Add NAPSTER AI medical skills

skills/medical-entity-extractor/SKILL.md

@@ -0,0 +1,144 @@
+---
+name: medical-entity-extractor
+description: Extract medical entities (symptoms, medications, lab values, diagnoses) from patient messages.
+license: MIT
+metadata:
+  author: "NAPSTER AI"
+  maintainer: "NAPSTER AI"
+  openclaw:
+    requires:
+      bins: []
+---
+
+# Medical Entity Extractor
+
+Extract structured medical information from unstructured patient messages.
+
+## What This Skill Does
+
+1. **Symptom Extraction**: Identifies symptoms, severity, duration, and progression
+2. **Medication Extraction**: Finds medication names, dosages, frequencies, and side effects
+3. **Lab Value Extraction**: Parses lab results, vital signs, and measurements
+4. **Diagnosis Extraction**: Identifies mentioned diagnoses and conditions
+5. **Temporal Extraction**: Captures when symptoms started, how long they've lasted
+6. **Action Items**: Identifies requested actions (appointments, refills, questions)
+
+## Input Format
+
+```json
+[
+  {
+    "id": "msg-123",
+    "priority_score": 78,
+    "priority_bucket": "P1",
+    "subject": "Medication side effects",
+    "from": "patient@example.com",
+    "date": "2026-02-27T10:30:00Z",
+    "body": "I've been feeling dizzy since starting the new blood pressure medication (Lisinopril 10mg) three days ago. My BP this morning was 145/92."
+  }
+]
+```
+
+## Output Format
+
+```json
+[
+  {
+    "id": "msg-123",
+    "entities": {
+      "symptoms": [
+        {
+          "name": "dizziness",
+          "severity": "moderate",
+          "duration": "3 days",
+          "onset": "since starting new medication"
+        }
+      ],
+      "medications": [
+        {
+          "name": "Lisinopril",
+          "dosage": "10mg",
+          "frequency": null,
+          "context": "new medication"
+        }
+      ],
+      "lab_values": [
+        {
+          "type": "blood_pressure",
+          "value": "145/92",
+          "unit": "mmHg",
+          "timestamp": "this morning"
+        }
+      ],
+      "diagnoses": [
+        {
+          "name": "hypertension",
+          "context": "implied by blood pressure medication"
+        }
+      ],
+      "action_items": [
+        {
+          "type": "medication_review",
+          "reason": "possible side effect (dizziness)"
+        }
+      ]
+    },
+    "summary": "Patient reports dizziness after starting Lisinopril 10mg 3 days ago. BP elevated at 145/92. Possible medication side effect requiring review."
+  }
+]
+```
+
+## Entity Types
+
+### Symptoms
+- Name, severity (mild/moderate/severe), duration, onset, progression (improving/stable/worsening)
+
+### Medications
+- Name, dosage, frequency, route, context (new/existing/stopped)
+
+### Lab Values
+- Type (BP, glucose, cholesterol, etc.), value, unit, timestamp, normal range
+
+### Diagnoses
+- Name, context (confirmed/suspected/ruled out)
+
+### Vital Signs
+- Temperature, heart rate, respiratory rate, oxygen saturation, blood pressure
+
+### Action Items
+- Type (appointment, refill, question, callback), urgency, reason
+
+## Medical Terminology Handling
+
+The skill recognizes:
+- Common abbreviations (BP, HR, RR, O2 sat, etc.)
+- Brand and generic medication names
+- Lay terms for medical conditions ("sugar" → diabetes, "heart attack" → MI)
+- Temporal expressions ("since yesterday", "for the past week")
+
+## Integration
+
+This skill can be invoked via the OpenClaw CLI:
+
+```bash
+openclaw skill run medical-entity-extractor --input '[{"id":"msg-1","priority_score":78,...}]' --json
+```
+
+Or programmatically:
+
+```typescript
+const result = await execFileAsync('openclaw', [
+  'skill', 'run', 'medical-entity-extractor',
+  '--input', JSON.stringify(scoredMessages),
+  '--json'
+]);
+```
+
+**Recommended Model**: Claude Sonnet 4.5 (`openclaw models set anthropic/claude-sonnet-4-5`)
+
+## Privacy & Security
+
+- All processing happens locally via OpenClaw
+- No data is sent to external services (except Claude API for LLM processing)
+- Extracted entities remain in your local environment
+

skills/medical-priority-scorer/SKILL.md

@@ -0,0 +1,108 @@
+---
+name: medical-priority-scorer
+description: Score medical messages using urgency, sentiment, and patient context to produce priority rankings (P1/P2/P3).
+license: MIT
+metadata:
+  author: "NAPSTER AI"
+  maintainer: "NAPSTER AI"
+  openclaw:
+    requires:
+      bins: []
+---
+
+# Medical Priority Scorer
+
+Produce deterministic priority scores for medical messages without mutating any state.
+
+## Quick Triggers
+
+- Rank messages by medical urgency for callback priority
+- Classify messages into P1/P2/P3 queue
+- Score follow-up priority from triaged messages
+
+## Recommended Chain
+
+`medical-triage -> medical-priority-scorer -> medical-entity-extractor`
+
+## Execute Workflow
+
+1. Accept input from medical-triage containing triaged messages
+2. Score each message with:
+   - `urgency_score` in range `[0, 1]` (based on triage category)
+   - `sentiment_score` in range `[-1, 1]` (anxiety, distress, frustration)
+   - `recency_score` in range `[0, 1]` (how recent the message is)
+   - `patient_context_score` in range `[0, 1]` (chronic conditions, known patient)
+3. Compute `priority_score` on a 0-100 scale:
+   - `priority_score = 100 * (0.50*urgency_score + 0.25*sentiment_score_risk + 0.15*recency_score + 0.10*patient_context_score)`
+   - `sentiment_score_risk = max(0, -sentiment_score)` (negative sentiment = higher risk)
+4. Assign buckets:
+   - `P1` for `priority_score >= 75` (immediate attention)
+   - `P2` for `priority_score >= 50 and < 75` (same-day)
+   - `P3` for `< 50` (routine)
+5. Produce plain-language `evidence` tokens that explain the score
+
+## Input Format
+
+```json
+[
+  {
+    "id": "msg-123",
+    "category": "urgent",
+    "subject": "Medication side effects",
+    "from": "patient@example.com",
+    "date": "2026-02-27T10:30:00Z",
+    "body": "I've been feeling dizzy since starting the new medication..."
+  }
+]
+```
+
+## Output Format
+
+```json
+[
+  {
+    "id": "msg-123",
+    "priority_score": 78,
+    "priority_bucket": "P1",
+    "urgency_score": 0.8,
+    "sentiment_score": -0.4,
+    "recency_score": 1.0,
+    "patient_context_score": 0.6,
+    "evidence": "Urgent triage category + negative sentiment (concern) + very recent message + known patient"
+  }
+]
+```
+
+## Enforce Boundaries
+
+- Never write to databases or files
+- Never send messages or trigger outbound channels
+- Never create reminders or execute actions
+- Never bypass routing or approvals
+
+## Handle Errors
+
+1. Reject schema-invalid inputs
+2. Return field-level reasons when scoring cannot be computed
+3. Fail closed if required scoring features are missing
+
+## Integration
+
+This skill can be invoked via the OpenClaw CLI:
+
+```bash
+openclaw skill run medical-priority-scorer --input '[{"id":"msg-1","category":"urgent",...}]' --json
+```
+
+Or programmatically:
+
+```typescript
+const result = await execFileAsync('openclaw', [
+  'skill', 'run', 'medical-priority-scorer',
+  '--input', JSON.stringify(triagedMessages),
+  '--json'
+]);
+```
+
+**Recommended Model**: Claude Sonnet 4.5 (`openclaw models set anthropic/claude-sonnet-4-5`)
+

skills/medical-triage/SKILL.md

@@ -0,0 +1,115 @@
+---
+name: medical-triage
+description: Classify medical messages (emails, iMessages) as critical, urgent, or routine based on medical urgency indicators.
+license: MIT
+metadata:
+  author: "NAPSTER AI"
+  maintainer: "NAPSTER AI"
+  openclaw:
+    requires:
+      bins: []
+---
+
+# Medical Triage
+
+Classify medical messages into priority categories based on urgency indicators, symptoms, and patient context.
+
+## Categories
+
+| Icon | Category | Description |
+|------|----------|-------------|
+| 🔴 | `critical` | Life-threatening symptoms, emergency keywords, severe pain |
+| 🟡 | `urgent` | Needs same-day attention, worsening symptoms, medication issues |
+| 🟢 | `routine` | Follow-ups, general questions, appointment requests |
+
+## How It Works
+
+This skill analyzes message content for:
+- **Emergency keywords**: chest pain, difficulty breathing, severe bleeding, etc.
+- **Symptom severity**: pain scale, duration, progression
+- **Patient context**: chronic conditions, medications, recent procedures
+- **Temporal urgency**: "right now", "getting worse", "can't wait"
+
+## Input Format
+
+The skill expects a JSON array of messages:
+
+```json
+[
+  {
+    "id": "msg-123",
+    "subject": "Chest pain",
+    "from": "patient@example.com",
+    "date": "2026-02-27T10:30:00Z",
+    "body": "I've been having chest pain for the last hour..."
+  }
+]
+```
+
+## Output Format
+
+Returns a JSON array with triage results:
+
+```json
+[
+  {
+    "id": "msg-123",
+    "category": "critical",
+    "reason": "Chest pain mentioned - potential cardiac emergency",
+    "confidence": 0.95
+  }
+]
+```
+
+## Usage
+
+This skill is designed to be invoked programmatically via OpenClaw's skill execution API.
+
+## Medical Urgency Indicators
+
+### Critical (🔴)
+- Chest pain, pressure, or tightness
+- Difficulty breathing or shortness of breath
+- Severe bleeding
+- Loss of consciousness
+- Stroke symptoms (FAST: Face, Arms, Speech, Time)
+- Severe allergic reaction
+- Suicidal ideation
+
+### Urgent (🟡)
+- High fever (>103°F / 39.4°C)
+- Persistent vomiting or diarrhea
+- Medication side effects
+- Worsening chronic condition
+- Moderate pain (7-8/10)
+- Infection signs (redness, swelling, pus)
+- Mental health crisis
+
+### Routine (🟢)
+- Appointment requests
+- Prescription refills
+- General health questions
+- Follow-up on stable conditions
+- Lab result questions
+- Mild symptoms (<3 days)
+
+## Integration
+
+This skill can be invoked via the OpenClaw CLI:
+
+```bash
+openclaw skill run medical-triage --input '[{"id":"msg-1","subject":"...","body":"..."}]' --json
+```
+
+Or programmatically:
+
+```typescript
+const result = await execFileAsync('openclaw', [
+  'skill', 'run', 'medical-triage',
+  '--input', JSON.stringify(messages),
+  '--json'
+]);
+```
+
+**Recommended Model**: Claude Sonnet 4.5 (`openclaw models set anthropic/claude-sonnet-4-5`)
+