node -v
npm -v
git --version
where.exe bash

# 如果知道 Git Bash 安装位置，也可以直接检查
Test-Path "D:\Program Files\Git\bin\bash.exe"

git clone https://github.com/openclaw/openclaw.git
cd openclaw

# 先确认你的 bash.exe 路径
Test-Path "D:\Program Files\Git\bin\bash.exe"

# 配置 npm 的 script-shell
npm config set script-shell "D:\Program Files\Git\bin\bash.exe"

npm config set script-shell "C:\Program Files\Git\bin\bash.exe"

npm install
npm run build
npm link

openclaw --version

openclaw onboard --install-daemon

listening on ws://127.0.0.1:18789

C:\Users\<用户名>\.openclaw\openclaw.json

{
  "gateway": {
    "mode": "local",
    "port": 18789,
    "auth": {
      "token": "${OPENCLAW_GATEWAY_TOKEN}"
    }
  },
  "models": {
    "mode": "merge",
    "providers": {
      "dashscope": {
        "baseUrl": "https://dashscope.aliyuncs.com/compatible-mode/v1",
        "apiKey": "${OPENAI_API_KEY}",
        "api": "openai-completions",
        "models": [
          { "id": "qwen-turbo", "name": "Qwen Turbo" },
          { "id": "qwen-turbo-latest", "name": "Qwen Turbo Latest" },
          { "id": "qwen-max-latest", "name": "Qwen Max Latest" }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "dashscope/qwen-turbo"
      },
      "workspace": "~/.openclaw/workspace"
    },
    "list": [
      {
        "id": "quant-analyst",
        "model": "dashscope/qwen-turbo-latest",
        "workspace": "~/.openclaw/workspace",
        "skills": ["akshare-kline", "talib", "backtrader"]
      }
    ]
  }
}

$env:OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1"
$env:OPENCLAW_PROVIDER="openai"
$env:OPENAI_API_KEY="sk-your_api_key"

C:\Users\<用户名>\.openclaw\.env

npm install -g @openclaw/feishu
openclaw configure

cd ui
npm install
npm run build
cd ..

dist/control-ui/

openclaw gateway --port 18789

openclaw dashboard

http://127.0.0.1:18789/#token=...

my-skill/
├── SKILL.md
├── scripts/
│   └── run.py
├── templates/
└── resources/

---
name: "skill-name"
description: "这个技能能做什么，在什么情况下触发，应该如何调用。"
version: "1.0.0"
---

# 技能指令

这里写执行流程、参数说明、禁止事项和输出格式。

npm install -g clawhub

clawhub search "PPT"
clawhub search "股票"

clawhub install ppt-generator

C:\Users\<用户名>\.openclaw\skills\ppt-generator

mkdir C:\Users\$env:USERNAME\.openclaw\workspace\skills

Copy-Item -Path "C:\Users\$env:USERNAME\.openclaw\skills\ppt-generator" `
  -Destination "C:\Users\$env:USERNAME\.openclaw\workspace\skills\" -Recurse

openclaw skills list

openclaw gateway --port 18789

python skills/akshare-kline/scripts/get_kline.py 600519.SH 1d 3

skills/
├── akshare-kline/
│   ├── SKILL.md
│   └── scripts/
│       └── get_kline.py
├── talib/
│   ├── SKILL.md
│   └── scripts/
│       └── calc_indicators.py
└── backtrader/
    ├── SKILL.md
    └── scripts/
        └── run_backtest.py

---
name: akshare-kline
description: "获取 A 股 K 线。禁止 pip install、禁止检查环境。唯一正确做法：直接执行 exec，命令 python skills/akshare-kline/scripts/get_kline.py ..."
metadata:
  openclaw:
    os: ["win32"]
    requires:
      bins: ["python"]
      packages: ["akshare", "pandas"]
    permissions: ["exec"]
    timeout: 60
---

## 执行流程（必须遵守）

1. 收到 K 线请求后，直接执行 exec。
2. 命令：python skills/akshare-kline/scripts/get_kline.py <stock_code> [period] [count]
3. workdir 为工作区根目录。
4. 将脚本输出的 JSON 直接作为结果返回给用户。

禁止：安装依赖、检查环境、验证包、直接调用 akshare API。

def get_kline_data(stock_code, period="1d", count=100, adjust=""):
    try:
        symbol = parse_stock_code(stock_code)
        ak_period = period_to_akshare(period)
        start_date, end_date = calc_date_range(period, count)

        df = ak.stock_zh_a_hist(
            symbol=symbol,
            period=ak_period,
            start_date=start_date,
            end_date=end_date,
            adjust=adjust,
        )

        records = format_records(df.tail(count))
        return {
            "stock_code": stock_code,
            "period": period,
            "data_count": len(records),
            "data": records,
        }
    except Exception as error:
        return {"error": f"获取失败: {str(error)}"}

python skills/talib/scripts/calc_indicators.py macd --stock 600519.SH
python skills/talib/scripts/calc_indicators.py rsi data.csv --period 14

INDICATORS = {
    "sma": (calc_sma, ["close"], {"period": 10}),
    "ema": (calc_ema, ["close"], {"period": 12}),
    "macd": (calc_macd, ["close"], {"fast": 12, "slow": 26, "signal": 9}),
    "rsi": (calc_rsi, ["close"], {"period": 14}),
    "atr": (calc_atr, ["high", "low", "close"], {"period": 14}),
    "bbands": (calc_bbands, ["close"], {"period": 20}),
}


def calc_macd(close, fast=12, slow=26, signal=9):
    macd, sig, hist = talib.MACD(
        close,
        fastperiod=fast,
        slowperiod=slow,
        signalperiod=signal,
    )
    return {
        "macd": macd.tolist(),
        "signal": sig.tolist(),
        "histogram": hist.tolist(),
    }

python skills/backtrader/scripts/run_backtest.py 600519.SH 2025-01-01 2025-12-31 double_ma

class DoubleMAStrategy(bt.Strategy):
    params = (("fast", 10), ("slow", 30))

    def __init__(self):
        self.ma_fast = bt.indicators.SMA(self.data.close, period=self.p.fast)
        self.ma_slow = bt.indicators.SMA(self.data.close, period=self.p.slow)
        self.crossover = bt.indicators.CrossOver(self.ma_fast, self.ma_slow)

    def next(self):
        if not self.position and self.crossover > 0:
            self.buy()
        elif self.position and self.crossover < 0:
            self.close()

def run_backtest(stock_code, start_date, end_date, strategy_name):
    df = load_stock_data(stock_code, start_date, end_date)
    strategy = STRATEGIES[strategy_name]

    cerebro = bt.Cerebro()
    cerebro.addstrategy(strategy)
    cerebro.adddata(bt.feeds.PandasData(dataname=df))
    cerebro.broker.setcash(1_000_000)
    cerebro.broker.setcommission(commission=0.0002)
    cerebro.addsizer(bt.sizers.PercentSizer, percents=95)

    cerebro.addanalyzer(bt.analyzers.SharpeRatio, _name="sharpe")
    cerebro.addanalyzer(bt.analyzers.DrawDown, _name="drawdown")
    cerebro.addanalyzer(bt.analyzers.TradeAnalyzer, _name="trades")

    results = cerebro.run()
    return extract_metrics(results, cerebro)

xcopy /E /Y "skills\akshare-kline\*" "%USERPROFILE%\.openclaw\workspace\skills\akshare-kline\"
xcopy /E /Y "skills\talib\*" "%USERPROFILE%\.openclaw\workspace\skills\talib\"
xcopy /E /Y "skills\backtrader\*" "%USERPROFILE%\.openclaw\workspace\skills\backtrader\"

skills/
└── stock-screener/
    ├── SKILL.md
    └── scripts/
        └── screener.py

---
name: stock-screener
description: "A 股选股筛选。直接执行 exec，命令 python skills/stock-screener/scripts/screener.py ..."
metadata:
  openclaw:
    requires:
      bins: ["python"]
    permissions: ["exec"]
    timeout: 60
---

# 选股技能

## 执行流程（必须遵守）

1. 收到选股请求后，直接执行 exec。
2. 命令：python skills/stock-screener/scripts/screener.py <条件参数>
3. 将脚本输出的 JSON 直接返回给用户。

禁止：安装依赖、检查环境。

{
  "agents": {
    "list": [
      {
        "id": "quant-analyst",
        "skills": [
          "akshare-kline",
          "talib",
          "backtrader",
          "stock-screener"
        ]
      }
    ]
  }
}

安装 Node / Git Bash
→ 源码安装 OpenClaw
→ onboard 初始化
→ 配置模型和工作区
→ 启动 Gateway 与 Dashboard
→ 安装或编写 Skills
→ 注册到 openclaw.json
→ 同步到 workspace
→ 通过对话验证调用链路