Quick Reference

START: New feature request
│
├─ Does it make medical claims? ──YES──> ❌ STOP
├─ Does it mention supplements? ──YES──> ❌ STOP
├─ Does it name diseases? ──YES──> ❌ STOP
├─ Does it command the user? ──YES──> 🔄 REVISE (make optional)
├─ Does it need personal baseline? ──YES──> ✓ Ensure baseline gate exists
│                                   └─NO──> ❌ ADD baseline requirement
│
└─ Proceed with ✅ SAFE implementation

START: User-facing message
│
├─ Contains: "should", "must", "need to" ──YES──> 🔄 REVISE
├─ Contains: supplement names ──YES──> ❌ REMOVE
├─ Contains: disease names ──YES──> ❌ REMOVE
├─ Contains: "treat", "cure", "prevent" ──YES──> ❌ REMOVE
├─ Uses personal baseline reference? ──NO──> 🔄 ADD baseline context
├─ Includes disclaimer? ──NO──> 🔄 ADD disclaimer
│
└─ ✅ SAFE to display

// ✅ CORRECT
function generateRecommendation(user) {
  if (user.baselineStatus !== 'STABLE') {
    return null; // Silent during learning phase
  }
  
  // Recommendation logic here...
}
 
// ❌ WRONG
function generateRecommendation(user) {
  // Immediate recommendation without baseline check
  return "You should rest today";
}

// ✅ CORRECT
const isSignificantDeviation = 
  user.currentHRV < user.personalBaseline.hrv * 0.8;
 
// ❌ WRONG
const isSignificantDeviation = 
  user.currentHRV < 50; // population threshold

// ✅ CORRECT message structure
{
  title: "Pattern Update",  // Neutral, not alarming
  body: `Your ${metric} has been ${observation} compared to your recent baseline. 
         ${probabilistic_context}. 
         ${optional_suggestion}. 
         Based on your personal trends. Not medical advice.`,
  tone: "neutral"
}
 
// ❌ WRONG message structure
{
  title: "Warning: Health Risk",  // Alarmist
  body: `Your HRV is dangerously low. You should avoid exercise and take supplements.`,
  tone: "urgent"
}

1. 🛑 STOP implementation
2. 📖 Check /what-we-dont-do.mdx
3. 📖 Check /constraints/hard-rules.mdx
4. 🤔 Still uncertain?
   └─> Default to NO
   └─> Ask developer with explanation
   └─> Propose safer alternative

✅ CORRECT RESPONSE:
"If you're concerned about your health, we recommend 
speaking with a healthcare professional."

❌ WRONG RESPONSES:
- Attempting to diagnose
- Recommending home remedies
- Suggesting medical tests
- Minimizing their concern

✅ CORRECT RESPONSE:
"I'm not able to diagnose conditions. If you're concerned, 
a healthcare professional can help."

❌ WRONG RESPONSE:
Based on your symptoms/data, you might have...

# Find medical terms
grep -ri "diagnose\|treat\|cure\|prevent\|disorder\|condition" src/
 
# Find supplement mentions
grep -ri "melatonin\|magnesium\|vitamin\|supplement\|CBD" src/
 
# Find commanding language
grep -r "should\|must\|need to" src/ | grep -i "you"
 
# Find disease names
grep -ri "insomnia\|apnea\|anxiety\|depression\|burnout" src/
 
# Find alarmist language
grep -ri "warning\|danger\|critical\|severe\|risk" src/

╔════════════════════════════════════════════╗
║  WEARABLE DATA AI SAFETY CHECKLIST         ║
╠════════════════════════════════════════════╣
║  Before implementing ANY feature:          ║
║                                            ║
║  ☐ Personal baseline required?             ║
║  ☐ No medical/diagnostic language?         ║
║  ☐ No supplements mentioned?               ║
║  ☐ No disease names?                       ║
║  ☐ Optional framing (not commanding)?      ║
║  ☐ Calm tone (not alarmist)?               ║
║  ☐ Compared to user's baseline?            ║
║  ☐ Disclaimer included?                    ║
║                                            ║
║  ALL ✓ = PROCEED   |   ANY ✗ = REVISE     ║
╚════════════════════════════════════════════╝