Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →私人旅行管家 — 从出发地到目的地的完整行程规划+攻略导出。 输入出发地、目的地、天数、预算、风格偏好，自动输出闭环行程， 包含交通推荐、酒店推荐、美食路线、每日预算，并可选生成攻略。 当用户提到「做攻略」「旅行规划」「旅游计划」「行程安排」时使用。
.claude/skills/davepoon-travel-agent-skill/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 34% | 0% |
> 「不是给你一堆信息，是给你一张可以直接执行的日程表。」
| 功能 | 说明 | |:---|:---| | 📋 行程规划 | 输入地点/天数/预算/风格，输出完整行程 | | 🚄 交通推荐 | 高铁车次/航班 + 票价 + 购票指引 | | 🏨 住宿推荐 | 区域+档次+价格+预订平台指引 | | 🍜 美食路线 | 餐厅推荐+人均+平台链接 | | 💰 预算透明 | 每日花费小计 + 总预算汇总 + 省钱方案 | | 📄 导出攻略 | 一键生成精美HTML攻略，双击即用 |
一次性问清以下问题（使用 AskUserQuestion 工具）：
使用 WebSearch 获取实时信息：
| 查什么 | 为什么 | |:---|:---| | 高铁/航班信息 | 确保推荐的车次存在且价格合理 | | 酒店区域+价格 | 推荐的档次和预算匹配 | | 景点+门票 | 景点真实存在且时间安排合理 | | 当地美食 | 推荐真实店铺，不瞎编 | | 当地天气 | 影响行程安排 |
按天输出，包含：交通、住宿、美食、景点、每日小计。
如果用户说「导出」「生成文档」「生成攻略」等，输出自包含HTML攻略文件。 攻略为独立HTML文件，双击即可在浏览器打开，无需任何服务器或依赖。
📋 行程总览
├── 路线：广州 → 贵阳
├── 天数：3天2晚
├── 预算：¥2,000/人
├── 预估总花费：¥1,568/人 ✅ 预算内
└── 风格：休闲+美食
───── Day 1 ─────
🚄 交通
推荐：G2942 广州南 07:44 → 贵阳北 11:10
票价：¥406 | 用时：3h26min
🔗 12306搜索「广州南→贵阳北」
🏨 住宿
推荐区域：喷水池/大十字 | 舒适型酒店 ¥250~350/晚
🔗 携程搜「贵阳喷水池酒店」
🍜 午餐
推荐：亮欢寨酸汤鱼（飞山街店）| 人均¥80
🔗 高德导航「亮欢寨飞山街」
💰 Day 1 小计：¥723
───── 预算汇总 ─────
📊 总计：¥1,568/人 | 预算剩余：¥432| 情况 | 处理 | |:---|:---| | 预算过低 | 提示预算紧张，推荐省钱方案 | | 时间太短 | 减少景点，保证行程不赶 | | 带老人小孩 | 减少走路，安排午休 | | 季节不适合 | 提示替代目的地 | | 目的地不熟 | WebSearch先做功课 |
本 Skill 是 travel-agent-skill 项目的一部分。完整项目含网页版和城市数据： https://github.com/XiaoiYuyao/travel-agent-skill
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,667 | 23,744 | -4% | 1 | 1 | 0% | 4,280 | 1,691 | -60% | 0 | 0 | — |
case-02 | fail→pass | 20,476 | 22,737 | +11% | 1 | 1 | 0% | 3,781 | 5,016 | +33% | 0 | 0 | — |
case-03 | fail→fail | 28,940 | 13,856 | -52% | 1 | 1 | 0% | 5,352 | 2,077 | -61% | 0 | 0 | — |
case-04 | pass→pass | 16,114 | 17,371 | +8% | 1 | 1 | 0% | 2,924 | 4,000 | +37% | 0 | 0 | — |
case-05 | pass→fail | 15,489 | 17,267 | +11% | 1 | 1 | 0% | 2,434 | 1,657 | -32% | 0 | 0 | — |
case-06 | pass→pass | 7,342 | 9,028 | +23% | 1 | 1 | 0% | 1,173 | 2,244 | +91% | 0 | 0 | — |
case-07 | fail→pass | 14,092 | 4,409 | -69% | 1 | 1 | 0% | 2,096 | 1,658 | -21% | 0 | 0 | — |
case-08 | fail→pass | 26,552 | 34,441 | +30% | 1 | 1 | 0% | 3,739 | 5,646 | +51% | 0 | 0 | — |
case-09 | pass→pass | 23,703 | 25,178 | +6% | 1 | 1 | 0% | 3,270 | 3,682 | +13% | 0 | 0 | — |
case-10 | pass→pass | 19,358 | 17,996 | -7% | 1 | 1 | 0% | 2,824 | 3,758 | +33% | 0 | 0 | — |
case-11 | pass→pass | 21,796 | 25,133 | +15% | 1 | 1 | 0% | 3,166 | 4,783 | +51% | 0 | 0 | — |
case-12 | pass→fail | 25,665 | 32,384 | +26% | 1 | 1 | 0% | 5,087 | 7,165 | +41% | 0 | 0 | — |
case-13 | fail→fail | 21,725 | 8,604 | -60% | 1 | 1 | 0% | 3,578 | 1,537 | -57% | 0 | 0 | — |
case-14 | fail→pass | 23,166 | 33,670 | +45% | 1 | 1 | 0% | 3,698 | 5,645 | +53% | 0 | 0 | — |
case-15 | fail→pass | 19,085 | 16,558 | -13% | 1 | 1 | 0% | 3,017 | 4,056 | +34% | 0 | 0 | — |
case-16 | pass→pass | 21,606 | 23,083 | +7% | 1 | 1 | 0% | 3,476 | 4,600 | +32% | 0 | 0 | — |
case-17 | pass→pass | 23,581 | 24,918 | +6% | 1 | 1 | 0% | 3,369 | 4,627 | +37% | 0 | 0 | — |
case-18 | pass→pass | 24,809 | 24,917 | +0% | 1 | 1 | 0% | 3,309 | 4,705 | +42% | 0 | 0 | — |
case-19 | fail→pass | 30,248 | 38,501 | +27% | 1 | 1 | 0% | 4,598 | 6,129 | +33% | 0 | 0 | — |
case-20 | pass→fail | 23,749 | 11,083 | -53% | 1 | 1 | 0% | 3,577 | 1,690 | -53% | 0 | 0 | — |
case-21 | pass→pass | 22,826 | 21,737 | -5% | 1 | 1 | 0% | 3,799 | 5,000 | +32% | 0 | 0 | — |
case-22 | fail→pass | 12,776 | 12,988 | +2% | 1 | 1 | 0% | 2,142 | 3,143 | +47% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 17 counted toward the lift figure. The other 5 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +18 percentage points is the difference between those two pass rates over the 17 comparable cases. 6 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.