Medical Appointments

Workflows

Find Doctors

1. If a specialty name is given, GET /api/specialties to resolve the name to an ID (case-insensitive match).
   • If no match exists, list available specialties and ask the user to choose one.
2. GET /api/doctors?specialtyId={id}&name={query} (both params optional).
3. Display each doctor: name, specialty, available days, slot duration (minutes), bio.

Check Available Slots

1. GET /api/doctors/{doctorId}/slots?date={YYYY-MM-DD}.
2. Filter to slots where available: true.
3. Display each as HH:MM – HH:MM. State total available vs total slots.
4. If zero available, note whether the doctor works on that day.

Book Appointment

Required inputs: patient ID, doctor ID, appointment date/time (ISO 8601), reason.

1. Fetch confirmation details in parallel:
   • GET /api/patients/{patientId} — patient name
   • GET /api/doctors/{doctorId} — doctor name, specialty, slot duration
2. Ask the user to confirm before proceeding. Present:
   • Patient name, doctor name, specialty, date, time, duration, reason
3. On confirmation: POST /api/appointments with body {"patientId","doctorId","dateTime","reason"}.
4. Return the created appointment ID and full details.
5. On refusal: abort without making any API call.

Note: The MCP server uses native elicitation (a UI confirmation dialog). This skill asks for confirmation through conversation. See Limitations in the README.

Cancel Appointment

Required input: appointment ID.

1. GET /api/appointments/{appointmentId} — retrieve current details.
2. Fetch patient and doctor names for a readable summary.
3. Ask the user to confirm cancellation. Show: appointment ID, patient, doctor, scheduled date/time, reason.
4. On confirmation: PATCH /api/appointments/{appointmentId}/cancel.
5. Return confirmation of cancellation with original scheduled details.
6. On refusal: leave the appointment unchanged.

Note: Same elicitation limitation as Book Appointment. See Limitations in the README.

List Appointments

All filters are optional.

1. Build query string from any provided filters: patientId, doctorId, status (scheduled | cancelled | completed), date (YYYY-MM-DD).
2. GET /api/appointments?{filters}.
3. Display each as: {id} | {date} {time} | status: {status} | {reason}.

Recommend Specialist

1. GET /api/specialties — load all specialties with descriptions.
2. Analyze the patient's symptoms against each specialty description using your own reasoning.
3. Identify the single most appropriate specialty; explain the reasoning briefly.
4. GET /api/doctors?specialtyId={id} — list matching doctors.
5. Present: recommended specialty, rationale, and the available doctors.

Note: The MCP server calls a sub-LLM (sampling) for this step. The skill uses the agent's own reasoning directly, producing an equivalent result. See Limitations in the README.

Schedule Appointment (Guided Workflow)

End-to-end booking for a requested specialty and preferred date.

1. GET /api/specialties — load all specialties.
2. Match the requested specialty (case-insensitive). If no match, list available options and ask the user to choose.
3. GET /api/doctors?specialtyId={id} — list doctors in the specialty with bios and available days.
4. For each doctor, GET /api/doctors/{id}/slots?date={preferredDate} — check open slots on the preferred date.
5. Present doctors that have availability, with their open time slots.
6. Ask the user to select a doctor and time slot.
7. Ask for the patient ID if not already known (available patient IDs: pat-1 through pat-5; or GET /api/patients to list them).
8. Ask for the reason for the visit.
9. Run the Book Appointment workflow above.

Patient History

1. GET /api/patients/{patientId} — retrieve patient info.
2. GET /api/appointments?patientId={patientId} — get the full appointment list.
3. Summarize:
   • Counts: total, scheduled, completed, cancelled.
   • Chronological list: date, doctor ID, status, reason.
   • Highlight upcoming scheduled appointments.
   • If last completed appointment was more than 6 months ago (or none exists), recommend scheduling a check-up.

Triage Symptoms

1. GET /api/specialties — load all specialties with descriptions.
2. Analyze the reported symptoms against the specialty descriptions. Ask clarifying questions if the symptoms are ambiguous or could map to multiple specialties.
3. Identify the most appropriate specialty with a brief rationale.
4. Run Recommend Specialist to surface matching doctors.
5. Offer to start the Schedule Appointment workflow if the patient wants to book.