tasks.txt for instructions and separate log.txt for status tracking

Real-time status logging - AI writes progress monitoring into dedicated log.txt with count-based monitoring

Never terminates automatically - maintains persistent operation until you explicitly stop it

Self-correcting behavior - when AI makes mistakes, it reads your corrections and fixes its mistakes

Works with any AI IDEs - GitHub Copilot, Cursor, Windsurf, Trae IDE, and more

1. Drag the tasksync instructions to chat and ask the agent to follow the tasksync.
2. Add tasks in your tasks.txt file
3. Write what you want - it checks tasks.txt for updates automatically
4. Change tasks.txt anytime to follow next instructions or make it fix its mistakes.

Start saving money today. Get the better results with way fewer premium requests.

Click install or copy-paste the installation commands for other IDEs:

Create tasks.txt and log.txt file inside instructions folder. Add your tasks in tasks.txt.

git clone --filter=blob:none --sparse https://github.com/4regab/TaskSync.git
cd TaskSync
git sparse-checkout set .cursor

git clone --filter=blob:none --sparse https://github.com/4regab/TaskSync.git
cd TaskSync
git sparse-checkout set .global

• GitHub Copilot (VS Code) - .github/instructions/ setup for maximum premium usage
• Cursor IDE - Modern .cursor/rules/*.mdc setup
• Global - global_rules.md for any IDE

• Infinite Monitoring: AI never terminates automatically - operates continuously until manually stopped
• Dual File System: AI uses tasks.txt for instructions and separate log.txt for status tracking
• Status Logging: AI writes check counts directly into dedicated log.txt file with each monitoring cycle
• Count-Based Monitoring: Systematic counting from Check #1 incrementing indefinitely
• File Editing Protocol: Mandatory physical file editing with each monitoring check
• Complete File Reading: Always reads entire files (minimum 1000 lines) for comprehensive analysis
• Real-Time Communication: Edit tasks.txt anytime to communicate with AI during execution
• Autonomous Execution: Independent task completion with persistent operation
• State Management: Active → Monitoring → Manual Termination Only

Real-time task communication with separate log file - edit tasks.txt anytime:

# Current Priority
Fix the authentication bug in login.tsx
Add TypeScript types for user profile

# New Feature Request  
Create a dashboard component with charts

# Quick Corrections
The button color should be blue, not red
Use const instead of let in the helper functions

Issue: AI models tend to end conversations quickly, especially after completing tasks.

Solution: Continuously add new tasks to tasks.txt before the AI finishes its current work to maintain persistent operation.

Best Practices:

• Queue multiple tasks in tasks.txt from the beginning
• Add new tasks while AI is working on current ones
• Use the STATUS LOG to monitor AI progress and add tasks proactively
• Keep a backlog of improvements, optimizations, or additional features ready

Example of Continuous Task Management:

# Current Task
Fix authentication bug in login.tsx

# Queued Tasks (add these before current task completes)
Add TypeScript types for user profile
Implement password reset functionality
Add unit tests for authentication
Optimize login page performance
Add accessibility improvements

Monitoring Tip: Watch the STATUS LOG check numbers - if they stop incrementing, the AI may have ended the session despite the infinite monitoring protocol.

See CONTRIBUTING.md for development setup, coding standards, and submission guidelines.

• Complete Protocol
• Usage Examples
• Implementation Guide
• Contributing Guidelines