xiaohongshu-search

Search for Viral Xiaohongshu Posts

1. Overview

Search for popular and viral Xiaohongshu posts by keyword, with recommendations ranked by relevance, popularity, and recency. The tool also recommends trending posts and niche topics, helping creators, brands, and MCN agencies discover trends and creative inspiration. Run this tool only in the main Agent; do not delegate to a sub-Agent.

2. Features

🔍 Intelligent keyword search — Supports exact keyword searches, multiple comma-separated keywords, and sitewide trending queries with an empty keyword.
📊 Three-dimensional ranking — Keyword searches use a weighted total of relevance (up to 10 points), popularity (up to 3), and recency (up to 2), for a maximum of 15 points. Sitewide trending results are ranked by interactions.
🧠 Detailed intent understanding — Extracts niche terms from the user's description first, identifies broad terms, and automatically recommends 10 narrower directions.
⏱️ Flexible date range — Defaults to the past 7 days and expands the range when data is insufficient (1 day → 3 days → 7 days → 30 days). Data is updated daily at 7 a.m.
🔥 Trending-post recommendations — Automatically shows recent popular posts and trending hashtags when few results are available.
📈 Niche-topic guidance — Proactively recommends 10 related niches after each query to support further exploration.
🏷️ Related-keyword recommendations — The script returns a relatedSearches field, which is used to display related search suggestions automatically.
📩 Scheduled subscription delivery — Supports calendar subscription tasks that deliver the latest trending posts at a specified time.
📄 HTML report generation — Automatically generates a {keyword}_热门数据.html visualization file.
🛡️ Explicit data notes — Trending posts must have 1,000+ interactions. Data notes and the ranking criteria are shown at the top.

3. One-click Installation

Authentication

Get an API Key

Visit Redfox Hub to obtain an API KEY.

Configure the API Key

Option 1: Using OpenClaw as an example, add REDFOX_API_KEY to ~/.openclaw/openclaw.json:

{ "env": { "REDFOX_API_KEY": "ak_xxxx..." } }

Option 2: Terminal configuration

export REDFOX_API_KEY="ak_xxxx..."

Dependency installation

This Skill uses the Python 3 standard library and requires no additional third-party dependencies. Make sure Python 3.x is installed.

Environment variable configuration

Environment variable	Description	Required?	How to obtain it
`REDFOX_API_KEY`	Redfox Data API Key	Yes	Redfox Hub

4. Usage Guide

⚠️ Core execution rules (mandatory)

Ask before querying a broad term: When a broad term is detected,never call the script directly. First recommend narrower terms and wait for the user's selection before querying.
Correct execution order: Extract keywords → determine whether they are broad → ask the user if they are → call the script after the user replies.
Mandatory waiting rule: After recommending narrower terms, stop and wait for the user's next message saying “拓展” (expand) or “不拓展” (do not expand).Do not continue with any script call in the same turn.

Common broad terms: Broad terms are highly abstract, wide-ranging categories without specific context or attribute modifiers, such as industry classifications containing multiple subcategories. Characteristics: (1) They are umbrella concepts, such as “beauty” covering “foundation/lipstick,” “sports” covering “running/yoga,” or AI. (2) They often appear with generalizing words such as “field” or “type,” as in “the beauty field” or “types of sports.”

Common specific terms: Specific terms are less abstract, clearly targeted instances with contextual or attribute modifiers, usually direct subcategories of a broad term. Characteristics: (1) They are subordinate concepts, such as “foundation” under “beauty products” or “ketogenic diet” under “dietary approaches.” (2) They often contain modifiers, such as “spring” → “spring outfits” or “ketogenic” → “ketogenic diet.”

Basic usage (query in 3 steps)

Step 1 — Extract keywords: Extract search keywords from the user's natural-language description. Prioritize niche terms with specific context or attribute modifiers over broad categories.

Step 2 — Call the script：

python scripts/fetch_xhs_hot_articles.py --keyword <关键词> --start-date <日期>

With a niche keyword:python scripts/fetch_xhs_hot_articles.py --keyword <关键词> --start-date <日期>
Without a niche keyword (sitewide trending query):python scripts/fetch_xhs_hot_articles.py --keyword "" --start-date <日期>
Separate multiple keywords with commas:python scripts/fetch_xhs_hot_articles.py --keyword "减脂餐,职场穿搭,健身" --start-date <日期>
Pagination parameters:--page-num 1 --page-size 50

Step 3 — View the results: The script returns structured JSON. Display the results according to this guide's presentation strategy.

Advanced usage

Understand user intent (before querying)

⚠️ Core rule: Interpret the meaning and prioritize niche terms from the user's description over broad categories.

1. Determine whether the user mentioned a niche keyword：

No niche keyword(e.g., “What posts are trending recently?”, “What content is popular lately?”, or “Show me trending data”) → call the script directly with an empty keyword string ""to query sitewide trending posts.
A niche keyword is present → continue extracting and classifying it.

2. Extract precise search keywords(only when the user mentions a topic):

Analyze the user's description: Extract clear niche terms from the user's introduction or requirements.
Example analysis:
- User input: “I'm an arts-focused Xiaohongshu creator with over 10,000 followers. I usually post about niche-film aesthetics, book reviews, music reviews, and Hong Kong and Taiwan culture. Help me find trending film topics.”
- Analysis: The user mentioned these niches: niche films, book reviews, music reviews, and Hong Kong and Taiwan culture.
- Relating the earlier context to “film” gives these narrower terms: niche films, Hong Kong and Taiwan films, and film-music reviews.
- Search keywords: niche films, Hong Kong and Taiwan films, film-music reviews.
- ❌ Incorrect: Search using only the broad term “film.”

3. Classify the keyword(only after extracting a keyword):

Specific term / vertical niche(a term with contextual or attribute modifiers, such as “work outfits,” “weight-loss meals,” or “outfits for petite people”) → search directly without asking to expand.
Broad term / category(an unmodified general category, such as “outfits,” “food,” or “beauty”) → apply the expansion strategy.
Classification principle: A modifier describing context, audience, style, or intent indicates a specific term: search directly. No modifier indicates a broad term: expansion is needed.

Broad-keyword expansion strategy

Broad-term workflow (⚠️ wait for an explicit user reply before calling the script!)：

Step 1: Generate narrower terms(do not call the script to search for data).

Principles for generating expanded terms:
- Use an appropriate level of specificity: Avoid combining too many terms or becoming overly narrow; for example, “middle-class outfits” may be too specific to return data.
- Cover different contexts: Include 2–3 terms each for trends, audiences, contexts, and intent.
Example output:
```
我识别到「中产」是较大的分类，已查询近期热门趋势，推荐以下细分方向：
老钱、轻奢、品质生活、松弛感、高级感穿搭、体面、法式穿搭、律师、医生、品质家居
回复「拓展」将同时搜索这10个词，回复「不拓展」将继续搜索「中产」
```
Step 2: Wait for the user's reply
- ❌ Prohibited: Calling the script before the user replies.
- ✅ Correct: Proceed only after the user explicitly replies “拓展” (expand) or “不拓展” (do not expand).
Step 3: Act on the explicit reply
- If the user replies “拓展” (expand) → call the script to search 10 narrower terms.
- If the user replies “不拓展” (do not expand) or “继续” (continue) → call the script with the original keyword.
- If the user has not replied or replies otherwise → identify the corresponding intent.

Date ranges and data queries

Date range：

The database contains data only from yesterday back to 30 days ago.
Default meaning of “recent”: The past 7 days (startDate = today - 7 days).
Date calculation(convert the user's wording into startDate):
- Today: Use yesterday's date directly; startDate = yesterday.
- Recent / past 7 days: startDate = today - 7 days.
- Past N days: startDate = today - N days.
- Example: The user says “past 15 days” → startDate = today - 15 days.

Automatic adjustment for insufficient data (⚠️ expand the date range first; never change keywords!)：

Handling principle: When data is insufficient, expand only the date range; do not replace or expand the keywords.
Adjustment order: Automatically expand the date range in this order:
1. Past 1 day → past 3 days
2. Past 3 days → past 7 days
3. Past 7 days → past 30 days
Inform the user: When adjusting automatically, say: “This keyword returned little data for the past X days, so the date range has been expanded to the past Y days.”
Prohibited behavior: ❌ Do not change keywords, recommend other terms, or trigger broad-term expansion merely because data is insufficient.

Apology for out-of-range or not-yet-updated data：

When the user says “today,” reply: “Sorry, today's data has not been updated yet. Here is the most recent available data.”
When the requested range exceeds 30 days, reply: “Sorry, only the past 30 days are supported. Here is the closest available data.”

Output files：

Filtered recommendation data:{keyword}_热门数据.html

Preliminary notes (must be shown before the data)

Data note: Trending posts are posts with more than 1,000 interactions. Yesterday's data is updated daily at 7 a.m. Interaction counts reflect the time of database ingestion, rather than live data; they may continue growing afterward.
Ranking note(keyword search): A weighted total combines relevance (up to 10 points), popularity (up to 3), and recency (up to 2), for a maximum of 15 points.
Ranking note(sitewide trending / no keyword): Results are ranked by interactions, with no scoring fields.

Data presentation strategy (core)

⚠️ Mandatory output rules：

✅ Follow this step's output format exactly.
❌ Do not add analysis or interpretation before the output.
❌ Do not invent advice or proposals.
❌ Do not ask about the user's underlying purpose or requirements.
✅ Read the JSON returned by the script and display it using the corresponding strategy.

Data field descriptions：

articles: Regular post data (the main content displayed).
latestHotArticles: Recommended trending posts (supplementary content; show 10 by default, with no scoring fields in the table).
hotTopics: Trending topics (returned by the API for reference; do not show them in the conversation).

A. articles count ≥ 10

Display:

Date-range note: State the queried range, e.g., “📅 Query date range: May 8–May 19.”
Regular posts (descending totalScore for keyword searches; descending interactions for sitewide trending queries).
Related-keyword recommendations (relatedSearches).

Markdown table format：

⚠️ Show table fields in exactly this order:：

Note: Show the total-score field in bold (using**分数**formatting).

Example (with a keyword): 📅 Query date range: May 8–May 19

Post title	Author	Interactions	Published at	Relevance	Popularity	Recency	Total score
Essential for New Employees: 5 Tips for Quickly Fitting into a Team	职场成长社 (author)	10.0w	2026-05-15	9.8	3.0	2.0	14.8

Example (sitewide trending): 📅 Query date range: May 8–May 19

Post title	Author	Interactions	Published at
Trending post title	Author name	10.0w	2026-05-15

🔤 Related-keyword recommendations: workplace communication, career advancement, employees.

B. 0 < articles count < 10

Display:

Date-range note: State the queried range, e.g., “📅 Query date range: May 8–May 19.”
Notice: “💡 Only X results were found for this keyword and date range. You can expand the terms or date range. We have also recommended recent trending posts.”
Regular post data
Related-keyword recommendations (relatedSearches)
Recommended trending posts (latestHotArticles, in a section titled “Recommended trending posts”; show 10 by default).
Recommended trending topics

Markdown format example：

📅 Viral posts must have more than 1,000 interactions. Query date range: May 8–May 19

Only 3 results were found for the current keyword and date range:

With a keyword:

Post title	Author	Interactions	Published at	Relevance	Popularity	Recency	Total score
Post 1	Author 1	10.0w	2026-05-15	9.8	3.0	2.0	14.8

Sitewide trending (no scoring fields):

Post title	Author	Interactions	Published at
Post 1	Author 1	10.0w	2026-05-15

🔤 Related-keyword recommendations: workplace communication, career advancement, employees.

💡 Here are some recent trending posts for reference that may be helpful.：

⚠️ Recommended trending-post tables need no scoring fields. Use this format:

Post title	Author	Interactions	Published at
Trending post 1	Author A	8.5w	2026-05-14
Trending post 2	Author B	6.2w	2026-05-13

📈 You can also search these trending topics:：

Outfits, food, makeup, film and television, careers, pets, home, travel, transport, hobbies, technology, the internet, healthcare, astrology and relationships, weddings, photography, education, parenting, personal care, fashionable shoes and bags, lifestyle, scientific exploration, news, and sports.

C. articles count = 0

⚠️ Use the following format exactly. Do not add unsolicited advice or analysis.

Display:

**🔍抱歉，爆款笔记收录原则为互动数1000以上的文章,该搜索词在查询时间范围（5月8日 - 5月19日）内太小众，未找到与"XXX"直接相关的内容，你可以尝试用更短/宽泛的关键词重试。**

**推荐搜索词**：从脚本返回的relatedSearches字段中提取推荐词，以加粗形式展示；

💡我们为您推荐了近期的热门笔记供参考，或许对您有帮助：

⚠️ 推荐热门笔记表格不需要评分字段，格式为：
| 笔记标题 | 作者 | 互动数 | 发布时间 |
|---------|------|--------|----------|
| [热门笔记1](url) | [作者A](https://www.xiaohongshu.com/user/profile/xxx) | 8.5w | 2026-05-14 |
| [热门笔记2](url) | [作者B](https://www.xiaohongshu.com/user/profile/xxx) | 6.2w | 2026-05-13 |

📈 您还可以尝试搜索以下热门赛道：
穿搭、美食、彩妆、影视、职场、萌宠、家居、旅行、交通、兴趣、科技、互联网、医疗保健、星座情感、婚庆婚礼、拍摄、教育、亲子育儿、个人护理、潮流鞋包、生活、科学探索、新闻资讯、运动

Output rules：

✅ Output the format above directly.
❌ Do not add extra analysis or advice.
❌ Do not explain why no data was found.
❌ Do not proactively offer other search approaches.
❌ Do not ask about the user's underlying purpose.

Presentation logic (pagination)

When articles count > 10：

Initially show only the first 10 records by default.
Prompt the user:

💡 当前共找到 X 条相关笔记（X为实际返回条数），已展示前10条。是否需要查看全部？

Wait for the user's reply:

If the user replies “是” (yes), “查看全部” (view all), or “全部” (all) → show all records.
If the user replies “否” (no), “不用” (no need), or “不需要” (not needed) → do not show more.

Format for displaying all records：

📊 全部结果（共X条，X为实际返回条数）：

有关键词时：
| 笔记标题 | 作者 | 互动数 | 发布时间 | 相关性 | 热度 | 时效 | **总分** |

全站热门时：
| 笔记标题 | 作者 | 互动数 | 发布时间 |

| ...（全部数据）...

Ask about subscriptions (mandatory)

When articles count > 0, ask after displaying the results.：

📬 订阅服务
1️⃣ 是否需要订阅当前搜索条件笔记，订阅后将定时推送给您？
2️⃣ 暂不需要

Handle the user's reply：

If the user selects 1️⃣ → use the calendar_create tool to create a calendar event subscribing to the current search conditions.
If the user selects 2️⃣ → end the current conversation.

Subscription implementation steps：

State the data update time and ask for the delivery time.：

📅 数据更新时间：每日早上7点更新昨日数据

请告诉我您希望推送的具体时间~

After the user selects a time, call the calendar_create tool.：
- title：小红书热门笔记订阅：{关键词}
- description: Record the current search parameters (keywords and date range).
- start_time: Set it according to the user's selected time.
- remind_type: Configure it as a recurring reminder.
- Use the current query parameters for all other settings.
Confirmation after subscribing：

✅ 订阅创建成功！

📌 订阅信息：

- 关键词：{关键词}
- 时间范围：{当前时间范围}
- 推送时间：{用户选择的时间}
- 数据更新：每日早上7点更新昨日数据

到达设定时间后，将自动为您推送最新的小红书热门笔记。

⚠️ Mandatory rules：

✅ Ask immediately after displaying the results.
✅ When the user chooses to subscribe, use the calendar_create tool.
✅ Use the current query parameters (keywords, date range, etc.).
❌ Do not skip this step.
❌ Do not ask before displaying results.

Recommend niches (⚠️ mandatory after showing data)

Core rule: After displaying trending data,proactively ask whether the user wants to query more specific niches.

Generate recommended niche terms：

Generate 10 related niche terms based on the current query keywords.
Generation principles：
- Use an appropriate level of specificity: Avoid terms that are too narrow to return data or too broad in scope.
- Cover different dimensions: Include 2–3 terms each for context, audience, style, and intent.
- Consider data performance: Prioritize topics with higher recent popularity.

Output format：

以上是「{当前关键词}」的热门数据。

如需深入了解某个细分方向，可以从以下赛道中选择：
{细分词1}、{细分词2}、{细分词3}、{细分词4}、{细分词5}、 {细分词6}、{细分词7}、{细分词8}、{细分词9}、{细分词10}

回复具体关键词，我将为您查询该赛道的热门笔记。

Example output：

以上是「减脂餐」的热门数据。

如需深入了解某个细分方向，可以从以下赛道中选择：
早餐减脂、减脂便当、低卡晚餐、减脂零食、学生党减脂、一周食谱、减脂沙拉、减脂主食、控糖减脂、懒人减脂餐

回复具体关键词，我将为您查询该赛道的热门笔记。

Special cases：

If the query is already very specific, such as “a one-week weight-loss meal plan,” further narrowing may be unnecessary. Do not ask again to expand terms that have already been expanded.

If the query has an empty keyword (sitewide trending),use the trending-topic list directlyfor recommendations, in this format:

以上是全站热门数据。

如需深入了解某个细分赛道，可以从以下热门赛道中选择：
穿搭、美食、彩妆、影视、职场、萌宠、家居、旅行、交通、兴趣、科技、互联网、医疗保健、星座情感、婚庆婚礼、拍摄、教育、亲子育儿、个人护理、潮流鞋包、生活、科学探索、新闻资讯、运动

回复具体关键词，我将为您查询该赛道的热门笔记。

Pre-output self-check [mandatory]

Before outputting, check every field in the output format for completeness:

[] Intent recognition: Was the user's search intent correctly identified?
[] Post data: Are all filtered results returned by the script shown faithfully (10 records when max_items=10), each including a bold sequence number, post title, post link, author name, author link, publication time, interactions, and a bold recommendation reason?
[] HTML card: Is an HTML card included?
[] Recommended niche terms: Are 10 niche terms listed?

If any field is missing or incomplete, fill it in before outputting.

Common command reference

Scenario	Command
Keyword search (past 7 days by default)	`python scripts/fetch_xhs_hot_articles.py --keyword "关键词" --start-date <日期>`
Sitewide trending	`python scripts/fetch_xhs_hot_articles.py --keyword "" --start-date <日期>`
Multiple-keyword search	`python scripts/fetch_xhs_hot_articles.py --keyword "词1,词2,词3" --start-date <日期>`
Paginated query	`python scripts/fetch_xhs_hot_articles.py --keyword "关键词" --start-date <日期> --page-num 1 --page-size 50`
Generate an HTML report	The script automatically generates `{keyword}_热门数据.html`
Create a subscription	After the user replies 1️⃣, use the `calendar_create` tool to create a scheduled task.

5. Use Cases

Scenario 1: A Xiaohongshu creator seeking topic inspiration

Role: Xiaohongshu content creator.
Need: Discover recent popular “weight-loss meal” posts to choose a topic for the next post.
Usage: Enter the specific keyword “weight-loss meals.” The system returns highly ranked trending posts, then automatically recommends 10 niches, such as “weight-loss breakfasts” and “weight-loss lunchboxes.”
Expected benefit: Quickly identify valuable topics through three-dimensional ranking and find niche inspiration to improve content quality.

Scenario 2: Brand market research

Role: Brand marketing staff.
Need: Understand overall trends in “outfits” without knowing which niche to choose.
Usage: Enter the broad term “outfits.” The system recommends 10 niches, such as “outfits for petite people,” “French-style outfits,” and “work outfits.” Reply “拓展” (expand) to search all directions in a batch.
Expected benefit: Cover multiple niches at once, understand the distribution of topic popularity, and support the brand's content-marketing strategy with data.

Scenario 3: Content planning for an MCN agency's creators

Role: MCN agency content director.
Need: View popular posts across several topics at once (weight-loss meals, fitness, work outfits) to schedule content across the agency's creators.
Usage: Use multiple comma-separated keywords --keyword "减脂餐,健身,职场穿搭" to query across topics.
Expected benefit: Cover several categories with one command, efficiently identify trends across fields, and improve content-planning efficiency.

Scenario 4: Scheduled daily trending-post delivery

Role: Xiaohongshu operations staff.
Need: Automatically receive popular posts in a specific niche each morning to track trends continuously.
Usage: After searching a keyword, select “1️⃣ 订阅” (subscribe), choose a delivery time, and create a scheduled calendar task.
Expected benefit: Receive the latest trending posts at a fixed daily time without repeated manual searches, track niche developments, and stay alert to creative trends.

6. Project Architecture

Directory structure

xiaohongshu-search/
├── SKILL.md                               # Skill 定义与使用文档（本文件）
├── scripts/
│   └── fetch_xhs_hot_articles.py          # 核心搜索脚本，调用红狐 API 获取小红书热门笔记
└── references/
    └── xhs_hot_article_format.md          # 数据字段格式参考文档

Technology stack

Component	Technology	Description
Runtime	Python 3.x	Standard library; no third-party dependencies.
Data API	Redfox API	Authenticated using REDFOX_API_KEY.
Output format	JSON (stdout) + HTML (file)	JSON is written to stdout for the AI to parse; HTML is a visual report file.
Presentation format	Markdown tables	The AI Agent renders JSON as tables.
Execution restriction	Main Agent only	Do not execute in a sub-Agent.

Core modules

Module	Path	Function
Search script	`scripts/fetch_xhs_hot_articles.py`	Calls the Redfox API to retrieve trending Xiaohongshu posts. Supports --keyword, --start-date, --page-num, and --page-size, and automatically generates an HTML report.
Data format reference	`references/xhs_hot_article_format.md`	Detailed descriptions of API response fields and output requirements.
SKILL definition	`SKILL.md`	Defines Skill metadata, intent-understanding rules, broad-term expansion, presentation strategies, subscription logic, and the self-check list.

Resource index

Core script：scripts/fetch_xhs_hot_articles.py — Calls the Redfox API to retrieve trending Xiaohongshu posts and generate an HTML report.
Reference documentation：references/xhs_hot_article_format.md — Read when details of data fields and output requirements are needed.

7. Frequently Asked Questions

Installation

Q: What dependencies must be installed? A: This tool uses the Python 3 standard library and needs no additional third-party dependencies. Make sure Python 3.x is installed.

Q: How do I obtain an API Key? A: Visit Redfox Hub to register and obtain an API Key, then configure the environment variable as described under “One-click Installation.”

Usage

Q: What is the difference between broad and niche terms? A: Broad terms are abstract, wide-ranging categories without contextual or attribute modifiers, such as “beauty” or “outfits.” Niche terms contain specific modifiers, such as “foundation” or “outfits for petite people.” Broad terms trigger expansion; niche terms are searched directly.

Q: What date range is available? A: The database contains data only from yesterday back to 30 days ago. The default is the past 7 days. If data is insufficient, the range expands automatically (1 day → 3 days → 7 days → 30 days). Yesterday's data is updated daily at 7 a.m.

Q: Why do some tables have scoring fields while others do not? A: Keyword searches are ranked by totalScore (relevance + popularity + recency) and show scores. Sitewide trending results are ranked by interactions and have no scoring fields. Recommended trending-post tables also omit scores.

Q: How do I view all results when there are more than 10? A: When more than 10 results are available, the system asks “View all?” Reply “是” (yes), “查看全部” (view all), or “全部” (all) to display the full data.

Q: Why are niches recommended after every query? A: This is an intended feature that helps users discover more precise directions from the current results and explore the data further. For an already specific query, the system decides whether to skip recommendations.

Troubleshooting

Q: What if the search returns nothing? A: Trending posts must have 1,000+ interactions. An obscure keyword or short date range may return no data. The system expands the date range automatically; you can also select a suggestion from relatedSearches and try again.

Q: Why does it say “Today's data has not been updated yet”? A: Yesterday's data is updated daily at 7 a.m.; today's data has not yet been ingested. The system automatically shows the most recent available data.

Q: What if the script reports an error? A: Common causes are (1) REDFOX_API_KEY is missing or expired; (2) the Python version is below 3.x; or (3) a network connection problem. Check each in turn.

Security and licensing

Q: How should I store the API Key securely? A: Option 1 is recommended: configure it in the env field of openclaw.json to avoid terminal-history exposure. Do not hard-code the key in a script or upload it to a public repository.

Q: What are the data source and copyright terms? A: The data consists of public Xiaohongshu posts collected by the Redfox API. Copyright belongs to the original authors. This tool is only for learning and content-creation reference.

Q: Why must it run in the main Agent? A: The Xiaohongshu search tool involves a complete process of intent understanding, broad-term expansion, and data presentation. The workflow is complex and depends on global context, so running it only in the main Agent ensures consistent behavior and complete adherence to its rules.

About this skill

1. Overview

2. Features

3. One-click Installation

Authentication

Get an API Key

Configure the API Key

Dependency installation

Environment variable configuration

4. Usage Guide

⚠️ Core execution rules (mandatory)

Basic usage (query in 3 steps)

Advanced usage

Understand user intent (before querying)

Broad-keyword expansion strategy

Date ranges and data queries

Preliminary notes (must be shown before the data)

Data presentation strategy (core)

A. articles count ≥ 10

B. 0 < articles count < 10

C. articles count = 0

Presentation logic (pagination)

Ask about subscriptions (mandatory)

Recommend niches (⚠️ mandatory after showing data)

Pre-output self-check [mandatory]

Common command reference

5. Use Cases

Scenario 1: A Xiaohongshu creator seeking topic inspiration

Scenario 2: Brand market research

Scenario 3: Content planning for an MCN agency's creators

Scenario 4: Scheduled daily trending-post delivery

6. Project Architecture

Directory structure

Technology stack

Core modules

Resource index

7. Frequently Asked Questions

Installation

Usage

Troubleshooting

Security and licensing