Personal Health Agent

Personal Health Agent

Overview

This skill helps an agent collect personal health data, keep it in a local JSONL log, summarize trends, and generate reminders. It is designed for routine self-tracking: blood pressure, blood tests, exercise, and body metrics.

It is not a medical device and does not diagnose, prescribe, or replace a clinician. Treat every result as a tracking aid for better conversations with qualified healthcare professionals.

When to Use

Use this skill when the user wants to:

Do not use this skill for emergency triage beyond telling the user to seek urgent medical care for red-flag situations. Do not make medication changes, diagnoses, or treatment decisions.

Quick Start

Natural language examples:

今天血压 126/82 心率 68
血检 LDL 142, HDL 52, HbA1c 5.9
record exercise: brisk walk 45 minutes
生成本周健康报告

Structured JSON examples:

{"action":"record","type":"blood_pressure","date":"2026-05-02","systolic":126,"diastolic":82,"pulse":68}

{"action":"record","type":"blood_lab","date":"2026-04-20","markers":{"LDL":142,"HDL":52,"HbA1c":5.9}}

{"action":"report","period":"weekly","end_date":"2026-05-02"}

Data Collection Workflow

1. Parse the user's entry as structured JSON first. 2. If it is not JSON, parse common natural-language patterns for blood pressure, exercise, and lab markers. 3. Normalize units and marker names where supported. 4. Append the record to the local JSONL store. 5. Return immediate analysis, safety flags, reminders, and next steps. 6. For reports, load the local store and summarize the requested weekly or monthly period.

The default store is:

~/.personal-health-agent/health_records.jsonl

Override it for tests or private project storage:

export PERSONAL_HEALTH_AGENT_DATA_DIR=/path/to/private/data

Supported Record Types

Blood Pressure

Fields: date, systolic, diastolic, optional pulse, context, and notes.

The handler groups readings into broad tracking ranges and flags very high readings. If systolic is at least 180 or diastolic is at least 120, the response instructs the user to recheck after quiet rest and seek urgent care if the reading persists or symptoms are present.

Blood Labs

Fields: date and markers.

Common aliases are normalized for LDL, HDL, triglycerides, total cholesterol, HbA1c, glucose, ALT, AST, creatinine, eGFR, uric acid, and hemoglobin. Built-in thresholds are generic reference aids only; official lab ranges and clinician context take precedence.

Exercise

Fields: date, activity, minutes, optional distance_km, intensity, and notes.

Weekly summaries compare logged moderate/vigorous minutes against a general 150-minute adult benchmark. This is a wellness benchmark, not a personalized prescription.

Body Metrics

Fields can include weight_kg, waist_cm, and body_fat_percent.

Body metrics are stored for future trend context. Avoid judgmental or appearance-focused language.

Report Output

Weekly and monthly reports include:

Safety Rules

Verification

From the skill directory:

python3 tests/test_handler.py
python3 handler.py '{"action":"help"}'

The first command runs behavior tests. The second verifies the handler entrypoint returns JSON.

Natural language examples:

今天血压 126/82 心率 68
血检 LDL 142, HDL 52, HbA1c 5.9
record exercise: brisk walk 45 minutes
生成本周健康报告

Structured JSON examples:

{"action":"record","type":"blood_pressure","date":"2026-05-02","systolic":126,"diastolic":82,"pulse":68}

{"action":"record","type":"blood_lab","date":"2026-04-20","markers":{"LDL":142,"HDL":52,"HbA1c":5.9}}

{"action":"report","period":"weekly","end_date":"2026-05-02"}