nucleus-apple-health

name	nucleus-apple-health
description	Query and analyze exported Apple Health metrics and raw samples with the `nucleus-apple` CLI. Use when the task involves trend summaries, daily metrics, raw sample inspection, sync change history, or diagnosing gaps between aggregates and raw exports.
metadata	{"openclaw":{"emoji":"❤️","homepage":"https://github.com/zish-rob-crur/nucleus-apple-mcp","os":["darwin"],"requires":{"anyBins":["nucleus-apple","uvx"]},"install":[{"id":"uv","kind":"uv","package":"nucleus-apple-mcp","bins":["nucleus-apple"],"label":"Install nucleus-apple (uv)"}]}}

Use nucleus-apple health ... to summarize exported Health data and drill into raw samples only when needed.

Operating Stance

• Prefer the installed nucleus-apple binary.
• Fall back to uvx --from nucleus-apple-mcp nucleus-apple ... when only uvx is available.
• Start with snapshot-based reads; escalate to raw sample reads only when the task needs record-level evidence or gap diagnosis.
• Use exact YYYY-MM-DD dates and keep ranges as narrow as the question allows.
• Narrow heavy reads with metric_keys, type_keys, tags, kinds, or max_records.
• Treat missing_dates, metric_status, and raw manifest status as separate signals.
• Reuse returned cursors when paginating raw reads instead of widening filters.

Core Workflow

1. If metric vocabulary is unclear, start with list-sample-catalog.
2. Route summary questions through snapshot reads using references/decision-rules.md.
3. Escalate to inspect-day before raw samples when the question is "why is this missing or inconsistent?"
4. Use references/commands.md for canonical command shapes.
5. Use references/failure-modes.md to interpret missing data, authorization gaps, and storage errors.

Task Routing

• Trend summaries or range overviews: read references/decision-rules.md#snapshot-routing.
• Single-day values or missing-metric diagnosis: read references/decision-rules.md#daily-and-diagnostic-routing.
• Record-level raw samples, workouts, or filtered manifests: read references/decision-rules.md#raw-escalation and references/decision-rules.md#filter-selection.
• Recent sync or export changes: read references/decision-rules.md#change-history.
• Unsupported metrics, aggregate-vs-raw gaps, or empty filtered reads: read references/failure-modes.md.

Output / Escalation

• Use --pretty only for human inspection.
• Keep raw reads bounded with --max-records unless full pagination is explicitly requested.
• Prefer inspect-day over direct raw reads when the task is to explain a missing or inconsistent daily aggregate.