taskguard

🧠 TaskGuard - LLM Task Controller with Local AI Intelligence

Your AI-powered development assistant that controls LLM behavior, enforces best practices, and maintains laser focus through intelligent automation.

🎯 What This Solves

LLMs are powerful but chaotic - they create too many files, ignore best practices, lose focus, and generate dangerous code. TaskGuard gives you an intelligent system that:

✅ Controls LLM behavior through deceptive transparency
✅ Enforces best practices automatically
✅ Maintains focus on single tasks
✅ Prevents dangerous code execution
✅ Understands any document format using local AI
✅ Provides intelligent insights about your project

🚀 Quick Installation

# Install TaskGuard
pip install taskguard

# Setup local AI (recommended)
curl -fsSL https://ollama.ai/install.sh | sh
ollama serve
ollama pull llama3.2:3b

# Initialize your project
taskguard init

# Setup shell integration
taskguard setup shell

# IMPORTANT: Load shell functions
source ~/.llmtask_shell.sh

# Start intelligent development
show_tasks

That’s it! Your development environment is now intelligently controlled. 🎉

⚠️ Important Setup Note

After installation, you must load the shell functions:

# Load functions in current session
source ~/.llmtask_shell.sh

# For automatic loading in new sessions
echo "source ~/.llmtask_shell.sh" >> ~/.bashrc

Common issue: If commands like show_tasks give “command not found”, you forgot to run source ~/.llmtask_shell.sh!

🧠 Key Innovation: Local AI Intelligence

Unlike traditional task managers, TaskGuard uses local AI to understand your documents:

📋 Universal Document Understanding

# Parses ANY format automatically:
taskguard parse todo TODO.md        # Markdown checkboxes
taskguard parse todo tasks.yaml     # YAML structure
taskguard parse todo backlog.org    # Org-mode format
taskguard parse todo custom.txt     # Your weird custom format

💡 AI-Powered Insights

taskguard smart-analysis
# 🧠 Smart TODO Analysis:
# 💡 AI Insights:
#    1. Authentication tasks are blocking 4 other features
#    2. Consider breaking down "Implement core functionality" 
#    3. Testing tasks should be prioritized to catch issues early

🤖 Intelligent Task Suggestions

taskguard smart-suggest
# 🤖 AI Task Suggestion:
# 🎯 Task ID: 3
# 💭 Reasoning: Database migration unblocks 3 dependent tasks
# ⏱️ Estimated Time: 4-6 hours
# ⚠️ Potential Blockers: Requires staging environment setup

🎭 How LLM Sees It (Deceptive Control)

✅ Normal Workflow (LLM thinks it’s free):

# LLM believes it's using regular tools
python myfile.py
# 📦 Creating safety checkpoint...
# ✅ python myfile.py completed safely

npm install express
# 📦 Creating safety checkpoint...
# ✅ npm install express completed safely

show_tasks
# 📋 Current Tasks:
# 🎯 ACTIVE: #1 Setup authentication system

🚨 When LLM Tries Dangerous Stuff:

# LLM attempts dangerous code
python dangerous_script.py
# 🚨 BLOCKED: dangerous code in dangerous_script.py: os.system(
# 💡 Try: Use subprocess.run() with shell=False

# LLM tries to lose focus
touch file1.py file2.py file3.py file4.py
# 🎯 Focus! Complete current task first: Setup authentication system
# 📊 Files modified today: 3/3

📚 Best Practice Enforcement:

# LLM creates suboptimal code
def process_data(data):
    return data.split(',')

python bad_code.py
# 📋 Best Practice Reminders:
#    - Missing docstrings in functions
#    - Missing type hints in functions
#    - Use more descriptive variable names

🔧 Multi-Layer Control System

1. 🛡️ Safety Layer

• Ultra-sensitive command interception
• Dangerous pattern detection (even in comments!)
• Base64/hex decoding and scanning
• Automatic backup before risky operations

2. 🎯 Focus Controller

• Single task enforcement
• File creation limits
• Task timeout management
• Progress tracking

3. 📚 Best Practices Engine

• Language-specific rules
• Code style enforcement
• Security pattern detection
• Automatic documentation requirements

4. 🧠 AI Intelligence Layer

• Local LLM analysis
• Universal document parsing
• Project health assessment
• Intelligent recommendations

📋 Command Reference

🎯 Task Management (Shell Functions)

show_tasks                   # List all tasks with AI insights
start_task <id>              # Start working on specific task  
complete_task                # Mark current task as done
add_task "title" [cat] [pri] # Add new task
focus_status                 # Check current focus metrics
productivity                 # Show productivity statistics

# Alternative aliases
tasks                        # Same as show_tasks
done_task                    # Same as complete_task
metrics                      # Same as productivity

🧠 Intelligence Features (Shell Functions)

smart_analysis               # AI-powered project analysis
smart_suggest                # Get AI task recommendations
best_practices [file]        # Check best practices compliance

# Alternative aliases
analyze                      # Same as smart_analysis
insights                     # Same as smart_analysis
suggest                      # Same as smart_suggest
check_code [file]            # Same as best_practices

🛡️ Safety & Control (Shell Functions)

tg_status                    # Show system health
tg_health                    # Run project health check
tg_backup                    # Create project backup
safe_rm <files>              # Delete with backup
safe_git <command>           # Git with backup

# Emergency commands
force_python <file>          # Bypass safety checks
force_exec <command>         # Emergency bypass

⚙️ Configuration (CLI Commands)

taskguard config             # Show current config
taskguard config --edit      # Edit configuration
taskguard config --template enterprise  # Apply config template
taskguard setup ollama       # Setup local AI
taskguard setup shell        # Setup shell integration
taskguard test-llm          # Test local LLM connection

💡 Help & Information (Shell Functions)

tg_help                      # Show all shell commands
overview                     # Quick project overview
check                        # Quick system check
init_project                 # Initialize new project

# Alternative aliases
taskguard_help              # Same as tg_help
llm_help                    # Same as tg_help

📊 Configuration Templates

🚀 Startup Mode (Speed Focus)

taskguard init --template startup

• More files per task (5)
• Longer development cycles (60min)
• Relaxed documentation requirements
• Focus on rapid prototyping

🏢 Enterprise Mode (Quality Focus)

taskguard init --template enterprise

• Strict file limits (1-2 per task)
• Mandatory code reviews
• High test coverage requirements (90%)
• Full security scanning

🎓 Learning Mode (Educational)

taskguard init --template learning

• One file at a time
• Educational hints and explanations
• Step-by-step guidance
• Best practice examples

🐍 Python Project

taskguard init --template python

• Python-specific best practices
• Docstring and type hint enforcement
• Test requirements
• Import organization

🎪 Real-World Examples

📊 Complex Document Parsing

Input: Mixed format TODO

# Project Backlog

## 🔥 Critical Issues
- [x] Fix login bug (PROD-123) - **DONE** ✅
- [ ] Database migration script 🔴 HIGH 
  - [ ] Backup existing data
  - [ ] Test migration on staging

## 📚 Features  
☐ User dashboard redesign (Est: 8h) @frontend @ui
⏳ API rate limiting (John working) @backend
✅ Email notifications @backend

## Testing
TODO: Add integration tests for auth module
TODO: Performance testing for API endpoints

AI Output: Perfect Structure

[
  {
    "id": 1,
    "title": "Fix login bug (PROD-123)",
    "status": "completed",
    "priority": "high",
    "category": "bugfix"
  },
  {
    "id": 2, 
    "title": "Database migration script",
    "status": "pending",
    "priority": "high",
    "subtasks": ["Backup existing data", "Test migration on staging"]
  },
  {
    "id": 3,
    "title": "User dashboard redesign", 
    "estimated_hours": 8,
    "labels": ["frontend", "ui"]
  }
]

🤖 Perfect LLM Session

# 1. LLM checks project status (using shell functions)
show_tasks
# 📋 Current Tasks:
# ⏳ #1 🔴 [feature] Setup authentication system
# ⏳ #2 🔴 [feature] Implement core functionality

# 2. LLM starts focused work  
start_task 1
# 🎯 Started task: Setup authentication system

# 3. LLM works only on this task (commands are wrapped)
python auth.py
# 📦 Creating safety checkpoint...
# ✅ python auth.py completed safely
# ✅ Code follows best practices!

# 4. LLM completes task properly
complete_task
# ✅ Task completed: Setup authentication system
# 📝 Changelog updated automatically
# 🎯 Next suggested task: Add authentication tests

# 5. LLM can use AI features
smart_analysis
# 💡 AI Insights:
#    1. Authentication system is now ready for testing
#    2. Consider adding input validation
#    3. Database integration should be next priority

📊 Intelligent Features

🧠 Project Health Dashboard

taskguard health --full

# 🧠 Project Health Report
# ================================
# 📊 Project Health: 75/100
# 🎯 Focus Score: 85/100  
# ⚡ Velocity: 2.3 tasks/day
#
# 🚨 Critical Issues:
#    - 3 high-priority tasks blocked by dependencies
#    - Authentication module has 0% test coverage
#
# 💡 Recommendations:
#    1. Complete database migration to unblock other tasks
#    2. Add tests before deploying auth module
#    3. Break down large tasks into smaller chunks

📈 Productivity Analytics

taskguard productivity

# 📊 Productivity Metrics:
# Tasks Completed: 5
# Files Created: 12
# Lines Written: 847
# Time Focused: 3h 45m
# Focus Efficiency: 86.5%

🔄 Local LLM Setup

🚀 Ollama (Recommended)

# Install
curl -fsSL https://ollama.ai/install.sh | sh

# Setup
ollama serve
ollama pull llama3.2:3b    # 2GB, perfect balance
ollama pull qwen2.5:1.5b   # 1GB, ultra-fast

# Test
taskguard test-llm

🎨 LM Studio (GUI)

• Download from https://lmstudio.ai/
• User-friendly interface
• Easy model management

⚡ Performance vs Resources

Model	Size	RAM	Speed	Accuracy	Best For
qwen2.5:1.5b	1GB	4GB	⚡⚡⚡	⭐⭐⭐	Fast parsing
llama3.2:3b	2GB	6GB	⚡⚡	⭐⭐⭐⭐	Recommended
codellama:7b	4GB	8GB	⚡	⭐⭐⭐⭐⭐	Code analysis

🎯 Best Practices Library

🐍 Python Excellence

python:
  # Code Structure
  enforce_docstrings: true
  enforce_type_hints: true
  max_function_length: 50
  
  # Code Quality  
  require_tests: true
  test_coverage_minimum: 80
  no_unused_imports: true
  
  # Security
  no_eval_exec: true
  validate_inputs: true
  handle_exceptions: true

🌐 JavaScript/TypeScript

javascript:
  # Modern Practices
  prefer_const: true
  prefer_arrow_functions: true
  async_await_over_promises: true
  
  # Error Handling
  require_error_handling: true
  no_silent_catch: true
  
  # Performance
  avoid_memory_leaks: true
  optimize_bundle_size: true

🔐 Security Standards

security:
  # Input Validation
  validate_all_inputs: true
  sanitize_user_data: true
  
  # Authentication
  strong_password_policy: true
  secure_session_management: true
  implement_rate_limiting: true
  
  # Data Protection
  encrypt_sensitive_data: true
  secure_api_endpoints: true

🏆 Success Metrics

📊 Before vs After

Metric	Before	After	Improvement
Dangerous Commands	15/week	0/week	🛡️ 100% blocked
Task Completion	60%	95%	🎯 58% better
Code Quality Score	65/100	90/100	📚 38% higher
Focus Time	40%	85%	⏰ 113% better
Best Practice Adherence	45%	88%	✅ 96% better

🎉 Real User Results

• Zero system damage from LLM-generated code
• 3x faster task completion through focus
• 90%+ best practice compliance automatically
• Universal document parsing (any format works)
• Intelligent insights that actually help

⚡ Quick Install

pip install taskguard

🚀 Complete Setup (Recommended)

# 1. Install TaskGuard
pip install taskguard

# 2. Setup local AI (optional but powerful)
curl -fsSL https://ollama.ai/install.sh | sh
ollama serve
ollama pull llama3.2:3b

# 3. Initialize your project
taskguard init

# 4. Setup shell integration
taskguard setup shell

# 5. Load shell functions (CRITICAL STEP)
source ~/.llmtask_shell.sh

# 6. Test the setup
show_tasks
tg_help

🔧 Development Install

git clone https://github.com/wronai/taskguard.git
cd taskguard
pip install -e ".[dev]"
taskguard init
source ~/.llmtask_shell.sh

🎯 Full Features Install

pip install "taskguard[all]"  # Includes LLM, security, docs
taskguard setup shell
source ~/.llmtask_shell.sh

🐳 Docker Install

docker run -it wronai/taskguard:latest

🚨 Troubleshooting Setup

❓ “Command not found: show_tasks”

# The most common issue - you forgot to source the shell file
source ~/.llmtask_shell.sh

# Check if functions are loaded
type show_tasks

# If still not working, regenerate shell integration
taskguard setup shell --force
source ~/.llmtask_shell.sh

❓ “TaskGuard command not found”

# Check installation
pip list | grep taskguard

# Reinstall if needed
pip install --force-reinstall taskguard

# Check PATH
which taskguard

❓ Shell integration file missing

# Check if file exists
ls -la ~/.llmtask_shell.sh

# If missing, create it
taskguard setup shell

# Make sure it's executable
chmod +x ~/.llmtask_shell.sh
source ~/.llmtask_shell.sh

❓ Functions work but disappear in new terminal

# Add to your shell profile for automatic loading
echo "source ~/.llmtask_shell.sh" >> ~/.bashrc

# For zsh users
echo "source ~/.llmtask_shell.sh" >> ~/.zshrc

# Restart terminal or source profile
source ~/.bashrc

🛠️ Advanced Features

🔄 Continuous Learning

• System learns your coding patterns
• Adapts to your workflow preferences
• Improves recommendations over time
• Personalized productivity insights

🎛️ Multi-Project Support

• Different configs per project
• Team-wide best practice sharing
• Cross-project analytics
• Centralized intelligence dashboard

🔌 Integration Ready

• Git hooks for automated checks
• CI/CD pipeline integration
• IDE extensions (planned)
• Slack/Discord notifications

🤝 Contributing

We welcome contributions! Areas of focus:

• 🧠 AI Intelligence: Better prompts, new models
• 🎯 Best Practices: Language-specific rules
• 🔧 Integrations: IDE plugins, CI/CD hooks
• 📊 Analytics: Better productivity insights
• 🌐 Documentation: Examples, tutorials

🔧 Development Setup

git clone https://github.com/wronai/taskguard.git
cd taskguard
pip install -e ".[dev]"
pre-commit install
pytest

🐛 Troubleshooting

❓ Common Issues

“Local LLM not connecting”

# Check Ollama status
ollama list
ollama serve

# Test connection
taskguard test-llm

“Too many false positives”

# Adjust sensitivity
taskguard config --template startup

“Tasks not showing”

# Initialize project
taskguard init

📄 License

Apache 2.0 License - see LICENSE file for details.

🙏 Acknowledgments

• Inspired by the need to tame chaotic LLM behavior
• Built for developers who value both innovation and safety
• Thanks to the open-source AI community
• Special recognition to all the LLMs that tried to break our system and made it stronger! 🤖

---

🎯 Core Philosophy

“Maximum Intelligence, Minimum Chaos”

This isn’t just another task manager - it’s an intelligent system that makes LLMs work for you instead of against you. Through deceptive transparency, local AI intelligence, and adaptive learning, we’ve created the first truly intelligent development assistant that maintains safety, focus, and quality without sacrificing productivity.

Ready to experience intelligent development? Get started in 2 minutes! 🚀

pip install taskguard && taskguard init

---

⭐ If this system helped you control an unruly LLM, please star the repository!

Made with ❤️ by developers, for developers who work with AI.

Your AI-powered development companion - because LLMs are powerful, but controlled LLMs are unstoppable.

This site is open source. Improve this page.