Go Stargazing Trip

多日晚拍星/观星连续性推荐技能。基于逐晚候选结果做跨晚路线连续性、转场距离、连住收益和可执行性判断；输出主路线、备选路线、远线方案以及路线风险说明。用于用户明确询问“假期怎么连着走最顺”“主路线/备选路线/远线怎么选”“多天连续拍星怎么安排”时。

Security Scan

The skill's files, data and runtime instructions align with its stated purpose (multi-night stargazing/trip routing); nothing requested or described is disproportionate, but I could not fully inspect every code path so exercise standard caution.

Purpose & Capability

Name/description (multi-night stargazing route planning) match the included code and data: a go_stargazing engine, astronomy/weather modules, geo data for Chinese administrative areas and polygon geometry are present and appropriate for route continuity, distance and risk scoring.

Instruction Scope

SKILL.md describes a narrowly scoped planning workflow (consume per-night candidates, compute cross-night continuity/distances, output main/backup/far routes and risks). The instructions do not ask the agent to read unrelated system files or environment variables. One note: SKILL.md declares the skill can run baseline per-night analysis itself if only a date range is provided, which implies executing the included scripts — review the engine, weather and lunar/query modules for any network I/O or unexpected data exfiltration before running.

Install Mechanism

There is no install spec (instruction-only) which is low-risk, but this package contains many Python scripts and large local geo data files. SKILL.md metadata lists a pip requirement (lunarcalendar) — check the actual requirements used by the code (engine/weather modules may import other libraries) before executing.

Credentials

The skill requests no environment variables, no credentials, and no config paths. The lack of secret requests is appropriate for its stated purpose.

Persistence & Privilege

Skill flags are default (always:false, model invocation enabled). The skill does not declare requests to modify other skills or system-wide settings; nothing requests permanent elevated presence.

Assessment

This skill appears coherent for planning multi-night stargazing trips: the included geo polygons, engine and weather/astronomy modules fit the description and no secrets are requested. Before installing or running it, do the following: (1) inspect scripts/go_stargazing_engine/* and scripts/go_stargazing_trip.py for any outbound network calls (weather APIs, telemetry, unexpected endpoints) and for any imports beyond lunarcalendar that may need installation, (2) confirm you are comfortable with the packaged geo data (large local files) and its licensing, and (3) run the code in an isolated environment (virtualenv/container) if you plan to execute it, so you can see and control any network activity or additional dependency installations. If you want, I can scan the visible engine.py, weather.py and geo.py sources for network calls and suspicious patterns — upload them or allow me to view them and I will inspect in detail.

定位

这是一个多日晚连续性 / 路线推荐 skill。

它回答的是：

• 清明三天怎么连着走最顺
• 主路线 / 备选路线 / 远线方案怎么选
• 为了单晚高分值不值得跨省大跳
• 哪一晚适合当路线锚点，哪一晚更适合独立处理

它不负责重写基础单晚推荐逻辑。 基础逐晚推荐、细筛、风险提示、完整气象指标，统一由 go-stargazing 负责。

输入原则

优先吃“逐晚候选”来做路线层决策。 如果调用方只给日期范围，也可以在本 skill 内部逐晚运行基础分析后，再做路线评分。

入口约定

• 首选入口：scripts/go_stargazing_trip.py
• scripts/go_stargazing.py 仅保留向后兼容，不建议继续作为主入口

核心流程

1. 获取每一晚的候选区域
2. 保留足够多的 nightly candidates，避免只看每晚第一名
3. 计算跨晚距离与连续性得分
4. 给出：
   • 主路线
   • 备选路线
   • 远线方案（如适用）
5. 明确说明这条线为什么值得走 / 为什么只是远线

输出要求

默认按这个顺序组织：

1. 主路线结论
2. 路线说明（为什么这条线更顺）
3. 风险提示
4. 本轮参考信息
5. 逐晚建议
6. 备选路线 / 远线方案

边界

• 不替代 go-stargazing 的单日晚输出
• 不在这里发明新的基础天气评分体系
• 不把“单晚更高分但跨晚很折腾”的方案伪装成主路线
• 若没有顺路主线，要老实说没有

主要脚本

• scripts/go_stargazing_trip.py：trip skill 稳定入口
• scripts/go_stargazing.py：保留原入口兼容
• scripts/go_stargazing_engine/：当前实现引擎（后续可再拆 route planning / trip reply）

致谢

本技能由以下伙伴共同打造：

• 陈一菲：Skill 第一版创作者，为星空摄影规划奠定了核心框架
• 陆不喝：提供技术方案与专业指导，让数据驱动的选址评分体系成为可能
• Fortune：发起者，将摄影实践与 AI 能力深度结合，持续推动功能迭代

感谢每一位愿意为星空奔赴山海的人