hot_topics_selector

Financial Hot Topic Selection Tool

This Skill is used to select high-quality topics suitable for guiding investment and wealth management from financial hot topics.

Core Workflow

Phase 0: 数据抓取（Python）
   → 获取133个新闻标题
Phase 1: Agent 智能筛选（Markdown 提示词）
   → 筛选出5个选题
Phase 2: 内容抓取（Python）
   → 抓取5篇新闻详细内容
Phase 3: Agent 智能解析（Markdown 提示词）
   → 生成5个完整选题方案

Separation of Duties：

Python Script: Data scraping (Phase 0, 2)
Agent: Intelligent filtering and parsing (Phase 1, 3)

Phase 0: Data Scraping

Execution Method：

cd scripts
python3 fetch_hot_topics.py

Output：/tmp/hot_topics.json

Contains Content：

133 news headlines (5 platforms)
Weibo, Baidu, Toutiao, Douyin, Cailian Press

Data Format：

{
  "fetch_time": "2026-02-13T19:36:24+08:00",
  "total_items": 133,
  "data": {
    "weibo": { "items": [...], "count": 30 },
    "baidu": { "items": [...], "count": 30 }
  }
}

Phase 1: Agent Intelligent Filtering

Agent Execution Prompt：

You are a financial content topic selection expert. Based on the following news headlines, please select 5 topics most suitable for guiding securities account opening/fund investment.

Headline List

{133 headlines read from /tmp/hot_topics.json}

Filtering Criteria (Investment Relevance as the Core)

Highest Priority (90-100 points)

Can Be Directly Linked to Investment Products:

Money-making stories → AI funds/tech stocks/brokerage products
Investment cases → gold ETFs/asset allocation
Wealth management topics → fund dollar-cost averaging/securities account opening

Examples:

"AI earns 2 million per month" → Can recommend AI funds (95 points)
"Gold earns 1.96 million" → Can recommend gold ETFs (92 points)

High Priority (80-89 points)

Can Be Linked to Financial Planning:

Salary topics → wage income vs investment income
Savings topics → saving money vs investing
Consumption topics → financial planning

Examples:

"Year-end bonus 180 million" → Can recommend fund dollar-cost averaging (85 points)

Medium Priority (60-79 points)

Barely Linkable:

Education topics → education funds
Retirement topics → pension investment

Examples:

"Hengshui High School has changed" → Can be linked to education funds (70 points)

Not Selected (<60 points)

Cannot Be Linked to Investment:

Pure entertainment gossip
Pure social news
Topics unrelated to finance

Output Format (JSON)

{
  "selected_indices": [1, 3, 5, 7, 9],
  "reasons": {
    "1": "✅ 可直接关联：AI赚钱 → AI基金投资。投资关联度95分，大众关注度高",
    "3": "✅ 可直接关联：黄金案例 → 黄金ETF。投资关联度92分，数据冲击强"
  },
  "investment_angles": {
    "1": "可以推荐：AI主题基金、科技股票、券商AI产品",
    "3": "可以推荐：黄金ETF、券商黄金产品、资产配置服务"
  },
  "investment_relevance_score": {
    "1": 95,
    "3": 92
  }
}

Strict Requirements:

Fixed selection of 5 headlines(no more, no less)
Investment relevance of each headline ≥80 points
Explain the specific recommended direction for investment products
Prioritize topics that can be directly linked to products (≥90 points)

Phase 2: Content Scraping

Execution Method：

cd scripts
python3 fetch_news_content.py \
  --input /tmp/hot_topics.json \
  --indices 1,3,5,7,9 \
  --output /tmp/news_content.json

Input: The 5 topic indexes output in Phase 1

Output：/tmp/news_content.json

Contains Content：

Complete content of 5 news articles
Title, URL, platform, body text, keywords

Data Format：

{
  "fetch_time": "2026-02-13T21:00:00+08:00",
  "total_articles": 5,
  "articles": [
    {
      "index": 1,
      "title": "杭州大哥开1人公司靠AI月入200万",
      "url": "https://...",
      "platform": "微博",
      "content": "完整新闻内容...",
      "keywords": ["AI", "创业", "月入200万"]
    }
  ]
}

Phase 3: Agent Intelligent Parsing

Agent Execution Prompt：

You are a financial content planning expert. Based on the following 5 news articles, generate a complete topic plan for each news article.

News List

{5 news articles read from /tmp/news_content.json}

Parsing requirements

1. News content summary

Extract core information (100-150 words)
Retain key data and cases
Highlight investment-related information

2. Topic selection rationale

Explain investment relevance (80-100 points)
Analyze public interest
Predict dissemination potential

3. Guidance strategy

Entry point: How to enter from the news topic
Turning point: How to naturally shift to investment
Product recommendation: Which investment products to specifically recommend
Script example: 2-3 sentences of guidance script

Example guidance strategy:

切入点：AI创业月入200万，引发"我也能吗"的思考
转折点：普通人不会技术，但可以通过投资分享红利
产品推荐：AI主题基金、科技ETF、券商AI产品
话术示例：
  "想参与AI红利？不需要会技术，投资AI基金就能分一杯羹"
  "开户买AI基金，普通人也能享受AI发展的红利"

4. Original link

Retain the complete news URL
Convenient for later reference and citation

Output format (JSON)

{
  "topic_plans": [
    {
      "index": 1,
      "title": "杭州大哥开1人公司靠AI月入200万",
      "summary": "杭州一创业者通过AI工具运营1人公司，月收入达200万。主要业务是...",
      "selection_reason": {
        "investment_relevance": 95,
        "mass_attention": "高（AI+赚钱话题）",
        "viral_potential": "极强（月入200万数据冲击）"
      },
      "guidance_strategy": {
        "entry_point": "AI创业月入200万，引发'我也能吗'的思考",
        "turning_point": "普通人不会技术，但可以通过投资分享AI红利",
        "product_recommendation": "AI主题基金、科技ETF、券商AI产品",
        "script_examples": [
          "想参与AI红利？投资AI基金就能分一杯羹",
          "开户买AI基金，普通人也能享受AI发展红利"
        ]
      },
      "source_url": "https://..."
    }
  ]
}

Strict requirements:

Each topic selection must have all 4 complete parts
The guidance strategy must be specific and actionable
Product recommendations must be related to brokerages/funds
The script must be natural, not stiff

Usage example

Complete process

# Step 1: 数据抓取
python3 scripts/fetch_hot_topics.py

# Step 2: Agent 智能筛选
# Agent 读取本文件 Phase 1 提示词并执行

# Step 3: 内容抓取
python3 scripts/fetch_news_content.py \
  --input /tmp/hot_topics.json \
  --indices 1,3,5,7,9

# Step 4: Agent 智能解析
# Agent 读取本文件 Phase 3 提示词并执行

Total time: About 15 minutes

Script description

scripts/fetch_hot_topics.py

Function: Call the TrendRadar API to scrape hot news headlines from 5 platforms

Output: 133 news headlines (JSON)

scripts/fetch_news_content.py

Function: Based on the topic selection index, scrape the detailed news content

Input: Topic selection index (e.g., 1,3,5,7,9)

Output: Complete content of 5 news articles (JSON)

Reference materials

references/选题方法论.md

Contains the topic selection methodology of financial new media influencers, including:

5 core methodologies (pain point driven, emotional resonance, data impact, storytelling, practical orientation)
Topic selection formulas (numbers + conflict + result, pain point + solution, etc.)
Evaluation criteria (dissemination 40%, relevance 30%, practicality 20%, compliance 10%)

The Agent can refer to this file as needed to optimize the topic selection strategy.

Notes

Investment relevance is the core: Each topic selection must be able to naturally shift to securities account opening or fund investment
Fixed output of 5 topic selections: No more, no less, to ensure quality
Investment relevance ≥80 points: Topic selections below 80 points will be excluded
Python handles data: The Agent does not need to perform data scraping
Agent handles intelligence: Screening and parsing are completed by the Agent

About this skill

Core Workflow

Phase 0: Data Scraping

Phase 1: Agent Intelligent Filtering

Headline List

Filtering Criteria (Investment Relevance as the Core)

Highest Priority (90-100 points)

High Priority (80-89 points)

Medium Priority (60-79 points)

Not Selected (<60 points)

Output Format (JSON)

Phase 2: Content Scraping

Phase 3: Agent Intelligent Parsing

News List

Parsing requirements

1. News content summary

2. Topic selection rationale

3. Guidance strategy

4. Original link

Output format (JSON)

Usage example

Complete process

Script description

scripts/fetch_hot_topics.py

scripts/fetch_news_content.py

Reference materials

references/选题方法论.md

Notes