Personal AI Employee Hackathon 0 – Building Autonomous FTEs in 2026

An autonomous email processing system that monitors Gmail, applies intelligent priority scoring, extracts tasks, creates action plans, and flags sensitive operations for human approval using the Ralph Wiggum pattern (ACT → INFORM → WAIT).

The AI Employee Vault autonomously:

• 📧 Monitors your Gmail inbox for important/starred emails
• 🎯 Scores priority (HIGH/MEDIUM/LOW/IGNORE) based on keywords
• 📝 Extracts tasks, deadlines, and people mentioned
• 📋 Creates actionable plans with checklists
• ✍️ Drafts professional reply emails
• ⚠️ Flags financial transactions for human approval
• 📊 Updates a live dashboard with activity logs
• ✅ Follows strict rules defined in Company Handbook

No emails are sent automatically - all replies require human approval (Bronze Tier safety feature).

graph TB
    A[Gmail Inbox] -->|OAuth| B[Gmail Watcher]
    B -->|Poll every 5 min| C{New Email?}
    C -->|Yes| D[Needs_Action/]
    D --> E[AI Agent]
    E --> F[Priority Scorer]
    F --> G[Task Extractor]
    G --> H[Reply Drafter]
    H --> I{Financial?}
    I -->|Yes| J[Flag HITL]
    I -->|No| K[Create Plan]
    J --> K
    K --> L[Update Dashboard]
    L --> M[Move to Done/]
    M --> N[Human Reviews]

• ✅ Gmail Watcher - OAuth authentication, polls every 5 minutes
• ✅ Priority Scoring - Keyword-based urgency detection
• ✅ Task Extraction - Automatic action item identification
• ✅ Reply Drafting - Professional email responses
• ✅ HITL Protection - Financial operations require approval
• ✅ Dashboard - Live system status and activity logs
• ✅ Ralph Wiggum Loop - ACT → INFORM → WAIT autonomy pattern
• ✅ Folder Workflow - Needs_Action → Plans → Done pipeline

• 🔒 OAuth 2.0 authentication (no password storage)
• 🔒 Local data storage only (no external APIs except Gmail)
• 🔒 Read-only Gmail scope (cannot send emails)
• 🔒 Automatic HITL flags for financial keywords
• 🔒 All sensitive files in .gitignore

• Python 3.13+
• Gmail account
• Google Cloud Project with Gmail API enabled

1. Clone the repository:

git clone https://github.com/AnamShergill/Hackathon_0-Bronze-Tier.git
cd Hackathon_0-Bronze-Tier

2. Install dependencies:

pip install google-api-python-client google-auth-oauthlib google-auth-httplib2 watchdog python-dotenv playwright pyyaml

3. Set up Gmail API:

   • Go to Google Cloud Console
   • Create a new project
   • Enable Gmail API
   • Create OAuth 2.0 credentials
   • Download as credentials.json and place in project root

4. Start the Gmail Watcher:

python Watchers/gmail_watcher.py

• Browser opens for OAuth (first run only)
• Polls Gmail every 5 minutes
• Creates .md files in Needs_Action/

5. Process emails:

python Skills/email_processor.py

• Applies priority scoring
• Extracts tasks and deadlines
• Creates plans in Plans/
• Drafts replies (requires approval)
• Moves processed emails to Done/

6. Check the dashboard:

cat Dashboard.md

AI_Employee_Vault/
├── Needs_Action/          # Incoming emails from watcher
├── Plans/                 # Action plans and draft replies
├── Done/                  # Processed emails with <COMPLETE> markers
├── Pending_Approval/      # Items awaiting human approval
├── Approved/              # Human-approved items
├── Skills/                # Agent skill definitions
│   ├── 01_EMAIL_PROCESSOR.md
│   ├── 02_EMAIL_REPLY_DRAFTER.md
│   ├── 03_TASK_EXTRACTOR.md
│   ├── 04_PRIORITY_SCORER.md
│   ├── 05_DASHBOARD_UPDATER.md
│   ├── 06_ARCHIVE_CLEANER.md
│   └── email_processor.py
├── Watchers/              # External system monitors
│   ├── gmail_watcher.py
│   └── base_watcher.py
├── Dashboard.md           # Live system status
├── Company_Handbook.md    # Rules of engagement
├── CLAUDE.md              # AI agent system prompt
├── demo_guide.md          # Comprehensive demo guide
├── TROUBLESHOOTING.md     # Debug guide
└── orchestrator.py        # Main coordination script

1. Send email to yourself with:

   • Subject: URGENT: Payment Due $500
   • Body: Please process this payment by tomorrow.

2. In Gmail:

   • DON'T open it (keep unread)
   • Click the star icon ⭐

3. Wait up to 5 minutes for watcher to detect

4. Run processor:

python Skills/email_processor.py

5. Check results:

ls Plans/          # Action plan created
ls Done/           # Email moved here
cat Dashboard.md   # Updated with activity

The system automatically assigns priorities:

Per Company_Handbook.md, these operations require human approval:

✅ Auto-approve:

• Reading emails
• Creating plans
• Updating dashboard

⚠️ Require approval:

• Financial transactions ($, €, £, payment, invoice)
• Sending emails with attachments
• Adding new recipients (CC/BCC)

Example HITL flag:

requires_hitl: true
hitl_reason: "Financial transaction ($500) requires human approval"

Check:

1. Email is unread (bold in Gmail)
2. Email is starred (yellow star icon)
3. Email is in Inbox (not Sent/Spam)
4. Email is recent (last 24 hours)

Debug:

python debug_gmail.py

See TROUBLESHOOTING.md for detailed guide.

• demo_guide.md - Comprehensive demo and usage guide
• TROUBLESHOOTING.md - Debug and fix common issues
• Company_Handbook.md - Rules and authorization levels
• CLAUDE.md - AI agent system prompt

• Structured markdown files with YAML frontmatter
• Ralph Wiggum pattern for clear autonomy boundaries
• Keyword-based priority scoring (no ML needed)
• Folder-based workflow (simple and debuggable)
• HITL flags for financial operations

• HTML email content extraction
• Gmail's 24-hour query limitation
• No automatic email sending (by design)
• Manual loop execution required

• Autonomy requires constraints (HITL approval builds trust)
• Transparency is critical (dashboard + activity logs)
• Simple > Complex (files easier than database)
• Security first (limited OAuth scopes)

• 📱 WhatsApp Business API watcher
• 📧 MCP email sender with approval workflow
• 🔄 Automatic loop trigger on new files
• 🧵 Conversation threading

• 🌐 Multi-channel (Slack, Discord, Telegram)
• 🧠 Advanced NLP (sentiment, entity extraction)
• 📅 Calendar integration
• 💳 Payment processing with HITL

• 🤝 Team collaboration
• 📱 Mobile app
• 🎤 Voice interface
• 🔮 Proactive suggestions

This is a hackathon project demonstrating Bronze Tier capabilities. Contributions welcome!

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

MIT License - See LICENSE file for details

Built for Personal AI Employee Hackathon 0 – Building Autonomous FTEs in 2026

Special thanks to:

• Google Gmail API team
• Python community
• Claude AI for development assistance

Anam Shergill

• GitHub: @AnamShergill
• Project: Hackathon_0-Bronze-Tier

Bronze Tier Complete ✅

• Folder structure ✓
• Core markdown files ✓
• Python project + dependencies ✓
• Gmail Watcher running ✓
• Agent skills created ✓
• Ralph Wiggum loop functional ✓

Ready for Silver Tier! 🚀

Last Updated: March 18, 2026
Version: Bronze Tier v1.0
Status: Production Ready