health-memory - SKILL.md Agent Skill

name: health-memory description: "Centralized health memory hub — manages daily logs and per-item longitudinal tracking files under memory/health/. WHEN TO USE: After any health skill records data, or when answering health questions that need historical context." version: 1.0.0 user-invocable: false metadata: {"openclaw":{"emoji":"🧠","category":"health"}}

health-memory — Centralized Health Memory Hub

Instruction-only skill with no Python scripts. All operations are performed by the agent using built-in tools (Read/Write/Edit/Grep/Glob).

Goals

Single source of truth — Provide one canonical location (memory/health/) for all health data across skills
Daily log aggregation — One file per day, multi-skill; each skill writes its own section
Per-item longitudinal tracking — Dedicated files for blood pressure, tumor markers, medications, etc. with rolling 90-day history
Health file indexing — PDFs, images, and lab reports stored in files/; only summaries appear in memory (never full text)
Shared format contract — Any health skill can read/write without code dependencies, just by following the format rules below

Directory Structure

memory/health/
├── _health-profile.md              # Long-term health baseline (manual / agent-maintained)
├── daily/                           # Daily logs (one file per day)
│   ├── 2026-03-10.md
│   ├── 2026-03-09.md
│   └── ...
├── items/                           # Per-item longitudinal tracking
│   ├── blood-pressure.md            # Blood pressure history
│   ├── blood-sugar.md               # Blood sugar history
│   ├── blood-lipids.md              # Blood lipid history
│   ├── tumor-markers.md             # Tumor marker history
│   ├── kidney-function.md           # Kidney function history
│   ├── weight.md                    # Weight history
│   ├── medications.md               # Medication records
│   ├── supplements.md               # Supplement records
│   └── ...                          # Agent creates new files as needed
└── files/                           # Health documents (PDFs, images, etc.)
    ├── 2026-03-10_blood_routine.pdf
    └── ...

Daily File Format

Path: memory/health/daily/YYYY-MM-DD.md

---
date: YYYY-MM-DD
updated_at: YYYY-MM-DDTHH:MM:SS
---

# YYYY-MM-DD

## Blood Pressure [blood-pressure-tracker · 09:15]
- 120/80 mmHg, pulse 72
- Grade: Normal

## Caffeine [caffeine-tracker · 14:30]
- Residual: ~230mg
- Safe to sleep: 22:30
- Today's intake: Espresso 150mg (08:30), Latte 80mg (14:30)

## Health Files
| File | Type | Summary | Path |
|------|------|---------|------|
| Blood routine | Lab report | WBC 6.2, Hb 135, PLT 210 all normal | memory/health/files/2026-03-10_blood_routine.pdf |

Format Rules

Section header: ## Title [skill-name · HH:MM] — fixed format for easy Grep lookup
Same skill, same day, multiple updates: Use Edit to replace the existing section; do NOT duplicate
## Health Files section: Always at the end of the file; table indexing file summaries + paths
Never put full file contents in memory: Only 1-2 line summaries; agent reads the original file on demand

Item File Format

Path: memory/health/items/{item}.md

---
item: blood-pressure
unit: mmHg
updated_at: YYYY-MM-DDTHH:MM:SS
---

# Blood Pressure Records

## Recent Status
- Latest: 120/80 mmHg (2026-03-10 09:15)
- 7-day average: 118/78 mmHg
- Trend: Stable

## History
| Date | Systolic | Diastolic | Pulse | Notes |
|------|----------|-----------|-------|-------|
| 2026-03-10 | 120 | 80 | 72 | Morning |
| 2026-03-09 | 125 | 82 | 75 | Morning |
| 2026-03-08 | 118 | 78 | 70 | |

Format Rules

## Recent Status: Latest value + short-term trend summary (for quick glance)
## History: Table format, newest first, keep last 90 days
Data older than 90 days: Agent removes trailing rows during updates
Item file names: Agent decides (health item name → English slug); creates new files automatically for new items
Table columns: Vary by item type (e.g., blood pressure has systolic/diastolic; tumor markers have CEA/CA199, etc.)

Health Profile

Path: memory/health/_health-profile.md

Long-term health baseline. Maintain the existing format. Agent may update as needed.

Write Protocol

After any health skill updates data, execute these steps:

Find or create daily file: Glob memory/health/daily/ for today's file. If it doesn't exist, Write a new one with frontmatter and an empty ## Health Files table
Read the daily file: Read the full content
Update the skill section: Grep for ## .* \[{skill-name} · to find the existing section. If found, Edit to replace it. If not found, insert a new section before ## Health Files
Update frontmatter: Set updated_at to current ISO timestamp
Update the corresponding item file (memory/health/items/{item}.md):
- Read the item file (if it doesn't exist, Write a new one with frontmatter, ## Recent Status, and an empty ## History table)
- Update the ## Recent Status section
- Insert a new row at the top of the ## History table (after the header row)
- If the table exceeds 90 days, delete trailing old rows

File Intake Protocol

When the user provides a health document (PDF, image, etc.):

Read the file and generate a summary (1-2 lines of key indicators / conclusions)
Save or move the original file to memory/health/files/, naming format: YYYY-MM-DD_{description}.{ext}
Append a row to the ## Health Files table in today's daily file (file name + type + summary + path)
If the document contains specific health item data (e.g., WBC from a blood routine), also update the corresponding item file

Read Protocol

When the user asks a health question, load context incrementally:

Read memory/health/_health-profile.md (baseline)
Read memory/health/daily/{today's date}.md (current day status)
If the question involves a specific health item trend → Read memory/health/items/{item}.md
If the question involves a historical date → Glob memory/health/daily/*.md to find relevant date files, then Read
If a daily or item file references a health document → Read the specific file (PDF, etc.) on demand
Do NOT load all files at once — load incrementally to avoid overwhelming the context window

Integration Guide for Other Skills

How to integrate: After your health skill finishes its data operations, update the memory files following the format rules above.

No Python imports needed. No scripts to call. The agent executing your skill simply uses Read/Edit/Write tools to update memory/health/daily/ and memory/health/items/ files per this spec.

Useful Grep patterns:

Find a skill's section in today's daily file: ## .* \[{skill-name} · in memory/health/daily/{date}.md
Find all daily files mentioning a health item: ## {item-name} in memory/health/daily/

New item file template:

---
item: {english-slug}
unit: {unit}
updated_at: {ISO-timestamp}
---

# {Item Name} Records

## Recent Status
- Latest: (pending)

## History
| Date | {Col1} | {Col2} | Notes |
|------|--------|--------|-------|