gourmet-research

# Gourmet Research

## Overview
Template-first workflow for traceable, comparable, auditable food recommendations across cities. Keep evidence, scores, and decisions synchronized.

## Core Rules (Non-Negotiable)
- If the user has not specified an **output language**, ask once at project start and record it in `overview.md`.
- Use **gourmet/<city-slug>/** with 6 files: `overview.md`, `inbox.md`, `candidates.md`, `notes.md`, `top-places.md`, `excluded.md`.
- **Never fabricate** sources, ratings, or claims. Use `unknown` when missing.
- Prefer **original-language place names** (not translated) unless the user requests otherwise.
- **Preserve audit trail**: never delete candidates; mark `rejected` and record why in `excluded.md`.
- Default **minimum sources = 4**. If the locale is information-sparse, allow **3** only when you record `evidence: limited` with the reason and attempted sources.

## Template-First Workflow (Summary)
1. **Initialize**: Ask for output language + city, then copy templates from `assets/templates/` into the city folder.
2. **Normalize**: Update language/city placeholders in the copied files before research begins.
2. **Discovery**: Capture raw ideas in `inbox.md`, then move top candidates into `candidates.md` with `status: inbox`.
3. **Evidence**: For each candidate, write a full evidence block in `notes.md` with sources + practical constraints.
4. **Score**: Apply the 50-point rubric and justify each component in `notes.md`.
5. **Decide**: Promote to Top Picks (>=35), Backups (30-34), or reject (<30).
6. **Publish**: Update `top-places.md` and `excluded.md` to match decisions.
7. **Verify**: Ensure no `inbox` statuses remain and required sections exist.

## Ranking Retrieval (When user asks for “highest score”)
Before extracting any “top N” list, **confirm the scope**:
- **Geography**: Okinawa *prefecture* vs *main island only* vs *specific subarea*.
- **Category**: overall vs cuisine category.
- **Source URL**: must match the user’s intent exac