Stock Theme Events

获取 A 股市场炒作题材对应的真实新闻事件。使用场景：(1) 分析近 N 日涨幅前列股票的题材分布，(2) 将相似题材聚类合并，(3) 从金十数据等新闻源获取对应题材的真实资讯事件，(4) 生成题材 - 事件关联分析报告。支持被其他 skill 调用或独立运行。

Security Scan

Skill mostly does what it says (map stock themes to news) but there are environment and invocation mismatches (calls an OpenClaw CLI via subprocess and writes to a hardcoded user path) that are not declared in the metadata and deserve caution.

Purpose & Capability

The skill's code and instructions align with the stated purpose (collect stock themes, cluster them, fetch news via akshare/browser, generate a Markdown report). Required Python packages (akshare, sentence-transformers, scikit-learn) are appropriate for the functionality. However, scripts call the OpenClaw CLI ('openclaw sessions spawn ...') to invoke the ths-stock-themes skill, but the skill metadata declares no required binaries — this is an undeclared runtime dependency and an environment mismatch.

Instruction Scope

SKILL.md instructs calling 'ths-stock-themes' (expected), using akshare and a browser fallback (expected). In practice, get_stock_themes.py spawns the OpenClaw CLI via subprocess to call that other skill and then attempts to extract JSON from CLI stdout with regex — this invocation method is not described in metadata and is fragile. search_news returns a 'browser_search' instruction object intended to be executed by an external browser tool; that is reasonable but means the agent or environment must actually execute browser actions. Several code paths read/write local cache and output paths and use hardcoded absolute paths (see run_full_analysis.py) which extend scope to filesystem writes beyond the working directory.

Install Mechanism

There is no install specification (instruction-only), which keeps risk lower. The README/SKILL.md recommend pip installing dependencies. No network download URLs or archive extracts are present in the bundle. The only notable install/runtime action is that sentence-transformers models (about 500MB) will be downloaded at first use — a large model download that the user should expect.

Credentials

The skill declares no required environment variables or credentials, which is reasonable. However, it implicitly requires: (1) an 'openclaw' CLI binary available in PATH for subprocess-based skill-to-skill calls, (2) network access for akshare and browser searches, and (3) disk write permissions to various cache/output paths. Additionally, run_full_analysis.py contains a hardcoded output_dir under '/Users/shinelp100/...' which may overwrite files in that location if present. These implicit requirements are not declared in the metadata and could surprise users.

Persistence & Privilege

The skill does not request permanent inclusion (always:false) and does not modify other skills' configurations. It will write its own cache and report files (normal for this kind of tool). It spawns other skills via the OpenClaw CLI rather than modifying their configuration.

What to consider before installing

This skill appears to implement the described functionality, but take precautions before installing/running it: - Verify runtime requirements: the code calls an 'openclaw' CLI via subprocess to invoke ths-stock-themes; ensure you have (or intentionally want) that CLI available. The skill metadata did not declare this binary dependency. - Review and adjust file paths: run_full_analysis.py uses a hardcoded author-specific path (/Users/shinelp100/...), and scripts create cache and output files — change those paths to a safe workspace to avoid unexpected overwrites. - Expect large downloads: the sentence-transformers model (~500MB) will be fetched on first run if you install that dependency. - Network and automation: the skill uses akshare (network calls) and may direct a browser automation tool to scrape Jin10/other sites; only run in an environment where outbound network and web automation are acceptable. - Trust boundary: the skill spawns another skill (ths-stock-themes). Inspect or vet that other skill before allowing subprocess-based invocations, since its output is parsed and used here. If you plan to use this skill, run it in a sandboxed environment (or inspect/modify get_stock_themes.py to replace the subprocess call with a direct API or validated interface), update output paths, and ensure required Python packages and the 'openclaw' CLI are intentionally provided.

Stock Theme Events - 股票题材与新闻事件关联分析

概述

本技能将市场炒作的题材概念与真实新闻事件对应起来，输出结构化的 Markdown 报告。核心流程：获取热门股票 → 提取题材 → 聚类合并 → 搜索新闻 → 生成报告。

核心功能

1. 获取个股题材

调用 ths-stock-themes skill 获取股票的题材/概念列表。

输入：股票代码列表 输出：{股票代码：[题材 1, 题材 2, ...]}

2. 题材聚类合并

使用语义相似度 + 同义词词典将相似题材合并。

示例：

• "人工智能" + "AI" + "大模型" → "人工智能/AI"
• "芯片" + "半导体" + "集成电路" → "半导体/芯片"

输出：{题材方向：[相关股票列表]}

3. 获取对应新闻事件

优先使用 akshare 财经新闻接口，缺失时用 browser 工具从金十数据搜索。

搜索范围：近 15 天 每题材选取：2-5 条高相关度新闻

输出：{题材方向：[新闻列表]}

4. 生成 Markdown 报告

结构化输出题材方向、涉及股票、对应新闻事件。

使用方法

被其他 skill 调用（推荐）

# 在调用方 skill 中
from stock_theme_events import analyze_theme_events

result = analyze_theme_events(
    stock_list=["600519", "000858", ...],  # 可选，默认获取近 10 日涨幅前 30
    output_path="~/Desktop/A 股每日复盘/",  # 可选
    date_range=15,  # 新闻搜索天数，默认 15
    top_themes=8    # 主流题材数量，默认 8
)

独立运行

生成题材事件分析报告
分析近 10 日涨幅前 30 股票的题材和对应新闻
获取市场炒作题材的真实事件驱动

执行流程

完整流程

1. 获取热门股票（如未提供）

   • 调用 ths-stock-themes 获取近 10 日涨幅前 30 股票
   • 排除 ST 股票

2. 获取个股题材

   • 遍历股票列表
   • 调用 ths-stock-themes 获取每只股票的题材
   • 缓存结果避免重复请求

3. 题材聚类

   • 统计题材频次
   • 使用语义相似度模型聚类
   • 应用同义词词典合并
   • 按频次排序取 TOP N

4. 搜索新闻

   • 对每个题材方向，先尝试 akshare 接口
   • 如结果不足，用 browser 工具搜索金十数据
   • 筛选近 15 天内的资讯
   • 去重并按相关性排序

5. 生成报告

   • 使用模板生成 Markdown
   • 保存到指定路径
   • 返回报告文件路径

输出格式

# A 股题材 - 事件关联分析报告

**数据日期**：2026-03-21
**统计周期**：近 10 个交易日
**新闻范围**：近 15 天
**样本股票**：涨幅前 30 只（排除 ST）

---

## 一、主流炒作题材方向

| 排名 | 题材方向 | 涉及股票数 | 代表股票 |
|------|----------|------------|----------|
| 1 | 人工智能/AI | 8 | 中科曙光、浪潮信息... |
| 2 | 半导体/芯片 | 6 | 中芯国际、北方华创... |
| ... | ... | ... | ... |

---

## 二、题材 - 事件对应分析

### 1. 人工智能/AI（8 只股票）

**涉及股票**：中科曙光、浪潮信息、科大讯飞...

**对应新闻事件**：

1. **[新闻标题 1]** (2026-03-20)
   - 来源：金十数据
   - 摘要：...
   - 链接：[查看详情](url)

2. **[新闻标题 2]** (2026-03-19)
   - 来源：财联社
   - 摘要：...
   - 链接：[查看详情](url)

---

## 三、总结

- 当前市场主要炒作方向：XXX、XXX、XXX
- 新闻事件驱动明显：XXX 政策/事件催化
- 风险提示：...

脚本说明

cluster_themes.py

题材聚类脚本，使用语义相似度 + 同义词词典。

python3 scripts/cluster_themes.py --input themes.json --output clustered.json

参数：

• --input: 输入题材 JSON 文件
• --output: 输出聚类结果
• --top: 保留 TOP N 个题材方向（默认 8）
• --similarity-threshold: 相似度阈值（默认 0.7）

search_news.py

新闻搜索脚本，akshare + browser 混合方案。

python3 scripts/search_news.py --theme "人工智能" --days 15 --limit 5

参数：

• --theme: 题材关键词
• --days: 搜索天数范围（默认 15）
• --limit: 返回新闻数量（默认 5）
• --output: 输出 JSON 文件路径

generate_report.py

报告生成脚本。

python3 scripts/generate_report.py --themes clustered.json --news news.json --output report.md

参数：

• --themes: 聚类后的题材 JSON
• --news: 新闻数据 JSON
• --output: 输出 Markdown 路径
• --template: 模板文件路径（可选）

配置说明

config/theme_synonyms.json

题材同义词配置，用于聚类合并。

{
  "人工智能/AI": ["人工智能", "AI", "大模型", "AIGC", "算力", "机器学习"],
  "半导体/芯片": ["芯片", "半导体", "集成电路", "光刻机", "存储芯片"],
  "新能源": ["光伏", "风电", "储能", "锂电池", "新能源汽车", "氢能"],
  "医药生物": ["创新药", "医疗器械", "生物医药", "CXO", "疫苗"],
  "金融": ["券商", "银行", "保险", "多元金融", "互联网金融"],
  "大消费": ["白酒", "食品饮料", "家电", "零售", "旅游"],
  "周期资源": ["钢铁", "煤炭", "有色", "化工", "石油"],
  "军工": ["军工", "航天", "船舶", "国防", "航空发动机"],
  "科技通信": ["5G", "通信", "物联网", "云计算", "数据中心"],
  "房地产": ["房地产", "物业", "园区开发", "REITs"]
}

依赖安装

pip install akshare sentence-transformers scikit-learn

注意事项

1. 数据时效性：新闻数据实时变化，建议在报告中注明数据获取时间
2. 浏览器自动化：金十数据搜索需要 browser 工具，避免高频请求
3. ST 股票：自动排除名称包含"ST"的股票
4. 语义模型：首次使用会下载 sentence-transformers 模型（约 500MB）
5. API 限制：akshare 有频率限制，建议添加请求间隔

错误处理

• 题材获取失败：重试 3 次，跳过该股票并记录警告
• 新闻搜索失败：降级到 browser 工具，仍失败则标记"暂无相关新闻"
• 聚类失败：降级到简单频次统计
• 报告生成失败：返回错误信息，提示检查模板文件

性能优化

1. 缓存机制：个股题材结果缓存到 cache/themes/ 目录
2. 并行搜索：多个题材的新闻搜索可并行执行
3. 增量更新：支持基于已有报告增量更新

示例调用

被 stock-daily-report 调用

# 在 stock-daily-report skill 中
result = analyze_theme_events(
    stock_list=top_gainers,  # 近 10 日涨幅前 30
    output_path=daily_report_dir,
    date_range=15
)
# 将 result 整合到每日复盘报告中

独立运行

用户：分析当前市场炒作的题材和对应新闻
助手：正在获取近 10 日涨幅前 30 股票...
     正在提取题材信息...
     正在聚类合并...
     正在搜索对应新闻...
     报告已生成：~/Desktop/A 股每日复盘/theme-events-2026-03-21.md