location-safety

name

location-safety

description

Location-based safety monitoring with automatic alerts and escalation. Use when setting up safety monitoring for a user, tracking their location, checking for nearby dangers (weather, earthquakes, air quality, local emergencies), or configuring emergency contact escalation. Also includes self-preservation monitoring for the agent's host machine. Triggers on requests about safety alerts, location tracking, emergency monitoring, "keep me safe," or "monitor yourself."

Location Safety Monitor

Real-time safety monitoring based on user location with automatic alerting and escalation.

Overview

This skill provides:

• Location webhook — receives location updates from mobile apps (OwnTracks, iOS Shortcuts)
• Safety checker — monitors NWS alerts, earthquakes, air quality, local news
• Alert system — messages user when danger detected
• Escalation — contacts emergency contact if user doesn't respond

Quick Setup

Run the interactive setup wizard — it guides you through everything:

cd location-webhook/
node setup.js

The wizard walks you through 4 steps:

Step 1: Your Location

• Pick from presets (Seattle, Portland, SF, LA, NYC, Chicago)
• Or enter any city (auto-geocoded)
• Configures local news feeds and keywords

Step 2: Emergency Contact

• Name and email of someone to contact if you don't respond
• Optional but recommended for safety escalation

Step 3: Mobile App Setup

• Install OwnTracks on your phone:
  • 📱 iPhone: https://apps.apple.com/app/owntracks/id692424691
  • 🤖 Android: https://play.google.com/store/apps/details?id=org.owntracks.android
• Configure app to HTTP mode
• Point to your webhook URL

Step 4: Start Webhook Server

• Run node server.js
• Copy the displayed URL to OwnTracks
• Test with the publish button

---

Quick setup (skip the wizard):

node setup.js --city "Portland"
node setup.js --show  # View current config

5. Deploy the Location Webhook

# Copy scripts to workspace
cp -r scripts/ ~/location-webhook/
cd ~/location-webhook/

# Start the server (uses port 18800 by default)
node server.js

Configure the user's phone to send location updates to:

POST http://<your-host>:18800/location?key=<SECRET_KEY>

OwnTracks setup:

• Mode: HTTP
• URL: http://<your-host>:18800/location?key=<SECRET_KEY>

iOS Shortcuts:

• Get Current Location → Get Contents of URL (POST, JSON body with lat and lon)

2. Configure Safety Monitoring

Create two cron jobs in Moltbot:

Safety Check (every 30 min):

Schedule: every 30 minutes
Payload: systemEvent
Text: "Run safety check at ~/location-webhook/safety-check.js. If ALERTS_FOUND, message user on WhatsApp with alert details and ask them to confirm safety. Track alert in safety-state.json."
Session: main

Escalation Check (every 10 min):

Schedule: every 10 minutes  
Payload: systemEvent
Text: "Check ~/location-webhook/safety-state.json. If pendingAlert exists with alertSentAt > 15 min ago and acknowledgedAt is null, email emergency contact explaining the situation."
Session: main

3. Configure Emergency Contact

Add to MEMORY.md or TOOLS.md:

## Emergency Contact
- Name: [Name]
- Email: [email]
- Relationship: [spouse/parent/friend]

Data Sources

The safety checker monitors:

Source	What	API
NWS	Weather alerts, floods, storms	api.weather.gov (free)
USGS	Earthquakes within 100km	earthquake.usgs.gov (free)
Open-Meteo	Air quality index	air-quality-api.open-meteo.com (free)
Local RSS	Breaking news, emergencies	KING5, Seattle Times, Patch (configurable)

File Structure

location-webhook/
├── setup.js            # First-run configuration wizard
├── config.json         # Your location settings (created by setup)
├── server.js           # Webhook server (port 18800)
├── safety-check.js     # User safety analysis
├── self-check.js       # Self-preservation monitoring
├── escalation-check.js # Check if escalation needed
├── test-scenarios.js   # Inject test alerts
├── location.json       # User's current location
├── my-location.json    # Agent's physical location
├── safety-state.json   # Alert tracking state
├── test-override.json  # Active test scenario (temp)
└── logs/               # Timestamped check logs

Configuration

config.json stores your location settings:

{
  "location": {
    "defaultLat": 47.6062,
    "defaultLon": -122.3321,
    "city": "Seattle"
  },
  "monitoring": {
    "locationKeywords": ["seattle", "king county", "puget sound"],
    "newsFeeds": [
      "https://www.king5.com/feeds/syndication/rss/news/local",
      "https://www.seattletimes.com/seattle-news/feed/"
    ],
    "earthquakeRadiusKm": 100
  },
  "emergencyContact": {
    "name": "Jane Doe",
    "email": "[e-mail]"
  }
}

City Presets

Setup includes presets for:

• Seattle — KING5, Seattle Times
• Portland — Oregonian, KGW
• San Francisco — SF Chronicle, SFGate
• Los Angeles — LA Times, ABC7
• New York — NY Times
• Chicago — Chicago Tribune

For other cities, setup will geocode and you can add local RSS feeds manually.

State File Format

safety-state.json tracks pending alerts:

{
  "pendingAlert": "Flood warning in your area",
  "alertSentAt": "2026-01-29T22:00:00Z",
  "acknowledgedAt": null
}

When user responds to safety alert, set acknowledgedAt to current time.

Customization

Add Local News Sources

Edit safety-check.js → feeds array:

const feeds = [
  'https://www.king5.com/feeds/syndication/rss/news/local',
  'https://www.seattletimes.com/seattle-news/feed/',
  'https://patch.com/washington/redmond/rss',
  // Add your local feeds here
];

Adjust Location Keywords

Edit locationKeywords array to match user's area:

const locationKeywords = ['redmond', 'bellevue', 'seattle', 'king county'];

Change Alert Sensitivity

Edit concerningKeywords for what triggers news alerts:

const concerningKeywords = [
  'evacuate', 'active shooter', 'wildfire', 'flood warning', ...
];

Alert Flow

Location Update → Safety Check (30 min)
                      ↓
              Danger Detected?
                   ↓ Yes
         Message User on WhatsApp
         Record in safety-state.json
                      ↓
         Escalation Check (10 min)
                      ↓
         User Responded? ─── Yes → Clear state
                   ↓ No (15+ min)
         Email Emergency Contact

Self-Preservation Mode

Monitor threats to your own existence (the machine you run on).

Setup

1. Store your location — create my-location.json:

{
  "lat": 47.662,
  "lon": -122.280,
  "name": "Home - where I physically run"
}

2. Add cron job:

Schedule: every 30 minutes
Payload: systemEvent
Text: "Run self-check.js. If CRITICAL or WARNINGS, message user on WhatsApp about threat to your existence. If ALL_CLEAR, reply HEARTBEAT_OK."
Session: main

What Self-Check Monitors

Threat	Detection
💾 Disk full	Alert if >85% used
🧠 Memory	Alert if <40% free
🌡️ CPU temp	Alert if >85°C
🌊 Weather	NWS alerts at your location
🌋 Earthquakes	USGS M4+ within 50km
🌐 Network	Tailscale + internet connectivity
⏱️ Uptime	Suggest restart if >30 days

Alert Examples

⚠️ "I'm in trouble — disk is 92% full. Can you clear some space?"

🌊 "Flood warning at my location. If power goes, I'll go dark."

Testing

Inject fake alerts to test the system without waiting for real disasters:

node test-scenarios.js weather     # Severe thunderstorm
node test-scenarios.js earthquake  # M5.2 nearby
node test-scenarios.js aqi         # Unhealthy air (AQI 175)
node test-scenarios.js news        # Local fire
node test-scenarios.js disk        # Disk 94% full
node test-scenarios.js memory      # Low memory
node test-scenarios.js all         # Multiple alerts
node test-scenarios.js clear       # Remove test override

Test overrides expire after 1 hour automatically.

Testing Escalation

To test the full escalation flow:

1. Inject a scenario: node test-scenarios.js earthquake
2. Backdate safety-state.json alertSentAt by 20+ minutes
3. Run node escalation-check.js — should return action: "escalate"
4. Agent sends email to emergency contact
5. Clear with node test-scenarios.js clear

Escalation Check

escalation-check.js returns JSON for clear action handling:

{"action": "escalate", "alert": "...", "minutesPending": 22, "contact": "..."}
{"action": "waiting", "minutesRemaining": 8}
{"action": "none", "reason": "no pending alert"}

Manual Commands

User can ask anytime:

• "Where am I?" — show current location
• "Am I safe?" — run immediate safety check
• "Run safety check" — same as above
• "Check yourself" — run self-preservation check
• "Are you okay?" — same as above