health-snapshot

---

name: health-snapshot description: Fetch the consolidated health snapshot from a DocVault instance (Apple Health + clinical labs + DNA). Use when the user wants a full health view, or asks to drill into specific lab trends, daily segment data, or DNA details. argument-hint: [person name or person-id] [archived|no-dna|no-clinical]

Fetch the consolidated health snapshot from a DocVault instance and present it.

Setup

The skill expects DocVault to be reachable at ${DOCVAULT_URL} (default http://localhost:3005). Override when running scripts directly:

export DOCVAULT_URL=http://docvault.local:3005

Instructions

1. Parse $ARGUMENTS:

   • Person name or person-xxxxxx id → set personId=<value> on the query string
   • archived → includeArchived=true
   • no-dna → includeDNA=false
   • no-clinical → includeClinical=false
   • Otherwise default: all non-archived people, clinical + DNA included

2. Fetch the markdown snapshot:

   curl -s "${DOCVAULT_URL:-http://localhost:3005}/api/health-snapshot?format=md${EXTRA_QS}"
   

3. If the curl fails (connection refused, timeout), tell the user the DocVault container may not be running. Check docker ps | grep docvault (or ssh <host> 'docker ps | grep docvault' for a remote deployment).

4. Present the full markdown output to the user — do NOT summarize or truncate

5. After presenting, offer to help interpret trends (e.g. "your RHR trend is improving — want me to explain why?"), flag concerning labs, suggest questions to bring to a doctor, or recommend follow-ups based on illness periods and overdue immunizations

Drilling deeper — bundled helper scripts

The snapshot only shows the latest value per lab and headline stats per segment. Use these scripts when the user asks for a trend, per-day data, or the full history of a single lab.

List all people on the instance

bash "${CLAUDE_SKILL_DIR}/scripts/person-ids.sh"
# → person-xxxxxx   <name>

Pass --include-archived to also list archived people.

Trend a single clinical lab over time

bash "${CLAUDE_SKILL_DIR}/scripts/lab-trend.sh" <personId> <labNameSubstring>

• labNameSubstring is case-insensitive. "LDL" matches "LDL, DIRECT*"; "hemoglobin" matches "HEMOGLOBIN A1C*".
• Prints every reading with date, value, unit, flag, and the reference range.
• If no lab matches, the script prints the first ~30 available lab names for the person so you can pick the right substring.

Raw endpoints — when scripts aren't enough

Question	Endpoint
Full daily activity / heart / sleep / workouts / body	GET /api/health/<personId>/snapshot/<segment> where segment is activity / heart / sleep / workouts / body / all
Every lab test with full history, conditions, medications, allergies, immunizations, procedures	GET /api/health/<personId>/clinical
Check if DNA exists before fetching	GET /api/health/<personId>/dna/status
Full DNA traits + health + polygenic + APOE	GET /api/health/<personId>/dna
Re-parse from the original zip	POST /api/health/<personId>/parse-export with {"filename":"export.zip"}

Prefer piping through node or jq rather than dumping the whole JSON into context — per-segment payloads can be hundreds of KB.

Notes

• Endpoint base: ${DOCVAULT_URL:-http://localhost:3005}/api/health-snapshot.
• Formats:
  • ?format=md — readable markdown tables (default for this command).
  • ?format=toon — compact flat key=value lines (~60% fewer tokens; default on the endpoint itself).
  • ?format=json — full structured response for downstream processing.
• Extra query params:
  • ?personId=person-xxxxxx — single person
  • ?includeArchived=true — include archived people
  • ?includeClinical=false — skip FHIR lab/panel/condition data
  • ?includeDNA=false — skip DNA (avoids decrypting results.json.enc)
  • ?includeDaily=true — (json only) include full daily arrays
• The snapshot aggregates:
  • Apple Health headlines: activity, heart (RHR/HRV trends), sleep, workouts, body weight
  • Auto-detected illness periods and user-annotated illness notes
  • FHIR clinical records: latest lab per test (flagged high/low), conditions, medications, allergies, immunizations, procedures
  • DNA: traits, health traits, polygenic scores, APOE genotype (decrypted on demand using DOCVAULT_MASTER_KEY)
  • Health-tagged reminders from .docvault-reminders.json (keyword match on title/notes)

Cross-reference the user's regimen before advising

If the health-snapshot response includes a nutrition section for the person, it contains their current supplement + nutrition regimen — what they're actively taking, what they're considering, past items, along with dose + time-of-day + notes. Read this section before recommending supplement additions, changes, or dose tweaks. Don't re-ask the user questions their catalog has already answered (current doses, brand preferences, what they've decided against).

The markdown format surfaces an aggregate micronutrient totals table across all active supplements — use that when flagging over/under-dosing (e.g. selenium stacking, vitamin E additive effects, Zn:Cu ratio).

Standing rule: upload shared labels to DocVault

When the user shares a supplement, medication, or nutrition label in a conversation, upload it to their DocVault nutrition catalog via the API so the regimen stays accurate. Use the ingestion endpoint:

# Upload a label (saves image, runs Claude Vision parser, stores structured data)
curl -X POST --data-binary "@/path/to/label.png" \
  -H "Content-Type: application/octet-stream" \
  "${DOCVAULT_URL:-http://localhost:3005}/api/health/<personId>/nutrition/upload?filename=<name>.png&status=considering"

# PATCH after upload to set status/dose/notes (product/brand names if parser missed them)
curl -X PATCH -H "Content-Type: application/json" \
  -d '{"status":"active","dose":{"amount":2,"unit":"capsules","frequency":"daily","timeOfDay":"morning"},"notes":"..."}' \
  "${DOCVAULT_URL:-http://localhost:3005}/api/health/<personId>/nutrition/<id>"

The catalog becomes the durable record — no label image lives in this skill folder.

Privacy reminder

The payload contains real clinical records and genetic data. Do NOT paste the output into external tools, chat interfaces, or commit it to a repository. Keep it in the session.