kairo-health-semantic-layer
Use Kairo's private health semantic layer to interpret its HealthKit ledger, daily profiles, trends, personal baselines, mood targets, learned factor scores, and evidence-backed insights without confu
SKILL.mdSkills/Skills/kairo-health-semantic-layer/SKILL.md
GitHubKairo health semantic layer
Use this skill when analysing Kairo health data, designing its dashboard or AI tools, or explaining its learned personal patterns.
- Read
references/semantic-layer.mdbefore constructing queries or claims. - Read
references/source-inventory.mdwhen assessing coverage or freshness. - Read
references/evidence.mdbefore interpreting sleep, mood, provenance, or predictive factors. - Prefer
health_daily_profilesandhealth_daily_featuresfor day/trend questions. Usehealth_current_observationsonly when a required feature is absent or being audited. - Use effective time for health behaviour and receipt time only for ingestion reliability.
- Treat mood self-reports as targets. Treat passive signals as predictors or context.
- Compare this owner with their own trailing baseline before invoking population reference ranges.
- Never imply causality, diagnosis, or certainty from
k; report target, horizon, lag, sample size, coverage, validation, and caveats. - Never interpret a missing record as zero or a denied HealthKit permission.
- Keep assistant context bounded and owner-authenticated. Do not expose unrestricted SQL or raw dumps to an AI model.
The canonical implementation and feature catalog live in the Kairo repository under docs/ and web/db/.