ClawGuard

A security-first Email → Sanitized Event → Agent pipeline that prevents prompt injection and unsafe content from ever reaching the LLM.

All external content is untrusted. Agents never see raw input.

Live at claw-guard.tech — closed beta, contact support@qubemc.com for access.

Problem

LLM agents are vulnerable to prompt injection, tool call manipulation, hidden HTML/script payloads, secret exfiltration attempts, malicious attachments, and markdown-based instruction hijacking. Most systems pass raw inbound data directly to the agent.

ClawGuard is a defensive middleware layer that sits between inbound email and your agent runtime.

How It Works

Gmail Inbox
  → OAuth Connect (per-user, encrypted)
  → Auto-Fetch (cron, every minute)
  → Deterministic Sanitization
  → Risk Scoring & Injection Detection
  → Structured Safe Payload
  → Agent (via API key)

Sanitization Guarantees

• No raw HTML reaches the agent
• No scripts or hidden elements
• Prompt injection patterns flagged and redacted
• API keys, tokens, and passwords redacted
• Attachments filtered by allowlist
• Content length-limited and normalized
• Risk metadata (0–100 score) included with every payload

Features

• JWT + API Key Auth — Closed beta with admin-managed users. API keys (cg_ prefix) for agent/automation access.
• Per-User Gmail OAuth — Each user connects their own Gmail(s). Tokens encrypted at rest with Fernet.
• Auto-Fetch — Cron-driven polling fetches new emails every minute with deduplication.
• Dashboard — Inbox viewer, API key management, Gmail account management, admin panel.
• Security Demo — Public demo at /demo with 8 attack vectors, no login required.
• Agent Skill — Bundled clawguard-skill for OpenClaw/agent integration with query scripts.

Tech Stack

• Python / FastAPI — API server
• SQLite — Storage (users, API keys, Gmail tokens, sanitized events)
• bcrypt + PyJWT — Auth
• Fernet (cryptography) — Gmail token encryption
• nginx + Let's Encrypt — HTTPS reverse proxy
• Deterministic regex pipeline — Injection detection and sanitization

Getting Started

# Create virtual environment
python -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Set up .env (required)
cat > .env << 'EOF'
CLAWGUARD_ADMIN_EMAIL=admin@example.com
CLAWGUARD_ADMIN_PASSWORD=your-secure-password
CLAWGUARD_JWT_SECRET=your-jwt-secret
CLAWGUARD_ENCRYPTION_KEY=your-fernet-key
EOF

# Run the server
uvicorn clawguard.main:app --host 0.0.0.0 --port 8000

API Endpoints

Endpoint	Auth	Description
POST /auth/login	None	Login (email + password → JWT)
GET /api/events	Bearer	List sanitized emails (filterable)
GET /api/events/risky	Bearer	List risky emails by score
GET /api/events/{id}	Bearer	Get single email detail
GET /api/accounts	Bearer	List connected inboxes
GET /api/senders	Bearer	List senders with counts
GET /api/stats	None	Inbox statistics
GET /api/timeline	Bearer	Daily email volume trends
POST /api/keys	Bearer	Create API key
POST /gmail/fetch	Bearer	Manual Gmail fetch
GET /demo/sanitize	None	Public security demo
GET /health	None	Health check

Deployment

# Deploy to server (pull + restart)
bash deploy.sh

Requires: nginx, systemd service, Let's Encrypt SSL, .env on server.

License

MIT