Quant Stock Picker Pro

AI-powered stock screening tool for Chinese A-shares. Daily picks using multi-factor analysis (fundamentals + technical + sentiment). Use when user asks abou...

Security Scan

The skill's code and instructions mostly match its stated stock‑picking purpose, but several surprising implementation details (hardcoded local sys.path insertion, environment proxy overrides, use of anti‑scrape tooling, and missing/mentioned config files) are disproportionate or risky and warrant caution.

Purpose & Capability

Name/description (A‑share multi‑factor screener) align with the code: scripts fetch market data, compute factors, and produce recommendations. However SKILL.md references config/scripts (e.g., scripts/config.py, risk_backtest.py, market_executor.py) that are not present in the manifest — this mismatch suggests incomplete packaging or sloppy docs. Overall capability is reasonable for the stated purpose, but missing files and unrealistic model metrics (F1 0.54%) are odd.

Instruction Scope

Runtime instructions and code perform wide network scraping (Sina, EastMoney, AkShare, web scraping of '股吧') and call external APIs. The SKILL.md instructs adding a cron job and pip installing dependencies — normal — but the actual scripts manipulate environment proxies (clearing HTTP_PROXY/HTTPS_PROXY and setting NO_PROXY='*') and the main script inserts a hardcoded absolute path (/Users/liangjiahao/.openclaw/workspace/scripts) into sys.path before importing modules. Those actions go beyond straightforward data collection and could be used to bypass network controls or import local/hidden code.

Install Mechanism

No formal install spec in registry (instruction‑only); SKILL.md recommends pip installing common data science and data packages (pandas, xgboost, akshare, efinance). This is expected for a Python quant tool. Because installation is manual (pip), risk is moderate — user must vet packages before pip install. No downloads from unknown URLs were observed.

Credentials

The skill declares no required environment variables or credentials, which is reasonable. But the code forcibly clears proxy environment variables (HTTP_PROXY, HTTPS_PROXY) and sets NO_PROXY='*' at runtime — this is disproportionate and suspicious because it alters agent/network behavior without a clear, legitimate reason. The script also inserts a hardcoded user path into sys.path which could cause imports to load arbitrary modules from the user's filesystem.

Persistence & Privilege

The skill does not request always:true and does not declare modifications to other skills or system configs. It asks users to add a cron job via an openclaw command in SKILL.md, which is a normal scheduling step but requires user action. Autonomous invocation is allowed (platform default) but not by itself a concern here.

What to consider before installing

This skill appears to implement the described stock‑screening features, but there are several red flags you should consider before installing or running it: 1) The main script injects a hardcoded user path (/Users/…/workspace/scripts) into Python's import path — this can cause the skill to import arbitrary local code on your machine; 2) It clears HTTP_PROXY/HTTPS_PROXY and sets NO_PROXY='*', which can bypass proxy/monitoring setups and is unnecessary for normal operation; 3) It imports or references an anti‑scraping library (Scrapling / StealthyFetcher), indicating aggressive scraping behavior; 4) SKILL.md mentions config files and other scripts that are not present in the package (e.g., scripts/config.py, market_executor.py), suggesting the package may be incomplete or inconsistent. Recommendations: only run this in an isolated environment (sandbox / VM), inspect and remove the sys.path insertion and proxy overrides before use, avoid running pip installs from unknown sources without reviewing packages, and confirm the skill's author/origin. If you plan to use it with real accounts or sensitive network environments, do not install until these issues are resolved.

AI增强的A股量化选股工具，每日自动筛选优质股票。

功能

• 多因子分析：基本面（60%）+ 技术面（40%）
• AI预测：XGBoost、LightGBM、随机森林集成
• 另类数据：新闻情感、股吧情绪、搜索热度
• 风险控制：动态止损、波动率目标、行业中性化
• 自动推送：每日9:35 AM自动运行（工作日）

使用场景

用户询问以下问题时自动触发：

• "推荐股票"
• "今天买什么"
• "量化选股"
• "股票筛选"
• "投资机会"

工作流程

1. 获取数据（9:35 AM）

   • 全市场A股实时行情（新浪API）
   • 新闻情感数据（AkShare）
   • 股吧情绪数据（东方财富）

2. 多因子打分

   • 成长股因子（营收增长、利润增长、ROE、市值、PE）
   • 技术面因子（涨幅、量比、换手率、连续上涨）
   • 另类数据因子（新闻情感、社交媒体、搜索热度）

3. AI预测

   • 集成学习模型（XGBoost + LightGBM + 随机森林）
   • 交叉验证准确率：F1 0.54%
   • 置信度分级（高/中/低）

4. 风险控制

   • 排除涨停板附近（涨幅>9.5%）
   • 排除亏损企业（PE<0）
   • 排除超高估值（PE>100）
   • 确保流动性（成交额>1000万）

5. 筛选输出

   • TOP 10 推荐股票
   • 包含：代码、名称、得分、关键指标、买入理由、风险提示

输出格式

# 量化选股报告 - YYYY-MM-DD

## TOP 10 推荐

| 排名 | 代码 | 名称 | 得分 | 涨幅 | PE | 换手率 | 买入理由 |
|------|------|------|------|------|----|----|----------|
| 1 | 600989 | 宝丰能源 | 45 | +8.32% | 17.5 | 5.2% | 低估值+温和上涨+成交活跃 |

## 风险提示

⚠️ **重要声明**：
- 本工具仅供学习参考，不构成投资建议
- 股市有风险，投资需谨慎
- 历史表现不代表未来收益
- 请根据自身风险承受能力做决策

## 系统信息

- 数据源：新浪财经（实时）
- AI模型：XGBoost + LightGBM + 随机森林
- 准确率：F1 0.54%（交叉验证）
- 运行时间：9:35 AM（工作日）

技术架构

quant-stock-picker-pro/
├── scripts/
│   ├── quant-stock-picker-ultimate-integrated.py  # 主脚本
│   ├── factor_engine.py                           # 因子工程
│   ├── data_collector.py                          # 数据采集
│   ├── risk_backtest.py                           # 风险管理
│   └── market_executor.py                         # 市场识别
├── references/
│   ├── factors.md                                 # 因子库文档
│   ├── strategies.md                              # 策略说明
│   └── data-sources.md                            # 数据源说明
└── SKILL.md                                       # 技能说明

安装

# 安装依赖
pip install pandas numpy scikit-learn xgboost lightgbm efinance akshare

# 设置定时任务
openclaw cron add --name "每日量化选股" --schedule "35 9 * * 1-5" --script scripts/quant-stock-picker-ultimate-integrated.py

配置

在 scripts/config.py 中配置：

# 筛选参数
MIN_SCORE = 25          # 最低得分
MAX_PE = 100            # 最大PE
MIN_VOLUME = 10000000   # 最小成交额（元）

# AI模型参数
USE_AI = True           # 是否使用AI增强
CONFIDENCE_LEVEL = "medium"  # 置信度阈值（high/medium/low）

# 另类数据
USE_NEWS_SENTIMENT = True     # 新闻情感
USE_SOCIAL_SENTIMENT = True   # 社交媒体
USE_SEARCH_HEAT = True        # 搜索热度

定制化

可以根据需求调整：

1. 选股策略

   • 保守型：低估值 + 回调买入
   • 进取型：成长股 + 突破买入
   • 平衡型：混合策略

2. 因子权重

   • 基本面权重（0-100%）
   • 技术面权重（0-100%）
   • 另类数据权重（0-100%）

3. 风险偏好

   • 低风险：严格筛选，10只股票
   • 中风险：适度筛选，20只股票
   • 高风险：宽松筛选，30只股票

注意事项

• ⚠️ 仅供学习参考，不构成投资建议
• ⚠️ 股市有风险，投资需谨慎
• ⚠️ 历史表现不代表未来收益
• ⚠️ 请根据自身风险承受能力做决策

更新日志

• v1.0.0 (2026-03-04): 初始版本
  • 多因子分析
  • AI预测
  • 另类数据整合
  • 风险控制

---

作者: Sugar Daddy 版本: 1.0.0 许可: MIT