L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Productivité · SKILL.md

Restaurants

Restaurants

Build a personal restaurant system for tracking places to try, favorites, and dining memories.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: Restaurants slug: restaurants version: 1.0.0 description: Build a personal restaurant system for tracking places to try, favorites, and dining memories. homepage: https://clawic.com/skills/restaurants metadata: clawdbot: emoji: 🍽️ os: - linux - darwin - win32 displayName: Restaurants

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 ~/Clawic/data/restaurants/ as workspace

File Structure

~/Clawic/data/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
Skills enregistrées