Done in one command: all-dimension cultural research → content distillation → org document + portable cards.
Methodology draws on archaeological Desk-Based Assessment (DBA): before arrival, exhaust all documentary evidence.
Mode
NATIVE mode enforced. This workflow is a multi-skill pipeline (Research → ContentAnalysis → ljg-card) and does not follow the Algorithm's seven-step process.
Parameters
| Parameter | Description | Example |
|---|---|---|
| City name | Required, target city | Xi'an, Luoyang, Datong |
-f |
Focus theme (optional) | -f 唐代 -f 石窟 -f 青铜器 |
-q |
Fast mode, skips content distillation, does only research + document |
Execution
1. Parse parameters
Extract the city name and optional parameters from the user message. If there is a focus theme, all subsequent searches revolve around that theme.
2. All-dimension research (Research extensive — single call, 12 Agents in parallel)
Call the Skill tool to execute Research, using extensive mode.
Core design: do not split into two steps of "knowledge base map" and "platform discovery" — they are different search angles of the same research operation. 12 Agents depart simultaneously, half doing academic/encyclopedic research, half doing platform content search.
Research outline (the prompt passed to Research):
对「{城市}」进行深度文化旅行研究。这不是旅游攻略，是出发前的考古学式案头研究（Desk-Based Assessment）。
研究覆盖以下维度，每个维度需要中英文双语搜索：
**维度 A — 历史分层**
该城市经历了哪些重要历史时期？每个时期在这座城市留下了什么物质遗存？朝代更迭如何影响城市格局？
**维度 B — 博物馆重点**
该城市有哪些重要博物馆？各博物馆的镇馆之宝和核心馆藏？哪些展品有重大考古意义？必须给出具体文物名称和展厅位置。
**维度 C — 古建遗存**
现存哪些重要古建筑和遗址？营造年代、建筑形制、结构特点。哪些是全国重点文物保护单位？看建筑时应关注什么细节（斗拱、彩画、碑刻等）？
**维度 D — 考古发现**
该城市及周边有哪些重大考古发现？出土文物现藏于哪些博物馆？发掘过程中有什么重要故事？
**维度 E — 人文脉络**
与该城市相关的重要历史人物、文学作品、文化传统。帮助理解这座城市的文化性格。
**维度 F — 深度内容发现**
搜索 B站（bilibili.com）、知乎（zhihu.com）、微信公众号（mp.weixin.qq.com）、抖音（douyin.com）、小红书（xiaohongshu.com）上关于该城市博物馆和古建的深度讲解内容。
筛选标准：
- 要：有知识增量的内容（讲背景、讲工艺、讲考古过程、讲建筑细节）
- 不要：纯打卡拍照、纯推荐无分析、广告软文
- B站视频优先 10 分钟以上的讲解类
- 公众号文章优先有参考文献或明确作者身份的
返回内容标题、URL、一句话摘要。
{如有聚焦主题：特别关注与「{聚焦主题}」相关的内容，其他维度作为背景补充。}
Wait for Research to complete and obtain the all-dimension research results.
3. Content distillation (ContentAnalysis — optional)
If the user specified -q fast mode, skip this step.
From the results returned in step 2, extract all valid URLs (article links, video links).
For each URL, start an Agent subagent in parallel:
Each subagent calls the Skill tool to execute ContentAnalysis, passing in the URL, using the fast depth level, to extract core knowledge points.
Degradation rules:
- If ContentAnalysis fails for a certain URL (inaccessible, no subtitles, etc.), skip that URL and do not block
- If all URLs fail, the process does not break — the research results from step 2 are already enough to generate the document
- ContentAnalysis is an enhancement layer, not a necessary layer
Collect all successfully distilled content summaries.
4. Synthesize an org-mode document
Synthesize step 2 (research results) and step 3 (content distillation, if any) into a structured org-mode document.
Document structure:
#+title: {城市}旅行研究
#+date: {当前日期}
#+filetags: :travel:museum:architecture:
* 城市概览
{城市}的文明坐标——为什么值得去，去了看什么。一段话勾勒这座城市在中国文明史中的位置。
* 历史分层
** {时期1}（{年代范围}）
核心事件、遗留痕迹、对应可看的实物。
** {时期2}
...
* 博物馆指南
** {博物馆1名称}
地址、开放时间、预约方式（如需要）。
*** 镇馆之宝
- {文物名}：{为什么重要} | 看什么细节：{具体观察点}
*** 重点展厅
- {展厅名}：{核心看点}
*** 容易错过的
- {被忽视但值得看的内容}
** {博物馆2名称}
...
* 古建遗存
** {古建1名称}（{朝代}，{保护级别}）
形制概述。
*** 看什么
- {具体观察点1}：{为什么值得注意}
- {具体观察点2}
** {古建2名称}
...
* 考古发现
** {遗址/发现1}
发现经过、意义、出土文物现藏地。如果有有趣的发掘故事，讲出来。
* 参观路线
** 路线一：{主题名}（{预计时间}）
适合谁：{描述}
1. {地点} → 重点看 {什么}（{停留建议时间}）
2. ...
** 路线二：{主题名}
...
* 深度内容推荐
从各平台发现的值得在出发前看的内容。
** 视频
- [[{URL}][{标题}]] — {一句话摘要}
** 文章
- [[{URL}][{标题}]] — {一句话摘要}
** 书籍（如有推荐）
- {书名} — {为什么值得读}
File naming: use the denote naming schema, save to the ~/Documents/notes/ directory:
{YYYYMMDDTHHMMSS}==z--{城市}旅行研究.org
Writing requirements：
- Every recommendation must have "why see it" and "what details to look at"; no vagueness allowed
- The tone is notes written for yourself, not tour-guide commentary
- Write exactly what there is exact information for; do not make up what there isn't
5. Forge portable cards (ljg-card)
Extract the core content from the org document in step 4, forge two cards,executed in parallel：
Card A — City civilization overview (infographic):
Call the Skill tool to execute ljg-card -i, with input content being: the essence summary of the city's historical layers + core museum list + must-see ancient architecture list. Understand the civilizational skeleton of this city in one image.
Card B — Quick reference for the visit route (long image):
Call the Skill tool to execute ljg-card -l, with the input content being: suggested visit route + the core highlights of each location. View anytime on your phone.
6. Summary report
════ 旅行研究完成 ═══════════════════════
🏛️ 城市: {城市名}
📝 知识文档: {org 文件路径}
🖼️ 文明概览卡: {PNG 文件路径}
🖼️ 路线速查卡: {PNG 文件路径}
📊 研究覆盖: {N}个博物馆 | {M}座古建 | {K}处考古遗址
📎 深度内容: {X}个视频 | {Y}篇文章
Key constraints
- Step 2 is the core — 12 Agents cover academic research and platform content in parallel, completed in one go
- Step 3 (content distillation) is an enhancement layer; failure does not block the process
- The two cards in step 5 run in parallel with each other
- The org document is the main output, and the cards are derivative outputs — document quality takes priority
- Do not produce generic travel guides; every recommendation must have "why see it" and "what details to look for"
- Research searches use bilingual Chinese and English keywords to expand coverage
- When there is no exact information, it is better to leave it blank than to fabricate
Known Pitfalls
(First creation, no records yet. Accumulate through use.)