Agentic Vitals Watch — Development skill for Claude Code

VitalsWatch

A simulated agent-to-equipment health monitoring demo — a monitoring agent watches a single wearable sensor, reports to a doctor agent, and escalates clearly when something looks critical. Built with the [Claude API](https://docs.claude.com).

⚠️ This is a simulated portfolio demo, not a real device

Every reading, the doctor agent's responses, and the "911 call" are entirely fake and generated in-browser for demonstration purposes. **Nothing is actually monitored, dialed, or contacted.** This is a prototype of an agent coordination *pattern* — not a certified medical device, and it must never be used, or relied on, for real home health monitoring or real emergencies. A real version of this concept would require certified medical-grade sensors, real telephony integration, regulatory clearance (e.g. FDA), and continuous human clinical oversight — none of which a browser demo can honestly provide.

Live demo

**[Try it here](https://claude.ai/artifact/GixyPdQKdUj2moJWT5Tb1i)** — note: trying the interactive part requires a free Claude account to sign in with (a platform requirement for any page that calls Claude live, not specific to this project).

The problem

Remote/in-home patient monitoring is a real, growing category — wearables that watch vitals and alert someone when something's wrong, so an at-risk person (elderly, chronically ill, recovering from surgery) can live independently with a safety net. The interesting technical problem isn't the sensor — it's the coordination: who gets notified, in what order, based on what thresholds, and what happens if it's a real emergency versus a false alarm. This project models that coordination logic directly.

What it does

Simulates a single wrist-worn sensor (heart rate, SpO2, fall/motion detection) and:

1. A Monitoring Agent checks in on a regular interval, classifying status as normal, watch, urgent, or critical — a detected fall is always treated as at least critical, since an unrespon