趋势启动扫描器

基于历史技术指标验证，实时扫描筛选处于上升趋势初期的潜力股票，评分≥60分为重点关注标的。

Security Scan

The package appears to implement the stated trend-scanner/backtest functionality, but the SKILL.md and registry metadata omit important runtime requirements and the code contains hard-coded local paths and network calls — the pieces are internally inconsistent and need clarification before use.

Purpose & Capability

The name/description promise a 'real-time trend launch scanner', but the repository contains many backtest and utility scripts focused on historical analysis and batch backtests (multiple backtest_*.py, top5_backtest, etc.). The code relies on external market-data APIs (baostock and Tencent web API) and writes/reads data under hard-coded Windows paths (e.g., C:/Users/Administrator/.qclaw/workspace-ag01/data/trend_scan), yet the skill metadata and SKILL.md do not declare those dependencies, data sources, or required filesystem access. That mismatch (light-weight README vs. heavyweight code assumptions) is a coherence concern.

Instruction Scope

SKILL.md gives a simple runtime instruction (python trend_scanner.py) and high-level descriptions of modules, but does not document that many scripts will: (a) make network requests to external APIs (baostock, Tencent), (b) require multiple Python libraries (pandas, numpy, requests, baostock) and (c) read/write files under specific absolute paths. Several scripts call bs.login()/bs.logout(), hit web.ifzq.gtimg.cn, and save JSON to DATA_DIR. The runtime instructions are incomplete and omit file/network actions that materially affect privacy and environment.

Install Mechanism

No install spec is provided (instruction-only), so nothing is packaged/installed automatically. However, the code clearly requires third-party Python packages (pandas, numpy, baostock, requests, possibly others) which are not declared. This is not an immediate supply-chain red flag (no arbitrary download URLs or extract operations), but it is an operational omission: users need to pip-install dependencies manually or the scripts will fail.

Credentials

The skill declares no required environment variables or credentials, and the code does not appear to expect secrets. However, it does perform network requests to public market-data endpoints (baostock, Tencent) and uses hard-coded filesystem locations under C:/Users/Administrator/.qclaw; those absolute paths may unintentionally read or overwrite local datasets. The absence of any declared config/paths in metadata contrasts with the code's reliance on local directories (DATA_DIR) and specific workspace layout.

Persistence & Privilege

Flags show always:false and normal autonomous invocation allowed (default). The skill does not request permanent 'always' inclusion and does not modify other skills. The main elevated behaviors are normal: network I/O and file read/write when executed, which is expected for a backtester but should be noted.

What to consider before installing

Before installing or running this skill: 1) Expect to need Python and packages (pandas, numpy, baostock, requests, etc.) — the SKILL.md/metadata do not list them. 2) Inspect or change the hard-coded DATA_DIR and sys.path entries (they point to C:/Users/Administrator/.qclaw/…), otherwise the scripts may read/write data in those locations or fail on non-Windows systems. 3) The code makes network calls to baostock and Tencent (web.ifzq.gtimg.cn) to fetch market data — ensure you are comfortable allowing those requests. 4) Run the code first in an isolated environment (sandbox or container) so you can observe file I/O and network traffic. 5) Ask the publisher to provide a clear dependency list, configurable data directory, and a concise README describing which scripts are intended for real-time scanning vs. offline backtesting. If you need this skill to run inside a restricted environment (no network or no file writes), request a version that documents and parameterizes those behaviors.

趋势启动扫描器 (Trend Launch Scanner)

简介

基于历史数据验证的趋势跟踪选股系统。通过对比"上涨组"与"对照组"的技术指标差异，找出真正有区分度的信号，用于筛选当前处于上升趋势初期的潜力股。

核心能力： 方向2（行业分散）+ 方向3（动态权重）+ 方向4（卖出信号）+ 方向5（确认信号）+ 方向6（情绪指标）

---

功能模块

1. 核心评分系统

满分 100 分，筛选标准 >= 60 分：

信号	分值	区分度
MACD柱线为正	25分	+26%
价格>MA20	20分	+21%
RSI上升	15分	+19%
RSI>50	15分	+18%
窗口内上升	10分	+18%
均线多头	10分	+16%
MACD柱增	5分	+15%

2. 动态权重系统 (dynamic_weights.py)

根据近期验证数据自动调整信号权重：

• 持续追踪推荐股票的收益表现
• 高收益信号权重上调
• 低收益信号权重下调

3. 确认信号系统 (confirmation.py)

增加二次确认，避免假信号：

检查项	条件	说明
不过热	5日涨幅<15%	避免追高
布林位置	<1.0	不在上轨
量比	0.5-2.0	温和放量
RSI	<80	避免超买
MACD柱线	<5%	避免过度强势
MA20趋势	连续5日上升	趋势确认

4. 卖出信号系统 (tracking.py)

动态止盈止损：

评分	止盈	止损
100分	+20%	-10%
90-99分	+15%	-8%
80-89分	+12%	-7%
60-79分	+10%	-5%

5. 行业分散系统 (industry_data.py)

• 每个行业最多推荐2只
• 显示行业标签和emoji
• 保证推荐多样性

6. 情绪指标系统 (sentiment.py)

• 大盘指数涨跌
• 涨跌停数量
• 热门板块排行

---

验证结果

基于 117 个上涨样本 vs 186 个对照样本：

✅ 启动组特征：

• MACD柱线为正（最强信号）
• 价格在MA20上方（趋势向上）
• RSI偏强且在上升（动能增强）

❌ 常见误区：

• 不是低位蓄势
• 不是缩量
• 不是大幅回调

核心结论：真正的启动点是"强势延续"，而不是"低位反弹"。

---

脚本说明

脚本	用途
trend_scanner.py	主扫描脚本
dynamic_weights.py	动态权重系统
confirmation.py	确认信号系统
tracking.py	卖出追踪系统
industry_data.py	行业分类数据
sentiment.py	情绪指标系统
verification.py	验证追踪系统
stock_pool.py	股票池（306只）

---

使用方法

python trend_scanner.py

---

免责声明

本系统仅供学习研究使用，不构成任何投资建议。 股票投资有风险，入市需谨慎。