SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsxhs-title-scorer

xhs-title-scorer

Data & analytics

View repository

Updated 10-16 07:27

About this skill

Xiaohongshu Viral Title Scorer

1. Introduction

A title scoring tool based on analysis of Xiaohongshu viral-hit data -- enter the title to be scored and related topic keywords, and the system automatically queries viral-hit data in the same niche, distills actionable viral-hit patterns, and performs weighted scoring on the user's title across 6 dimensions (topic match, structural compliance, clarity of benefit, emotional arousal, perceived scarcity, compliance and safety), outputting an S/A/B/C grade scoring report and optimization suggestions.

Target users: Xiaohongshu creators, content operators, copywriters, and writers who need to verify title quality.

2. Features

Core features：

  • Multi-dimensional weighted scoring -- a 6-dimension weighted system (topic 15%, structure 20%, benefit 25%, emotion 20%, scarcity 15%, compliance 5%), with a maximum score of 10
  • Viral-hit pattern distillation -- automatically distills actionable patterns (structure / wording / benefit points) from the actual viral hits recalled
  • Grade rating -- S grade (9.0+), A grade (7.0-8.9), B grade (5.0-6.9), C grade (below 5.0)
  • Optimization suggestion generation -- outputs a core issue list, reference viral-hit titles, and suggested revised versions according to the score grade

Highlights：

  • Title overlap detection: if the user's title overlaps with a viral-hit title, directly give it an S-grade score
  • Data-driven scoring: all statistical data must be calculated based on the actual viral-hit data recalled; fabrication is strictly prohibited
  • Multiple revision style suggestions: three versions -- practical style / emotional style / curiosity-driven style
  • Anti-empty-talk rule: all conclusions must be supported by data

3. One-click installation

Authentication

Get API Key

Please go to RedFox Hub to get the API KEY

Configure 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..."

export REDFOX_API_KEY="ak_xxxx..."

Dependency installation

Requires the requests library (standard Python library); no additional installation is needed.

4. Usage Guide

Basic usage

Input format: Provide the "title to be scored" and "related topic / keywords".

Input examples：

  • "Help me check how this title 'This sunscreen is really so good' is, the keyword is sunscreen"
  • "Score 'You must try this in summer', the topic is summer outfits"
  • "How is this title 'Skincare tips beginners must read'"

Complete operation process

Step 1: Understand the user input and extract key information

Identify two core elements:

  1. The title to be scored
  2. Related topic / keywords (if the user does not provide them, extract them from the title)

Step 2: Determine the keyword type

  • Niche term / vertical niche -> query data directly
  • Specific product term -> query that product's data directly
  • Generic term / broad category -> execute the generic term expansion strategy to query

Step 3: Time range and data query

  • The database only contains data from yesterday to 30 days ago
  • The default time range is the last 30 days
  • Date calculation: first obtain the current system time, then calculate startDate based on the user's expression
python scripts/fetch_xhs_trends.py --keyword <关键词> --start-date <日期>

When data is insufficient, truthfully display the existing data and notify the user.

Step 4: Extraction of viral content patterns (executed internally only, not shown to the user)

All statistical data must be calculated based on the actually retrieved viral content data. It is strictly prohibited to fabricate any percentages, ratios, or numerical values.

The pattern dimensions correspond one-to-one with the scoring dimensions:

  1. Topic match pattern: Extract the trending keywords hit in the actual viral content and their frequency of occurrence
  2. Structural compliance pattern: Count the proportion of mainstream structures in viral titles (starting with numbers, interrogative sentences, exclamatory sentences, etc.) and the average word count
  3. Benefit clarity pattern: Count the specific types of benefits explicitly stated in viral titles
  4. Emotional arousal pattern: Count the types of emotional hooks in viral titles (curiosity, anxiety, satisfaction, surprise, etc.)
  5. Scarcity perception pattern: Analyze the differentiated expressions and exclusivity/limited-edition expressions in viral titles
  6. Compliance and safety pattern: Check whether viral titles contain violating/rate-limiting words, and count the proportion of compliant titles

Data calculation requirements：

  • All proportion data must be labeled with the denominator (e.g., "8/10")
  • Calculations must be based on the actually retrieved viral titles
  • If the retrieved data is fewer than 5 items, it is necessary to state "The data sample is relatively small, and the patterns are for reference only"

Step 5: Multi-dimensional scoring (shown to the user)

Title overlap determination (checked first before scoring)：

  • If the user's title is completely identical to a viral title, or differs only in punctuation or spaces -> directly give an S-level score(above 9.5 points)
  • Label it as "This title overlaps with a viral title" and skip the regular scoring process

Total score is 10 points, calculated by weighting the dimensions：

Scoring dimension Weight Scoring standard
Topic match 15% 10 points: hits current trending keywords in the same track; <7 points: no trending topic, outdated theme
Structural compliance 20% 10 points: conforms to the mainstream structure of viral content in the same track (has a hook + has a benefit); <5 points: plain narration with no hook
Benefit clarity 25% 10 points: explicitly states specific benefits/timeliness/cost; <5 points: users cannot see clear value
Emotional arousal 20% 10 points: has a clear emotional hook (curiosity/anxiety/satisfaction/surprise); <5 points: bland with no emotional fluctuation
Scarcity perception 15% 10 points: has differentiated/exclusive expressions; <5 points: severely homogenized
Compliance and safety 5% 10 points: no violating/rate-limiting words; 0 points: contains clearly violating content

Comprehensive score = Topic match x 0.15 + Structural compliance x 0.20 + Benefit clarity x 0.25 + Emotional arousal x 0.20 + Scarcity perception x 0.15 + Compliance and safety x 0.05

Output format：

## [用户输入的标题]-标题评分报告

**数据说明**：查询到相关小红书爆款标题[X]个，已据此总结相关爆款规律，以下是对您标题的评分。

### 综合得分
**[S/A/B/C]级**-**[X.X]分**（满分10分）

[该等级的具体说明]

### 各维度评分
| 维度 | 得分 | 权重 | 加权得分 |
|------|------|------|----------|
| 主题匹配度 | [X]分 | 15% | [X.X]分 |
| 结构合规度 | [X]分 | 20% | [X.X]分 |
| 利益清晰度 | [X]分 | 25% | [X.X]分 |
| 情绪唤醒度 | [X]分 | 20% | [X.X]分 |
| 稀缺性感知 | [X]分 | 15% | [X.X]分 |
| 合规安全性 | [X]分 | 5% | [X.X]分 |

Step 6: Generation of optimization suggestions (shown to the user)

When at S level (9.0+ points): Do not output optimization suggestions; simply praise according to viral content patterns. List 3-5 highly liked titles on the same topic and explain the similarities.

At A/B/C level：

  1. Core problem list -- the most impactful problems, each of which must correspond to specific evidence of viral patterns
  2. Reference viral titles -- list 3-5 highly liked titles on the same topic, and explain reusable points
  3. Suggested revised versions -- provide 2-3 revised versions (practical style / emotional style / curiosity style), and indicate applicable scenarios and expected effects

S-level output format：

## 爆款分析

**[若重合：该标题与爆款标题重合，已被市场验证为爆款！]**
**[若不重合：这个标题非常优秀，完美踩中了爆款规律！]**

### 爆款亮点
- **主题匹配度**：[基于实际爆款规律说明]
- [其他五个维度...]

### 参考爆款标题

!!!受小红书风控规则限制，部分作品链接可能无法正常跳转...

1. [参考标题1](链接)（作者：[作者名](作者链接)，互动数：XXX）
   - [相似之处]
...（共3-5条）

A/B/C-level output format：

## 优化建议

### 核心问题
1. [问题1]
2. [问题2]
3. [问题3]

### 参考爆款标题

!!!受小红书风控规则限制...

1. [参考标题1](链接)（作者：[作者名](作者链接)，互动数：XXX）
   - 可复用点：[说明]
...（共3-5条）

### 修改版本建议

**版本1：干货风**
- 修改后标题：[新标题]
- 适用场景：[说明]
- 预期效果：[说明]

**版本2：情绪风**
- 修改后标题：[新标题]
- 适用场景：[说明]
- 预期效果：[说明]

**版本3：猎奇风**（可选）
- 修改后标题：[新标题]
- 适用场景：[说明]
- 预期效果：[说明]

Advanced usage

Wording rules (must be followed 100%): In all output content, the words "crawl" and "scrape" are absolutely prohibited and must be replaced with "retrieve".

Common commands quick reference

Scenario Input example Expected output
Title scoring "Help me score 'This sunscreen is really amazing', keyword sunscreen" 6-dimension weighted score + grade rating + optimization suggestions
Topic scoring "How is this title 'Wearing this in summer looks super good', the topic is summer outfits" Scoring report combined with seasonal characteristics

5. Use cases

  1. Creators verify titles before publishing: Before publishing a note, submit candidate titles for scoring, and based on the score rating and optimization suggestions, choose or revise a title more likely to become viral.
  2. Content operations teams evaluate in batches: Score multiple candidate titles to quickly screen out the titles with the greatest viral potential and improve team output efficiency.
  3. Copywriting newcomers learn viral patterns: By comparing titles they wrote with reference viral titles, they can intuitively understand the structure, wording, emotional expression, and other characteristics of viral titles.
  4. Title pre-screening before A/B testing: Before investing traffic in testing, use the scoring system to pre-screen 2-3 high-scoring titles for testing, improving testing efficiency.

6. Project architecture

Directory structure

xhs-title-scorer/
├── SKILL.md                      # 本文件
└── scripts/
    └── fetch_xhs_trends.py       # 数据查询脚本，调用红狐API获取爆款标题数据

Tech stack

Technology Purpose
Python + requests HTTP requests and data retrieval
RedFox API Data source for Xiaohongshu viral titles
AI intelligent analysis Viral pattern extraction, multi-dimensional scoring, and optimization suggestion generation

Core module description

  • fetch_xhs_trends.py: Accepts keyword and time range parameters, calls the RedFox API to query popular Xiaohongshu note title data, and outputs a Markdown-format viral data file for scoring analysis.

Resource index

File Purpose
scripts/fetch_xhs_trends.py Data query script

7. FAQ

Installation

Q: What dependencies need to be installed? A: Requires requests library; if not installed, run pip install requests。

Q: How to configure the API Key? A: Please go to RedFox Hub to obtain the API KEY, and configure it through the environment variable REDFOX_API_KEY .

Usage

Q: What are the scoring dimensions? A: 6 dimensions: topic relevance (15%), structural compliance (20%), clarity of benefit (25%), emotional arousal (20%), perceived scarcity (15%), compliance and safety (5%).

Q: What happens if the title overlaps with a viral hit? A: Directly assign an S-tier score (above 9.5), mark it as "This title has been market-validated as a viral hit," and skip the regular scoring process.

Q: What is the data range? A: The database only contains data from yesterday to 30 days ago, with a default time range of the most recent 30 days.

Q: Are there optimization suggestions for S-tier scores? A: S-tier does not output optimization suggestions; instead, it switches to praising viral highlights and displaying reference titles.

Troubleshooting

Q: What should I do if there is too little recall data? A: If the recall data is fewer than 5 items, it will state "The data sample is small; patterns are for reference only." If the data for the past 30 days is still insufficient, display the existing data as is and notify the user.