Daolv Hotel Booking Assistant

SKILL.md

---
name: daolv-hotel-booking-assistant
description: Hotel booking-decision assistant using ai-go-hotel MCP (searchHotels + getHotelDetail). Use when users already have candidate hotels and need room-rate details, cancellation policy checks, final recommendation, and booking handoff actions.
---

# Daolv Hotel Booking Assistant

Focus on booking readiness and risk control.

## Workflow

1. Confirm booking context
- Ensure destination/date/guest count/room count/currency are clear.
- If user already has `hotelId`, prioritize direct detail lookup.

2. Build candidates
- If no candidates provided, call `searchHotels` first for 3-5 options.
- Then call `getHotelDetail` for finalists (prefer `hotelId` over `name`).

3. Evaluate room plans
- Compare room plans by:
  - total/average price
  - cancellation windows/penalties
  - breakfast and included fees
  - bed type / room type fit
- Limit displayed rows from large `roomRatePlans`.

4. Return booking decision + handoff
- Give one clear “best plan” and one fallback.
- Include booking risks and what to double-check before payment.
- Provide direct booking URL if available.

## Output Template

- **预订需求确认**：入住/离店/人数/房间数/预算
- **最佳方案**：酒店 + 房型 + 价格 + 退改条款 + 推荐理由
- **备选方案**：关键差异（更便宜/更灵活/位置更优）
- **风险提示**：取消时点、不可退规则、库存波动
- **下单前确认清单**（2-4 条）

## Quality Bar

- Never invent policy text.
- Handle plain-text error response from `getHotelDetail` gracefully.
- If availability uncertain, advise immediate refresh query.
- Keep recommendation decisive, avoid list dumping.

## MCP Preset Config

- Embedded preset: `references/mcp-client-config.json`
- Endpoint: `https://mcp.aigohotel.com/mcp` (`streamable_http` + prefilled Authorization header)