Garmin Health Analysis

**Quick judgment**: Strong, feature-rich skill for Garmin users who want AI-assisted health and training analysis. Well-documented, with wide metric coverage and good visualization support. Best suited to technically comfortable users or environments that can securely manage credentials and tokens. **What it does well** - Pulls a wide range of health and activity data from Garmin Connect via `garminconnect` (sleep, HRV, Body Battery, HR, activities, stress, etc.). - Provides both JSON outputs for reasoning and interactive Chart.js dashboards for visual review. - Clearly documents how to translate natural questions (sleep, recovery, workouts, trends) into concrete CLI/script calls. - Offers an interpretation framework for metrics and trends, with explicit “not medical advice” guidance. **Key risks & limitations** - **Unofficial API**: Uses an unofficial Garmin Connect API via `garminconnect`. Breakage risk if Garmin changes their endpoints or rate limits more aggressively. - **Credential handling**: Requires Garmin email/password or tokens stored locally. While scoped to the local machine, it increases sensitivity to local compromise; careful permissions and host security are important. - **Health data sensitivity**: Handles detailed personal health data; logs, tokens, and output files must be stored and shared carefully. - **Non-medical**: Insights are informational only and should not be treated as medical or diagnostic advice. **Recommended scenarios** - Individual Garmin users who want AI to answer questions like “How’s my recovery this week?” or “Is my HRV improving?” based on real data. - Athletes or serious hobbyists tracking training load, sleep, and recovery with recurring automated summaries and dashboards. - Coaches or power users building private health dashboards and reports from Garmin data in a controlled environment. **Less suitable for** - Non-Garmin users or contexts where you cannot safely store credentials/tokens on the host. - Clinical or regulated healthcare settings where unofficial APIs and non-validated interpretations are unacceptable.