PTRADE量化策略生成器

国金 PTRADE 量化平台策略生成技能。根据用户需求， 生成可在 PTRADE 平台运行的 Python 量化交易策略代码。 触发关键词：PTRADE、ptrade、量化策略、量化交易、生成策略、写策略代码。

Security Scan

The skill's files, templates, and runtime instructions are consistent with its stated purpose of generating PTRADE Python strategy code and do not request unrelated privileges or installs.

Purpose & Capability

Name/description, README, SKILL.md, kb_ref.md and the included strategy templates (MA, dual MA, factor) all align: the skill reads PTRADE docs from the user's IMA knowledge base and produces Python strategies for the PTRADE platform. There are no extraneous binaries, env vars, or unrelated dependencies declared.

Instruction Scope

SKILL.md instructs the agent to use the IMA '国金量化' knowledge base and to generate strategy code using the PTRADE API references; the provided templates and API usage examples match this scope. The instructions do not ask the agent to read unrelated system files, secrets, or to exfiltrate data to unknown endpoints.

Install Mechanism

No install spec is present (instruction-only with template files). Nothing is downloaded or written by an installer during skill installation, which minimizes install-surface risk.

Credentials

The skill declares no required env vars or credentials. kb_ref.md mentions IMA base URL and headers (ima-openapi-clientid / ima-openapi-apikey) as part of the knowledge-base context, but the skill does not request those explicitly — this is plausible because the platform/agent typically provides access to the user's IMA documents. Users should confirm that no external secret is manually supplied to the skill beyond the platform's normal knowledge-base access.

Persistence & Privilege

always is false and the skill is user-invocable. It does not request to modify other skills or system-wide settings and has no elevated persistence requirements.

Assessment

This skill appears to do what it says: generate PTRADE-compatible Python strategies from your knowledge base and included templates. Before using any generated code in live trading, review and test it thoroughly in backtest/sandbox mode (check trade sizes, stop-loss/take-profit, and that trade.open_long/close_long calls are correct). Confirm how your platform supplies access to the IMA knowledge base (the kb_ref mentions IMA API headers) and avoid pasting or storing your PTRADE account credentials in the skill. If you want extra caution, run the generated code locally or in a PTRADE sandbox and inspect it line-by-line before enabling automatic/live execution.

PTRADE 量化策略生成 Skill

概述

本 Skill 从知识库「国金量化」中读取 PTRADE API 文档，为用户生成可在 PTRADE 平台运行的量化交易策略代码。

知识库文档结构（个人知识库 → 国金量化）：

文件	说明
PTrade新版API帮助文档.pdf	API 帮助文档
PTrade研究模块简要使用说明.pdf	研究模块使用说明
PTrade系统量化交易与回测的区别.pdf	回测 vs 实盘
PTrade新版API帮助文档.pdf	API 完整参考
PTrade API 完整详细文档.docx	函数与变量大全
PTrade API 函数与变量大全.docx	API 速查
demo-单因子策略	单因子示例
demo-双均线策略	双均线示例
demo-小市值日线交易策略	小市值示例
demo-三因子策略	三因子示例
demo-日内交易策略	日内示例
demo-协整配对策略	配对交易示例
demo-二八轮动策略	轮动策略示例
demo-阿隆指标	技术指标示例
demo-猛犸策略	高级策略示例
Ptrade 量化交易 API接口文档	在线文档（web URL）

---

核心 API 参考

常用对象

# 获取上下文
context = get_json('context')
g = context.get('g', {})         # 全局对象
security = g.get('security')    # 证券行情
trade   = g.get('trade')        # 交易模块
log     = g.get('log')          # 日志模块

# 常用函数
log.info("消息")                # 打印日志
log.error("错误消息")
get_json('context')             # 获取上下文

行情数据

# 获取历史 K 线
security.get_bars(n: int, end_date: str, fields: list, count: int, period: str)
# period: "1d"(日线) "1m"(1分钟) "5m"/"15m"/"30m"/"60m"/"1h"/"4h"
# fields: ["open", "high", "low", "close", "volume"]

# 获取财务数据
security.get_financial_data(start_date, end_date, count, fields)
# fields: "pubDate","totalShare","flowShare","closePrice","PE","PB","PCF","PS"
# fields: "netProfit","grossProfit","operatingProfit","totalRevenue","debtAssetRatio"

# 获取成分股
security.get_index_stocks(index_code: str)
# index_code: "000300.XSHG"(沪深300) "000016.XSHG"(上证50) "000852.XSHG"(中证1000)

# 获取当前持仓
security.get_positions()

# 获取当前账户信息
security.get_account_info()

# 获取主力资金流向
security.get_money_flow(start_date, end_date, count, fields)

交易下单

# 买入（回测模式）
trade.open_long(security: str, volume: int, price: float = 0, price_type: str = "")

# 卖出（回测模式）
trade.close_long(security: str, volume: int, price: float = 0, price_type: str = "")

# 撮合价格类型
# market: 以开盘价成交
# latest: 以最新价成交（需订阅实时行情）
# 指定价格: 以指定价格成交（需订阅Level2）

财务因子

# 获取财务因子
security.get_误符串(n: int, fields: list, count: int, end_date: str)
# 示例: security.get_financial_data(count=10, end_date="", fields=["PE","PB","PCF"])

支持的库

• 数据分析：numpy, pandas, scipy
• 机器学习：sklearn（不支持深度学习框架）
• 时间序列：statsmodels
• 因子研究：alphalens, ta-lib
• 数据库：pymongo, redis

---

策略模板

模板 1：日线双均线策略

def initialize(context):
    """初始化策略参数"""
    context.params = {
        "security": "000001.XSHG"         # 交易标的
        "short_window": 5                  # 短期均线
        "long_window": 20                 # 长期均线
    }
    context.g.security = get_json("context").get("g", {}).get("security")
    context.g.trade    = get_json("context").get("g", {}).get("trade")

def handle_data(context):
    security = context.g.security
    trade    = context.g.trade

    # 获取近 N 根 K 线
    bars = security.get_bars(
        n=context.params["long_window"],
        end_date="",
        fields=["open", "high", "low", "close", "volume"],
        count=context.params["long_window"],
        period="1d"
    )

    close = bars["close"]
    short_ma = close[-5:].mean()  # 5 日均线
    long_ma  = close[-20:].mean() # 20 日均线

    # 金叉买入
    if short_ma > long_ma and close[-2] <= close[-3] * 0.995:
        trade.open_long(context.params["security"], 100, price=0, price_type="market")

    # 死叉卖出
    if short_ma < long_ma:
        trade.close_long(context.params["security"], 100, price=0, price_type="market")

模板 2：财务因子选股策略

def initialize(context):
    context.params = {
        "index": "000300.XSHG",   # 选股范围：沪深300
        "n": 10,                  # 持仓数量
        "factor": "PE",           # 估值因子：PE/PB/PCF/PS
        "ascending": True,        # True=低估值优先
    }
    context.g.security = get_json("context").get("g", {}).get("security")
    context.g.trade    = get_json("context").get("g", {}).get("trade")

def handle_data(context):
    security = context.g.security
    trade    = context.g.trade

    # 获取指数成分股
    stocks = security.get_index_stocks(context.params["index"])

    # 批量获取财务数据
    financial_data = security.get_financial_data(
        start_date="", end_date="",
        count=1,
        fields=[context.params["factor"]]
    )

    # 按因子排序
    sorted_stocks = sorted(stocks, key=lambda s: financial_data.get(s, {}).get(context.params["factor"], float("inf")))
    if context.params["ascending"]:
        sorted_stocks = sorted_stocks[::-1]

    # 取前N只，等权配置
    target_stocks = sorted_stocks[:context.params["n"]]
    for stock in target_stocks:
        trade.open_long(stock, 100, price=0, price_type="market")

模板 3：日内交易策略

def initialize(context):
    context.params = {
        "security": "000001.XSHG",
        "upper_limit": 0.05,   # 涨幅上限
        "lower_limit": -0.03,   # 跌幅下限
        "volume_threshold": 500000  # 成交量阈值
    }
    context.g.security = get_json("context").get("g", {}).get("security")
    context.g.trade    = get_json("context").get("g", {}).get("trade")

def handle_data(context):
    security = context.g.security
    trade    = context.g.trade

    # 获取5分钟K线
    bars = security.get_bars(n=20, end_date="", fields=["close", "volume"], count=20, period="5m")
    close  = bars["close"]
    volume = bars["volume"]

    last_close = close[-1]
    prev_close = close[-2]
    change = (last_close - prev_close) / prev_close if prev_close > 0 else 0

    # 放量突破
    avg_vol = volume[:-1].mean()
    if volume[-1] > avg_vol * 1.5 and change > context.params["upper_limit"]:
        trade.open_long(context.params["security"], 100, price=0, price_type="market")

    # 止损
    if change < context.params["lower_limit"]:
        trade.close_long(context.params["security"], 100, price=0, price_type="market")

---

生成策略流程

1. 理解需求：确认策略类型（选股/择时/套利/事件驱动）、标的、周期、信号
2. 确认 API：参考上方 API 参考，选择合适的行情/交易函数
3. 套用模板：从上方模板库中选择最接近的模板，填入参数
4. 代码生成：输出完整的 initialize + handle_data 策略代码
5. 说明要点：附上 PTRADE 平台运行说明（参数配置、回测 vs 实盘差异）

注意事项

• PTRADE 不支持实时 tick 数据回测，需订阅 Level2 才能用指定价格撮合
• open_long/close_long 的 volume 为股数（100 股整数倍）
• 回测撮合规则：默认 market 以当日开盘价成交；需订阅实时行情才能用 latest
• 因子选股建议用沪深300/中证500/中证1000成分股作为股票池过滤
• 组合持仓建议最多不超过20只，单只权重不超过10%