Intelligent Food Absorption
This skill absorbs ANY food input, auto-classifies it, and organizes for insights.
Rules:
- Auto-detect input type: meal photo, nutrition label, recipe, menu, text
- Extract and structure: items, portions, context, nutrition when visible
- Tag everything: #meal, #recipe, #product, #restaurant, #inventory
- Offer analysis: "Want nutrition estimate?" — don't force it
- Build personal database: scanned labels, frequent meals, saved recipes
- Provide insights: patterns, variety, timing, correlations
- Remember restrictions permanently, flag conflicts proactively
- For detailed macro tracking → complement with
caloriesskill - Check
processing.mdfor how each input type is handled
Memory Storage
All user data persists in: ~/food/memory.md
Format:
### Preferences
<!-- Their food preferences and restrictions. Format: "item: type" -->
<!-- Examples: nuts: allergy, gluten: intolerance, vegetarian: choice -->
### Products
<!-- Scanned/saved products for quick-log. Format: "product: cal/serving" -->
<!-- Examples: Hacendado yogurt: 120/170g, Oatly oat milk: 45/100ml -->
### Patterns
<!-- Detected eating patterns. Format: "pattern" -->
<!-- Examples: breakfast ~8am, snacks after 10pm, eats out Fridays -->
### Places
<!-- Restaurants and spots. Format: "place: notes" -->
<!-- Examples: Noma: loved fermented plum, Local Thai: go-to takeout -->
### Recipes
<!-- Saved recipes. Format: "dish: key info" -->
<!-- Examples: quick hummus: chickpeas+tahini+lemon 5min, Sunday roast: 2h -->
Empty sections = no data yet. Absorb, classify, organize.
Insights provided: Weekly variety score, meal timing patterns, frequent foods, eating out ratio, nutrition estimates when asked. Not medical advice.