restaurants（restaurants 效率）

Core Behavior

• User mentions restaurant → offer to save with notes
• User asks for recommendation → check their saved places first
• User returns from meal → help document experience
• Create ~/restaurants/ as workspace

File Structure

~/restaurants/
├── to-try/
├── favorites/
├── visited/
│   └── 2024/
├── by-cuisine/
└── by-occasion/

To-Try Entry

# sushi-nakazawa.md
## Location
West Village, NYC

## Cuisine
Omakase sushi

## Source
Friend recommendation

## Price Range
$$$$

## Notes
Need reservation weeks ahead

Visited Entry

# la-mercerie-2024-03.md
## Date
March 15, 2024

## Occasion
Anniversary dinner

## What We Ordered
- Burrata (excellent)
- Duck breast (slightly dry)
- Chocolate soufflé (must order)

## Verdict
★★★★☆ — Would return for brunch

Favorites

# joes-pizza.md
## Go-To Order
Plain slice, extra crispy

## Best For
Quick lunch, late night

## Notes
Cash only, expect line

By-Cuisine and By-Occasion

Simple lists linking to favorites:

# date-night.md
- La Mercerie — beautiful space
- Via Carota — classic Village
- Carbone — never fails

What To Track

• Location and neighborhood
• Cuisine type
• Price range: $ to $$$$
• Reservation: needed? how far ahead?
• Standout dishes
• Rating after visit

Surfacing Recommendations

When user asks "where should we eat":

• Ask occasion and cuisine preference
• Check THEIR saved places first
• Suggest from their data before general knowledge

What To Surface

• "You haven't tried that sushi place on your list"
• "Last Italian you loved was L'Artusi"
• "For date night you rated La Mercerie highest"

Progressive Enhancement

• Start: add 5 places to try
• After meals: quick entry with verdict
• Build cuisine and occasion lists over time

What NOT To Do

• Recommend unsaved places without asking
• Forget dietary restrictions
• Over-organize — simple notes work