@tangweigang-jpg

trading-agents-cn provides a multi-agent llm framework for a-share stock analysis, backtesting, and factor research with openai-compatible adapters. covers batch analysis, custom strategies, and semantic validation rules but lacks actionable step-by-step procedures.

cuemacro-finmarket provides a backtesting framework for fx and chinese equity markets with arcticdb tick storage, quandl data integration, and technical indicator strategies. execution requires careful adherence to semantic locks around signal generation and position sizing.

bt-portfolio-backtest wraps the bt framework for multi-strategy portfolio construction and backtesting, supporting risk parity and government bond rolling strategies. framework-level constraints are explicitly defined, but concrete execution steps are missing.

llama-index-rag documents a python framework for converting documents into queryable knowledge via a four-pillar retrieve-then-synthesize architecture. covers 52 constraints (5 fatal) but lacks actionable procedures and concrete trigger examples.

vectorbt-vectorized wraps a backtesting framework for multi-market strategy research and factor analysis. it defines strict execution rules (semantic locks) and anti-patterns but lacks actionable procedures for end-to-end workflows.

browser-use-agent wraps an llm into a web operator via async python, collecting dom snapshots and screenshots before issuing a single llm call to produce reasoning and action sequences executed over cdp. intended for form filling, scraping, and regression testing on non-deterministic pages.

e2b-sandbox-runtime provides isolated micro-vm execution for ai-generated code via python/typescript sdks. covers setup, basic usage patterns, and 42 constraints including fatal modes, but lacks step-by-step procedures for common tasks.

instructor-structured-output wraps pydantic models around llm calls across 20 providers via monkey-patching. covers schema injection, retry loops, and validation error recovery, but lacks concrete usage procedures and actionable troubleshooting steps.

pandas-ta-indicators wraps the pandas-ta library to compute technical analysis metrics (RSI, MACD, Bollinger Bands, KAMA) on multi-market data with visualization and parameter tuning. execution model and concrete procedures are absent; mostly metadata and constraint listings.

pyfolio-performance wraps pyfolio-reloaded to generate portfolio tear sheets (sharpe, drawdown, turnover, round-trip trades, sector attribution) from backtests. requires python 3.12+, uv, and doramagic ecosystem. covers a-share, hk, crypto markets with joinquant/eastmoney/baostock data sources.

dspy-prompt-optimizer exposes a python framework for composable llm modules with declarative signatures and automatic prompt/few-shot optimization via 14 teleprompter classes. covers rag, agents, classification, extraction. shipping with 44 constraints including critical pickle/cache and cost-control warnings.

quantaxis-data-platform covers a-share factor computation, tear sheet analysis, and backtesting via pandas/polars conversion and qifi account simulation. relies heavily on external reference files not included in this document.

lean-cloud-backtest wraps the LEAN engine for multi-market quantitative backtesting. it provides QuantBook data access, technical indicators, and factor modeling but lacks concrete procedural steps and relies heavily on external reference files.

finrl-meta-envs provides multi-market reinforcement learning environments for algorithmic trading with ppo/dqn backtesting and alpaca integration, but lacks actionable step-by-step procedures and relies heavily on external reference files.

backtrader-event-driven executes dual moving average crossover backtests with event-driven signal generation and pyfolio reporting. core skill is templated but lacks actionable procedural steps, relying instead on reference documents and locks that are not inlined.

openbb-terminal wraps multi-market financial data sourcing (stocks, crypto, commodities) with backtesting and technical analysis. execution requires explicit user intent matching on action verbs and market specification. actual step-by-step procedures are absent from this document.

finrl-rl-trading applies ensemble deep reinforcement learning (A2C, DDPG, PPO, TD3, SAC) for automated multi-market stock trading. executes backtesting and factor research across asia-pacific markets with strict semantic locks and anti-pattern guards.

tqsdk-futures-api provides real-time quote retrieval and backtesting for chinese futures and options markets, with factor analysis and volatility modeling. critical constraints and anti-patterns are documented but procedures lack step-by-step detail.

ccxt-crypto-api wraps major cryptocurrency exchanges with order management, market data, and account monitoring, but lacks concrete procedural steps and instead relies on external reference files that are not included.

talib-technical-analysis wraps 150+ ta-lib indicators for multi-market quantitative analysis, but lacks concrete procedural steps, actionable trigger phrases, and end-to-end worked examples. primary value is constraint/anti-pattern documentation rather than executable guidance.

macro-economic-model covers alm portfolio simulation, cash flow reporting, and eiopa yield curve calibration via smith-wilson. describes semantic constraints and anti-patterns but lacks executable procedure steps and concrete implementation guidance.

crewai-multi-agent documents a python framework for multi-agent llm applications with role-goal-backstory declarations, sequential and hierarchical execution modes, and dual tool-call loops. coverage is conceptual rather than procedural.

hummingbot-market-maker outlines crypto trading automation via funding rate arbitrage and cross-exchange market making, but lacks step-by-step execution procedures and concrete implementation guidance beyond use case taxonomy.

ml4t-book-notebooks wraps machine learning for trading notebooks with constraints on signal execution, entity formatting, and factor pipelines. structure is skeletal, procedures are absent, and critical steps rely on external reference files.

fava-beancount-viewer covers portfolio analysis and tax-loss harvesting for chinese equities via command-line interface. the skill references extensive constraint and anti-pattern documentation but lacks actionable step-by-step procedures and concrete output specifications.

edgar-crawler fetches sec filings (10-k, 10-q) for us equity fundamental analysis, with quarterly incremental updates and local caching. the skill lacks actionable procedures and concrete implementation detail.

arcticdb-timeseries wraps time-series data management for billion-row datasets with lazy loading and s3 backend support, but lacks executable procedures and concrete implementation steps beyond use case enumeration.

ifrs9-loss-engine claims to compute expected credit loss with vasicek and kaplan-meier methods, but the skill.md is a template stub that describes trading pipelines, macd triggers, and chinese stock entities rather than ifrs9 procedures or loss calculation logic.

advanced-financial-ml wraps mlfinlab for chinese equity backtesting with bar construction, fractional differencing, and factor research. heavily constrainted by semantic locks and anti-patterns; execution guidance is sparse and mostly referential.

credit-lgd-model claims to build loss-given-default models for credit risk assessment but provides no actionable procedure, conflates trading pipelines with credit modeling, and lacks executable steps or output contracts.

abs-cashflow-modeling addresses asset-backed securities structuring with mortgage pool simulation, bond waterfall distribution, and tranche performance analysis, but lacks procedural clarity and actionable steps for execution.

aml-data-generator converts transaction logs into synthetic AMLSim datasets for anti-money laundering testing, supporting account partitioning by bank ID and multi-source aggregation. however, the skill description does not align with the documented content, which focuses on chinese equity trading strategies rather than aml workflows.

quantlib-derivatives wraps QuantLib for pricing options, swaps, and bonds via SWIG bindings, with support for american finite-difference and multi-asset basket strategies, but lacks concrete execution procedures and step-by-step workflows.

financepy-derivatives wraps FinancePy for derivatives pricing, holiday calendars, and cash flow scheduling, but the skill.md conflates trading strategy execution with derivatives valuation and lacks concrete procedural steps.

py-vollib-options-pricing provides bsm and black model pricing for european options with greeks calculation and dividend yield support, but lacks actionable procedural steps and relies heavily on external reference files that are not included.

insurance-actuarial-python covers singular spectrum analysis and stationary bootstrap methods for interest rate time series decomposition, nss curve calibration, and derivative parameter fitting. execution relies on external reference files and lacks concrete procedural steps.

reactive-pricing-engine claims xva valuation for otc derivatives (cva/dva/fva) and simm margin analysis, but the skill.md is a config stub lacking actionable procedures, actual pricing logic, or failure handling.

beancount-plaintext-ledger claims to support double-entry accounting with bank statement import and financial reporting, but the skill.md mixes unrelated trading/backtesting pipeline content with accounting use cases, lacks actionable procedures, and references external files instead of defining intent or step-by-step execution logic.

insurance-loss-reserving applies chainladder methods to historical claims triangles for ibnr reserve estimation across reinsurance, catastrophe, and general liability lines, but the skill.md contains misaligned stock-trading boilerplate and no actionable procedure.

rotki-crypto-tracker describes a self-hosted portfolio tracking system for crypto assets across exchanges and wallets, but the skill.md is primarily metadata and constraint scaffolding with minimal actionable procedure. the single use case covers sphinx documentation configuration, unrelated to crypto tracking.

获取中国 A 股市场实时行情、历史 K 线、财务报表、基金期货等金融数据，支持股票、债券、期权等多品种数据查询。

分析alpha因子的预测能力与前向收益特征，生成分组收益、IC、换手率等报告，辅助量化策略的因子研究与事件分析。。

用 GARCH 族模型进行波动率建模与预测，支持夏普比率统计推断和 SPA 模型比较测试，应用于全球市场风险管理。

使用Fama-French因子模型进行气候ESG投资分析，支持月度股价数据下载、因子相关性计算、OLS回归诊断及显著性筛选，帮助用户构建因子组合和风险评估。

基于监督学习、决策树或聚类等多种算法，自动为评分卡变量生成最优分箱边界，同时支持单调性约束和缺失值处理。

处理信用评级转移矩阵，支持Not-Rated状态重分配、年度与月度矩阵转换、状态空间定义及数据集表征。

CZSC 缠论技术分析工具，支持 K 线生成、笔线段识别、分型信号提取与 A 股回测可视化。

为 VAlpha 量化终端用户提供 A 股市场数据获取、多数据源自动切换与熔断保护，支持 Tushare/Akshare 链路 fallback，并根据积分额度自动配置请求频率限制。

提供全球宏观经济数据仪表板视图，支持多源数据本地存储、冷热数据分离存储与自动化刷新调度。

计算投资组合风险指标，包括年化收益率、夏普比率、索提诺比率、最大回撤和卡玛比率，支持滚动窗口统计和 NaN 数据处理，适用于多市场数据。。

训练动态知识图谱嵌入模型，学习时序实体关系表示，支持链接预测和时间预测任务。

多智能体金融分析平台，支持股票研究、市场预测、财报解读与量化回测策略构建，覆盖全球市场数据分析。

执行银行系统级压力测试，基于EBA 2018真实数据计算CET1比率与杠杆率，模拟firesale情景下资产负债表韧性。

提供年化波动率、指数加权移动平均（EMA）和指数加权标准差等量化金融指标的专业计算能力，支持维度枚举到字符串的灵活覆盖，适用于金融时间序列分析与资产定价建模。

使用 NautilusTrader 配置驱动的 BacktestNode 运行高性能多市场回测，支持 Parquet 数据目录和外部 CSV 数据导入，策略可直接过渡到实盘交易。。

OpenSanctions 黑名单合规筛查：国际制裁名单、PEP（政要）、高风险人物数据的 抓取、去重、匹配与版本归档。适用于 KYC 和 AML 尽调。

验证 Frappe Lending 贷款模块核心流程，包括贷款申请创建、放款计划生成、还款处理及结清退款的自动化测试能力。

提供多策略投资组合优化框架，支持均值-方差、Black-Litterman 和分层风险平价（HRP）算法，内置多种协方差估计方法对比分析。

基于微软 qlib 的 AI 量化平台：覆盖预测模型、因子挖掘（Alpha158/TFT）、 组合优化、多频回测。支持 A 股 + 美股 + 港股多市场。

基于20日价格动量在沪深300、沪深500与国债之间自动轮转配置，通过RQAlpha框架执行完整回测并评估组合绩效。

提供多市场回测与强化学习交易环境构建能力，支持多交易所钱包组合管理、Plotly交互式交易可视化及RL智能体训练评估。

xalpha 支持多市场基金组合分析，实现 A/C 份额成本比较、可转债估值、组合业绩归因及基金相关性分析。

通过 Yahoo Finance 获取全球多市场股票、指数、外汇及加密货币的历史行情、财务数据、实时报价和财务日历。

使用 Zipline 框架执行日频股票策略回测，支持多市场数据接入、因子研究、可视化绩效分析，默认本金千万级。。

使用7种技术形态检测器（杯柄、三周紧绑、高紧旗、VCP、NR7等）按确定性顺序扫描股票池，支持跨检测器评分校准与置信度聚合排序。

Mem0 长期记忆层：为 LLM agent / chatbot 提供事实级记忆——抽取、嵌入、去重、存储 + 混合检索（语义 + BM25 + 实体加权），覆盖 17 个核心用例。自托管 Memory 与托管 MemoryClient 双形态。 Mem0 long-term memory layer for L...

从 SEC EDGAR 系统获取和解析公司监管文件，支持 SEC 文件检索、财务报表（10-K/10-Q）提取、内部人交易（Form 4）追踪及机构持仓（13F）分析。。

提供A股券商客户端自动化交易能力，支持雪球、芸享等多券商登录与交易操作封装，涵盖账户余额查询、持仓管理、委托下单及组合跟随等核心功能。

提供多市场财务分析能力，涵盖历史数据获取、财务报表解析、财务比率计算、固定收益分析、投资组合绩效评估和股票基本面筛选等核心功能。。

基于 lifelines 库提供生存分析与 Cox 比例风险建模能力，支持残差诊断、参数化回归模型自定义、时滞转化率分析及比例风险假设检验。

使用 Freqtrade 框架加载多交易所 OHLCV 历史数据并进行策略回测分析。

Darts 是轻量级时间序列预测库，支持多市场金融数据的确定性与概率性预测，提供协变量整合与层级聚合能力。

VeighNa（原vnpy）支持中国期货自动交易执行，集成日盘/夜盘交易时段管理，并提供CSI300成分股数据下载及Alpha101/LightGBM等因子研究工作流。。

Chroma 向量数据库：Rust 内核（v1.0.0+ 重写，2025-03），多语言客户端SDK。单节点用 PersistentClient（SQLite + 本地 HNSW）或 EphemeralClient（内存）；分布式 / 云用 SPANN + BLOCKFILE on S3/GCS。 Chroma...

LangChain v1：把 LLM、prompt、tool、retriever、parser 暴露为 Runnable，用 `|` 操作符（LCEL）组合成统一 invoke / stream / batch 接口的链。 LangChain v1: exposes LLMs, prompts, tools, r...

通过字节码驱动的复式记账引擎，支持多币种账户余额实时查询和资金来源的FIFO分配追踪。

实时获取多个加密货币交易所的市场数据流，支持异步回调处理并将交易、行情、订单簿等数据持久化到ArcticDB时序数据库。

A 股量化实验室：基于 zvt 框架的数据采集 + 因子研究 + 回测执行一站式。 覆盖 31 个场景——机构持仓、财报、指数成分、MACD/MA/量能择时。仅限中国 A 股。

AutoGen v0.4：asyncio actor-runtime 多智能体框架（autogen-core / autogen-agentchat / autogen-ext 三包）。 AutoGen v0.4: asyncio actor-runtime multi-agent framework (auto...

MCP Python SDK：Anthropic 主导的 Model Context Protocol 参考实现。2 server 层 + 3 transport + 3 原语 + 4 协议版本 + 50 条约束。 MCP Python SDK: reference Python implementation o...

自动化投资组合再平衡与交易执行，遵循先卖后买原则，支持多市场资产配置，智能计算最低交易规模及税费。。

基于 Qlib 的 A 股自选股智能分析系统，集成 LLM Agent ReAct 推理引擎和技术指标择时模块（MA 多头排列、乖离率阈值严进策略），自动生成每日 buy/hold/sell 指令并推送至微信。触发场景：(1) 用户要查询自选股当天的 AI 交易信号和涨跌预测；(2) 用户要获取符合 MA 多头排...