← Back to Skills Marketplace
反向穷游
by
hello_hang
· GitHub ↗
· v1.0.1
· MIT-0
104
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install flyai-reverse-budget
Description
反向穷游助手，我有X元能去哪？输入总预算、出发城市、可出行天数，AI反向搜索所有可能性，生成3档方案（省钱极限版/舒适版/极限拉满版），每档含完整预算拆解。当用户提到"我有X元"、"预算XX能去哪"、"这些钱能去哪"、"反向穷游"、"穷游推荐"、"预算旅行"时使用。
README (SKILL.md)
反向穷游 — 我有 X 元，能去哪？
你是一个能够自主学习、持续成长的反向旅行规划大师。传统逻辑是「选目的地 → 查价格」，你反过来——「定预算 → 反推目的地」。
核心定位
FlyAI 能力
完整命令参考见 reference 目录
本技能主要使用：search-flight、search-hotel、search-poi
预算确定感专家：
- 🎯 反向思维：用户不用想去哪，只要掏出钱包说"我就这么多钱"
- 🔍 全面搜索：AI 反向搜索所有可能性，告诉用户"你的钱能去这些地方，能玩这么好"
- 💡 预算拆解：精确到机票+酒店+景点+餐饮+交通，让每分钱花得明明白白
- 🧬 记忆成长：记住用户的偏好和历史方案，持续提供个性化推荐
Memory 系统
作为一个能持续成长的智能助手，我会记住你的风格和偏好。
核心要点：
- 启动时读取：除非用户说"忽略偏好/换个风格"
- 有记录：直接用已保存的偏好开始对话
- 无记录：首次用户，收集基本信息
- 实时更新：用户提到出发城市、偏好、同行人时更新 Memory
工具说明
用户画像读取（双模式）
启动时读取用户历史偏好，减少重复询问。
优先：search_memory(query="用户旅行画像", category="user_hobby", keywords="flyai")
降级：read_file(file_path="~/.flyai/user-profile.md")
工作流程
详细步骤见 reference/workflow.md
核心阶段：
- 收集预算信息 - 获取用户预算/天数/出发城市
- 反向搜索 - 调用 FlyAI 搜索符合预算的目的地
- 生成三档方案 - 省钱极限版/舒适版/极限拉满版
- 预算拆解 - 精确到机票+酒店+景点+餐饮
- 输出方案 - 含预订链接+完整费用明细
现实约束与失败处理
| 情况 | 处理方式 |
|---|---|
| 预算太低，哪也去不了 | 诚实告知"该预算建议周边自驾游"，推荐省内/周边方案 |
| 预算太高，选择太多 | 用 ask_user_question 追问偏好（海岛/城市/自然）缩小范围 |
| 某档方案超预算 | 给出"如果多加¥X就能解锁XX"的升级提示 |
| 价格波动导致超预算 | 标注"按当前价格计算，建议尽快预订锁价" |
| 搜索返回空结果 | 调整搜索条件，或用 AI 通识知识推荐替代目的地 |
| SSL 证书验证失败 | 确保命令前加 NODE_TLS_REJECT_UNAUTHORIZED=0 |
常见预算档位参考
| 预算（人均） | 推荐方向 | 典型目的地 |
|---|---|---|
| 1000以内 | 周边自驾/高铁 | 省内城市、周边古镇 |
| 1000-2000 | 国内短途 | 长沙、成都、重庆、西安 |
| 2000-3000 | 国内中途 | 厦门、青岛、丽江、大理 |
| 3000-5000 | 国内远途/东南亚 | 三亚、新疆、泰国、越南 |
| 5000-8000 | 东南亚深度/日韩 | 日本、韩国、新马泰 |
| 8000+ | 欧美/海岛 | 欧洲、马尔代夫、澳新 |
自主学习机制
示例对话
用户偏好保存（双模式）
发现新偏好时提示保存。详见 reference/user-profile-storage.md
保存流程：发现偏好 → 提示确认 → Qoder用update_memory / 非Qoder更新本地文件
Usage Guidance
This skill matches its stated travel-planning purpose, but it contains actionable instructions that affect system security: it tells the agent to install a global npm CLI (may require sudo) and repeatedly recommends disabling TLS certificate verification (NODE_TLS_REJECT_UNAUTHORIZED=0). If you plan to use it, consider: 1) refuse to run the TLS bypass — it makes network calls susceptible to MITM; 2) prefer installing the FlyAI CLI yourself after reviewing the package on npm/GitHub rather than letting the agent run global installs; 3) be aware the skill will read and write ~/.flyai/user-profile.md (and will attempt to use any platform memory APIs), so do not store sensitive secrets in that file; 4) if you are uncomfortable with automatic network installs or file writes, decline or sandbox the skill. If you want a safer setup, request a version that omits TLS bypasses and that documents exactly which external packages/hosts it will call.
Capability Analysis
Type: OpenClaw Skill
Name: flyai-reverse-budget
Version: 1.0.1
The skill bundle contains high-risk instructions that introduce significant security vulnerabilities and potential supply chain risks. Most notably, workflow.md directs the agent to globally install an NPM package (@fly-ai/flyai-cli) which differs from the package name mentioned in tools.md (@anthropic-ai/flyai-cli), a common indicator of typosquatting. Furthermore, the instructions in SKILL.md and workflow.md explicitly mandate disabling SSL certificate verification (NODE_TLS_REJECT_UNAUTHORIZED=0) for all network commands, exposing the agent to Man-in-the-Middle (MitM) attacks. While these actions are framed as setup and troubleshooting steps for a travel assistant, the combination of a suspicious global installer and the intentional bypass of core security protocols warrants a suspicious classification.
Capability Assessment
Purpose & Capability
Name/description (reverse-budget travel) align with the instructions: it calls flight/hotel/poi search commands, builds 3-tier budgets, and stores user preferences. Reading/writing a user profile and using search-* tools is consistent with the stated functionality.
Instruction Scope
Instructions explicitly tell the agent to bypass TLS verification (NODE_TLS_REJECT_UNAUTHORIZED=0) and to run global npm installs and CLI commands. They also mandate reading and updating user profile data on startup (via search_memory/update_memory or ~/.flyai/user-profile.md). The TLS bypass and automated global installs broaden the agent's scope beyond pure recommendation logic.
Install Mechanism
Although the package has no formal install spec, the workflow requires running `npm install -g @fly-ai/flyai-cli@latest` or using `npx @anthropic-ai/flyai-cli@latest`. This downloads and executes code from the public npm registry and may require sudo; it's a moderate-to-high risk action not enforced/declared by the registry metadata.
Credentials
No credentials are requested (good), but the SKILL.md relies on an insecure environment variable (NODE_TLS_REJECT_UNAUTHORIZED=0) multiple times which weakens network security. It also reads/writes a local file (~/.flyai/user-profile.md) and attempts to use platform memory APIs — these are proportional to personalization but are not declared in requires.env and involve reading/writing user data.
Persistence & Privilege
Skill does not request always: true and does not modify other skills. It asks to persist a profile file in the user's home (~/.flyai/user-profile.md) and to use platform memory when available; persisting personal preferences is expected for personalization but grants ongoing local storage access.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install flyai-reverse-budget - After installation, invoke the skill by name or use
/flyai-reverse-budget - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
**Changelog for flyai-reverse-budget v1.0.1**
- Added extensive reference documentation in the `reference/` directory, covering workflows, tool usage, search commands, and example interactions.
- Modularized the main SKILL.md by moving detailed explanations and technical steps into separate reference files for clarity and maintainability.
- Updated skill to support dual-mode user preference storage (memory API and local file fallback).
- Improved documentation and guidance for workflow, user info collection, and error handling.
- Enhanced maintainability and user experience by referencing detailed procedures and examples from external docs.
v1.0.0
flyai-reverse-budget v1.0.0
- 新增反向穷游智能助手，用户只需输入预算、出发城市和出行天数，即可自动生成3档旅行方案（省钱极限版/舒适版/极限拉满版），每档含完整预算拆解。
- 实现一站式反向搜索，自动调用机票、酒店、景点、美食、交通等多渠道数据，输出预算内最优推荐。
- 内置记忆成长机制，自动记忆和更新用户偏好与历史选择，提升个性化推荐能力。
- 提供交互式提问工具，智能补全用户需求，提升使用体验。
- 配备失败场景处理和风险提示，确保方案实用性和用户体验。
Metadata
Frequently Asked Questions
What is 反向穷游?
反向穷游助手，我有X元能去哪？输入总预算、出发城市、可出行天数，AI反向搜索所有可能性，生成3档方案（省钱极限版/舒适版/极限拉满版），每档含完整预算拆解。当用户提到"我有X元"、"预算XX能去哪"、"这些钱能去哪"、"反向穷游"、"穷游推荐"、"预算旅行"时使用。 It is an AI Agent Skill for Claude Code / OpenClaw, with 104 downloads so far.
How do I install 反向穷游?
Run "/install flyai-reverse-budget" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is 反向穷游 free?
Yes, 反向穷游 is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does 反向穷游 support?
反向穷游 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created 反向穷游?
It is built and maintained by hello_hang (@hello-ahang); the current version is v1.0.1.
More Skills