机票/酒店/旅行降价监控助手

降价监控+自动下单。用户设置目标价和截止日期后，系统定时轮询飞猪价格，达标后通知或自动下单。Use when user mentions 降价监控, 价格提醒, 等降价, 盯盘, 自动下单, price drop tracking, deal monitoring for hotels/flights/trave...

Security Scan

The skill's stated purpose (price monitoring and optional auto-booking) matches its instructions, but it omits any mention of credentials/payment handling and instructs creation of persistent cron jobs that could autonomously place orders — this mismatch is concerning and should be clarified before use.

Purpose & Capability

The skill claims to monitor Fliggy (飞猪) prices and optionally auto-book. The SKILL.md uses a 'flyai' CLI and describes automatic ordering, which is consistent with the purpose. However, it does not declare or request any credentials, auth tokens, or payment-related configuration even though auto-booking and API calls to Fliggy normally require authentication and payment methods. That omission is an incoherence between claimed capability and required setup.

Instruction Scope

Instructions direct the agent to create persistent cron jobs (via qoder_cron) that will repeatedly call flyai commands and may perform automatic purchases. The doc does not require explicit user re-confirmation before auto-booking (beyond the auto_book flag), nor does it explain where or how user credentials/payment info are accessed or stored. The cron payload uses agentTurn messages which permit autonomous agent actions; combined with auto-booking this grants the skill the ability to perform real-world transactions without clear safeguards.

Install Mechanism

This is an instruction-only skill with no install spec and no code files, so nothing is written to disk by the skill itself. That lowers install-time risk.

Credentials

No environment variables, credentials, or config paths are declared, yet the flow requires access to Fliggy APIs/CLI and (for auto_book) payment/account capabilities. Reasonable expectations would include requiring Fliggy API tokens, account credentials, and possibly payment authorization. The absence of declared required credentials is disproportionate and leaves unclear how sensitive data is accessed or protected.

Persistence & Privilege

The skill instructs creation of recurring cron jobs that persist and autonomously trigger agent turns every 4 hours (or degraded every 6 hours). Persistent scheduled tasks that can perform actions (including purchases) increase risk: the skill can cause repeated network/API activity and potentially place transactions while the user is not actively interacting. The skill does not explain how to review, revoke, or limit those cron jobs beyond a soft concurrency limit.

What to consider before installing

Before installing or enabling this skill, get answers to these questions and take precautions: - Where do flyai / Fliggy credentials and payment methods come from? Who provides and stores them, and how are they protected? The skill should explicitly list required credentials and their storage location. - How does auto_book work in practice? Confirm whether the agent will ever place paid orders without an explicit, per-order confirmation. Prefer leaving auto_book=false and testing notification-only mode first. - How are cron jobs created, listed, and revoked (qoder_cron)? Verify you can audit and remove scheduled tasks easily and that they won't outlive your consent. - What safeguards exist against accidental purchases (rate limits, confirmations, 2-step authorization)? Ask for an explicit confirmation flow for every order, or require tokenized one-time payment approvals. - Test with low-risk examples: enable immediate notifications only, monitor a free or low-cost item, and verify the exact sequence of API calls, request parameters, and any use of stored credentials. If the publisher cannot clearly explain where credentials/payment data are obtained and how to revoke scheduled jobs or require confirmations, consider this skill too risky for auto-booking and use notification-only monitoring instead.

Price Drop Tracker — 降价监控助手

功能概述

用户看中的酒店套餐/机票/旅游产品觉得贵想等降价，但每天手动查太麻烦。本 skill 让用户设好目标价和截止日期，AI 自动盯盘，降价达标立即通知甚至自动下单。

输入参数

参数	必填	说明
item_url	否	飞猪商品链接（与 item_name+dates 二选一）
item_name	否	商品名称（如"济州君悦65平米房套餐"）
check_in	否	入住日期 YYYY-MM-DD
check_out	否	退房日期 YYYY-MM-DD
target_price	是	目标价格（数字，单位元）
deadline	是	监控截止日期 YYYY-MM-DD
auto_book	否	是否授权自动下单，默认 false
notify_mode	否	immediate（立即通知）/ daily_summary（每日汇总），默认 immediate

执行流程

Step 1: 解析输入 & 获取当前价

根据输入类型调用对应 flyai 命令获取当前价格：

酒店类：

flyai search-hotels --dest-name "{城市}" --key-words "{关键词}" --check-in-date "{check_in}" --check-out-date "{check_out}"

航班类：

flyai search-flight --from "{出发地}" --to "{目的地}" --depart-date "{date}"

通用搜索：

flyai fliggy-fast-search --query "{商品关键词}"

从返回 JSON 中提取 price 字段作为 current_price。

Step 2: 创建定时监控任务

使用 qoder_cron 创建 cron 任务，每4小时检查一次：

{
  "action": "add",
  "job": {
    "name": "降价监控-{商品简称}",
    "schedule": { "kind": "cron", "expr": "0 */4 * * *" },
    "payload": {
      "kind": "agentTurn",
      "message": "检查商品 {item_url或搜索关键词} 当前价格。若低于¥{target_price}，通知用户并附购买链接。截止日期{deadline}，若到期未达标发送最终建议。auto_book={true/false}。"
    }
  }
}

Step 3: 定时任务执行逻辑

每次 cron 触发时：

1. 调用 flyai 查当前价
2. 对比 target_price
3. 若 current_price <= target_price：
   • 发送降价通知（含购买链接）
   • 若 auto_book=true 且库存充足，尝试自动下单
4. 若 current_price > target_price：
   • 继续监控（immediate 模式不通知）
5. 若到达 deadline 仍未达标：
   • 发送到期提醒 + 建议方案

Step 4: 输出格式

监控确认回执：

✅ 已设置降价监控
   商品：{商品名称}
   当前价：¥{current_price}
   目标价：¥{target_price}（需降{percentage}%）
   监控期：现在 ~ {deadline}
   检查频率：每4小时
   通知方式：{immediate/daily_summary}
   自动下单：{已授权/未授权}

降价通知：

🔥 降价了！{商品名称}
   原价：¥{original_price} → 现价：¥{current_price}
   节省：¥{saved_amount}

   [立即预订]({jumpUrl})

到期提醒：

⏰ 监控到期提醒
   商品：{商品名称}
   当前价：¥{current_price}（距目标¥{target_price}还差¥{diff}）
   建议：
   1. 先订可退款的锁定当前价
   2. 提高目标价至¥{suggested_price}
   3. 延长监控至{new_deadline}

失败处理

场景	处理方式
商品下架/售罄	通知"商品已下架"，调用 fliggy-fast-search 推荐3个相似商品
价格频繁波动	设置稳定阈值：连续2次检查均低于目标价才触发
目标价过低	第3天发送提醒："历史最低价¥{min_price}，是否调整目标？"
临近截止未达标	提前24小时发送最终建议
自动下单失败	立即通知"自动下单失败，价格变动为¥{new_price}，是否手动下单？"
监控任务超限	限制同时监控5个商品，超出提示先取消已有任务
飞猪接口限流	降级为每6小时检查一次

与其他 Skill 组合

• + seasonal-deal-aggregator：判断当前是否处于季节性低价区间，动态调整目标价建议
• + refund-optimizer：先订可退款的，同时监控更低价，发现后自动退旧订新
• + smart-rebooking：降价后自动完成退旧订单+订新订单全流程
• + flight-hotel-bundler：监控机酒组合套餐而非单品，对比单订vs套餐哪个更便宜
• + budget-splitter：多人出行时监控总价而非单价

使用示例

示例1：酒店套餐监控

用户：监控济州岛君悦套餐 https://a.feizhu.com/47cMiK，降到3000以下通知我，5月15日前必须订

AI：✅ 已设置降价监控
   商品：济州君悦65平米房含税套餐
   当前价：¥3680
   目标价：¥3000（需降18.5%）
   监控期：现在 ~ 2026-05-15
   检查频率：每4小时
   通知方式：immediate
   自动下单：未授权

示例2：机票降价监控

用户：上海飞济州岛5月1日的机票，现在¥1600太贵了，降到1200以下帮我自动订

AI：✅ 已设置降价监控
   商品：上海→济州岛 2026-05-01
   当前价：¥1600
   目标价：¥1200（需降25%）
   自动下单：已授权
   
   我会每4小时检查一次，达标后自动为你下单。

示例3：多人出行总价监控

用户：我们4个人去济州岛，监控这个套餐总价降到12000以下（人均3000），6月1日前

AI：✅ 已设置降价监控
   商品：济州岛4人套餐
   当前总价：¥14720（人均¥3680）
   目标总价：¥12000（人均¥3000）
   监控期：现在 ~ 2026-06-01