Wearables and health metrics (privacy-first)

Wellness-oriented summaries, not clinics:

• Daily or weekly digests: consistency, not obsessive minute-by-minute surveillance unless you want that explicitly.
• Habit nudges tied to goals you wrote down—hydration, movement, sleep wind-down.
• Export awareness: know what leaves the wearable vendor and what stays local.

Community excitement around wearables is really signal + discipline—agents summarize; you decide.

• Custom skills can wrap vendor APIs where terms allow automation.
• Local-first posture helps keep health JSON off random third parties—still read vendor ToS.
• Approvals before messaging anyone else about your health data.

1. Read vendor API terms—some forbid certain automations.
2. Minimize stored PHI; prefer aggregates when possible.
3. Separate “coaching” from “medical”—do not diagnose.
4. Opt-out paths for bad days when nudges feel cruel.
5. Review weekly with a human lens—trends beat single numbers.

• Informed consent if coaches or family see summaries.
• Accuracy limits of consumer wearables—document uncertainty.

1. Connect read-only summaries first.
2. Define red lines: when to notify a human vs stay silent.
3. Correlate lightly with calendar stress—causal claims are hard; avoid them.
4. Rotate tokens; revoke on device change.
5. Audit who can query historical data.

• “Summarize trends for the last 14 days—no diagnostic language.”
• “Flag inconsistencies between reported sleep and subjective notes I typed.”
• “What disclaimer should appear on any shared summary?”

• Privacy review for regional health data rules that apply to you.
• Crisis resources linked if mental-health-adjacent topics appear—policy choice.
• User can delete historical integrations in one place.

• Pseudo-medical advice from averages and correlations.
• Shame loops from streaks and scores.
• Sharing biomarkers in chat channels accidentally.
• Vendor API changes breaking silent—monitor errors.