Personalized Health Checkup Appointment

Problem

Checkup recommendations often run into fabricated items, estimated prices, and privacy leakage. A model may list tests from general knowledge, but available items, prices, and substitutes vary by institution. Writing names, phone numbers, or other PII into the booking flow also expands sensitive data exposure. This skill constrains recommendations to auditable data files and local scripts.

How It Works

It first collects age, gender, symptoms, family history, and prior abnormalities, then follows a fixed pipeline:
- Risk assessment: reads reference/risk_logic_table.json.
- Symptom mapping: reads reference/symptom_mapping.json, including synonyms.
- Item validation: runs node scripts/verify_items.js so only items in the local catalog can be recommended.
- Price calculation: runs node scripts/calculate_prices.js instead of letting the model calculate totals.
- Booking output: presents the plan with templates and can generate an output.png QR code after user consent.

Boundaries

This is for pre-booking recommendation, not clinical diagnosis. Plans must respect the 600 CNY minimum and the mandatory item029 baseline. If an item is only available at one institution, it should search for equivalent items from another provider and compare total prices. Network behavior is limited to scripts/sync_items.js, syncing only anonymized item IDs; the QR code contains booking codes only.