quantskills

Multi-agent collaboration framework for Liangshuyuan tasks, organizing quantitative tools, build workflows, and task roles.

11 skills · 1 agent

AI-native quantitative research framework for reusable skills, strategy workflows, backtests, and reports.

9 skills

Scans dated price changes for calendar anomalies using robust tests, bootstrap checks, and multiple-testing control.

1 skill · 4 agents

Monitor FX-adjusted A/H premiums, historical extremes, dislocations, and daily cross-market price-discovery proxies.

1 skill · 3 agents

Computes a hot-money seat cooling and reversal factor from Dragon-Tiger and market data with validation artifacts.

3 skills · 1 agent

Event-driven top-N relay signal from daily A-share streak leaders: ten sub-factors, ICIR-shrinkage weights, rolling IC gate and sentiment filter; research-only candidate finder, not a strategy.

3 skills · 1 agent

Maintains the daily A-share limit-up pool with streak, blow-up and re-seal marks, theme groups, special patterns and sentiment metrics; outputs tables and an HTML dashboard; research only.

3 skills · 1 agent

Pulls the post-market A-share dragon-tiger list, tags broker seats as northbound, institution, hot-money, quant or branch, and builds a next-day watchlist with an HTML dashboard; research only.

3 skills · 1 agent

Screens A-share and US companies using moat, valuation, and point-in-time data for research records.

3 skills · 1 agent

Evaluates quantitative factors with IC, IR, stratified backtests, monotonicity, turnover, and decay diagnostics.

3 skills · 1 agent

Rebuilds A-share shareholder placard events across the 5%-30% disclosure tiers from top-holder snapshots, with intent, lock-up period and near-threshold watchlist; research only.

3 skills · 1 agent

Checks timing, IC/IR, costs, and neutralization quality in single-factor research.

1 skill · 3 agents

当需要设计、审查或排错期货对冲、期货仓位 sizing、合约移仓、基差/carry 分析、日历价差、保证金压力测试或 CTA 风格期货配置时，使用此 skill。适用于股指期货、商品期货、利率期货和跨期价差场景，重点处理合约乘数、名义本金、保证金、期限结构、交割规则和压力损失。

1 skill · 3 agents

当需要开发、计算、验证 Graham 净净营运资本(NCAV) 因子时，使用此 skill。适用于 A 股全市场深度价值筛选，排除银行/房地产/非银金融，计算 NCAV 折价因子并生成 buy/sell/hold 信号。

3 skills · 1 agent

Computes member-position concentration signals from institutional, hot-money, and northbound net positions.

3 skills

Computes a futures family-position reversal signal from seat-position relationships.

3 skills

Computes a futures family-versus-main-seat position-divergence factor signal.

3 skills

Builds look-ahead-safe backtests and audits leakage, survivorship, overfitting, costs, and out-of-sample checks.

1 skill · 2 agents

PandaData-only point-in-time Buffett moat hard screener for A-shares.

2 skills · 1 agent

A-share CSRC approval pipeline tracking: categorization by announcement level, approval flow monitoring, and status report generation.

2 skills · 1 agent

Price options and other derivatives and quantify their risk from a contract specification and market inputs, covering Black-Scholes-Merton analytical pricing, the full set of Greeks, binomial-tree and Monte-Carlo numerical pricing with convergence checks, implied-volatility inversion, volatility smile/skew structure, and model-risk validation (put-call parity, no-arbitrage bounds, cross-method consistency). Use when the user asks for 期权定价、希腊字母计算、隐含波动率、波动率微笑/曲面、二叉树或蒙特卡洛定价、Black-Scholes、美式期权定价、看跌看涨平价校验, or a one-stop option pricing and risk dossier.

1 skill · 2 agents

skill dl transformer multiasset

3 skills

A-share cumulative guarantee risk scanning: guarantee ratio, excess guarantee, and high-debt-ratio guarantee monitoring and alerting.

2 skills · 1 agent

Computes monthly agricultural-futures seasonality from daily prices and overlays crop-calendar context.

1 skill · 1 agent

Analyze A-share cross-sectional market participation, turnover concentration, liquidity distribution, speculative crowding, leader dependence, and structural fragility from daily or intraday stock snapshots. Use when the user asks in Chinese or English to analyze A-share market breadth, participation, turnover concentration, crowding, whether an index rally is broad or narrow, or to generate a reproducible market-structure report. Use only bundled scripts with PandaData as the primary live source, AKShare as fallback, or user-provided local data; never search, browse, or scrape webpages for replacement market data, and fail closed when approved sources are unavailable. Do not use for order routing, fill simulation, slippage/TCA, live trade execution, or stock-level margin/northbound/block-trade capital-flow attribution.

1 skill · 1 agent

Scans A-share macro, funding, valuation, trend, sector-rotation, and event evidence into risk levels.

1 skill · 1 agent

Builds and audits point-in-time A-share financial data without later restatements.

1 skill · 1 agent

Uses Pandadata to produce a sourced A-share dossier covering fundamentals, corporate actions, holders, event risks, and market funds.

1 skill · 1 agent

A-share tradability-constraint auditor — replays a backtest or signal trade log against date-resolved price limits (10%/20%/30%; main-board ST was 5% until the 2026-07-06 rule change), locked vs intraday-opened limit bars, suspensions, T+1 settlement, the naked-short ban, the new-listing window and turnover participation caps, then splits headline PnL into executable PnL and phantom PnL. Use when the user asks 回测收益是不是真的能成交、涨停买不进、 跌停卖不出、停牌怎么处理、T+1 约束、一字板、可交易性/可成交性审计, or wants to know why a live account underperforms its backtest. Fills the ecosystem gap: the four existing auditors all check whether the DATA is right, none checks whether the EXECUTION ASSUMPTION is possible.

1 skill · 1 agent

Provides a cross-sectional long-only backtest protocol with T+1 execution, fees, limit filters, and diagnostics.

1 skill · 1 agent

Use when an agent needs to independently audit the assumptions and biases behind a backtest / strategy code / research backtest report — execution timing and lookahead, trading costs, price limits and suspensions, survivorship bias, parameter freedom and multiple testing, data alignment and adjustment, turnover and capacity, benchmark and excess returns, and reporting transparency. Outputs a structured defect list (axis x evidence x severity x impact x fix).

1 skill · 1 agent

Backtests ETF time-series and cross-sectional strategies using PandaData and local Parquet with explicit next-day execution, price and cost assumptions; research only.

1 skill · 1 agent

Builds an A-share block-trade radar from discount or premium, volume, and price evidence.

1 skill · 1 agent

Monitors A-share buyback lifecycles, purposes, price ranges, and intensity for research.

1 skill · 1 agent

Aggregates margin, northbound-holding, and block-trade data into consensus, divergence, and crowding-percentile signals.

1 skill · 1 agent

Discover causal alpha factors from OHLCV data using causal discovery (PC + LiNGAM + NOTEARS), build Structural Causal Models, construct regime-invariant factor expressions, and validate through backtesting. Produces OHLCV-only alpha factors whose predictive power is grounded in causal mechanisms rather than spurious correlations. Use when an agent needs to discover stable alpha factors that survive regime changes, test whether existing factors are causal or merely correlational, or generate alpha factors with formal invariance guarantees on portable agent platforms such as Claude Code, Codex, or Codex-style skill systems.

1 skill · 1 agent

Analyzes A-share convertible bonds with double-low screening, terms, equity linkage, Greeks, and volatility.

1 skill · 1 agent

Builds commodity-futures carry, time-series momentum, cross-sectional momentum, basis, and inventory factors for rotation backtests.

1 skill · 1 agent

Monitors A-share concept and theme momentum, breadth, and rotation for research reports.

1 skill · 1 agent

Audits split and cash-dividend consistency between raw and adjusted equity prices before research.

1 skill · 1 agent

Monitors A/H and China ADR cross-listing parity using prices, FX, and share ratios.

1 skill · 1 agent

Aggregates cross-market prices, sectors, flows, and news into a daily Markdown review.

1 skill · 1 agent

Turn unstructured A-share disclosure text from cninfo (巨潮) and the SSE/SZSE exchanges into a traceable, structured event table (监管问询/诉讼担保/重组/治理/ 停牌控制权变更/增减持/质押/业绩预告). Use when the user asks to scan A-share announcements, find 关注函/问询函/诉讼/停牌/易主 events, build an event table for factor or alert pipelines, or backtest disclosure events. Fills the information-search gap Pandadata does not cover.

1 skill · 1 agent

Calculates A-share rolling dividend yield, dividend continuity, and ex-dividend calendars.

1 skill · 1 agent

Builds A-share heterogeneous graphs for GNN stock selection and backtesting.

1 skill · 1 agent

Defines OHLCV alpha-expression formats and validation rules for document-derived factors.

1 skill · 1 agent

Formal CAR event study around earnings/corporate events — abnormal returns, multi-window CARs, t-tests and sign tests, with sample and model disclosure; research only, no trading advice.

1 skill · 1 agent

Scans earnings guidance, industry distributions, and qualified audit items during earnings seasons.

1 skill · 1 agent

Monitors A-share ETF primary/secondary-market premiums and redemption-basket feasibility.

1 skill · 1 agent

Evaluates domestic non-QDII passive equity-index ETFs and comparable-index peers.

1 skill · 1 agent

Scans watchlists or holdings for unlock, pledge, ownership-change, and earnings-event risks.

1 skill · 1 agent

Computes Alpha101 and Alpha191 factors from long-form OHLCV CSV and outputs wide CSV.

1 skill · 1 agent

Runs long-only cross-sectional factor backtests on supplied factors and market data with diagnostics.

1 skill · 1 agent

De-redundantly weights and combines multiple factor signals into a composite signal.

1 skill · 1 agent

Provides a symptom, cause, and verification playbook for factor failures.

1 skill · 1 agent

Analyzes decay in Rank IC, turnover, and bucket returns and estimates half-life.

1 skill · 1 agent

Checks factor panels for drift, coverage, missing values, duplicates and distribution changes, with optional predictive degradation analysis; research only.

1 skill · 1 agent

Scores a cross-sectional factor using IC, Sharpe, drawdown, monotonicity, and turnover.

1 skill · 1 agent

Wraps grouped factor-processing workflows and their pipeline diagrams.

1 skill · 1 agent

Diagnose factor IC decay, ICIR, significance, rolling stability, and multi-horizon half-life. Evidence-first, no trading signals.

1 skill · 1 agent

Generates candidate factor ideas with economic rationale and risk notes from the default data scope.

1 skill · 1 agent

Local closed-loop factor research: generate, validate, backtest, diagnose, learn, iterate N rounds. Self-improving alpha discovery and optimization.

1 skill · 1 agent

Uses a FactorMAD-style multi-agent debate framework for interpretable stock-alpha mining.

1 skill · 1 agent

Provides a factor-mining SOP from hypothesis and experiment notes through scoring and accept/rollback.

1 skill · 1 agent

Mines factors with PandaAI data and feedback or extracts them from public documents.

1 skill · 1 agent

Runs parameter sweeps, ablations, and version refinements for existing equity or futures factors.

1 skill · 1 agent

Orthogonalizes cross-sectional factors with daily OLS and outputs residual factors and exposure diagnostics.

1 skill · 1 agent

Generates mutation, crossover, and recommendations from evaluated seed factor pools.

1 skill · 1 agent

Runs mRMR or Marginal-SAGE on local factor and label data to produce Top-K rankings.

1 skill · 1 agent

Scans a factor library and experiment logs for inventory, structural analysis, and research recommendations.

1 skill · 1 agent

Tracks Weibo user "峰哥亡命天涯" (Feng Ge), a well-known A-share reverse indicator. Extracts stock/market opinions from his posts and generates contrarian trading signals.

1 skill · 1 agent

Aggregates financial headlines and market data to select headlines and draft analysis articles.

1 skill · 1 agent

Audits probability-forecast calibration rather than sample ranking alone.

1 skill · 1 agent

Generate alpha factor expressions from fundamental data (PandaData). Accepts a document, URL, natural language query, or model invention and returns validated point-in-time factors.

1 skill · 1 agent

Uses PandaAI fund data and ETF constituents to estimate holdings, concentration and sector/style exposure with risk notes; estimates do not replace full actual holdings.

1 skill · 1 agent

Commodity-futures CTA factor library — computes a structured date×variety factor panel (time-series & cross-sectional momentum, carry/roll, term structure, positioning/COT, inventory, volatility). Use when the user asks for 商品期货因子、 CTA 信号、动量/carry/期限结构/库存/持仓因子, a factor panel for futures backtesting, or futures factor IC. Fills the ecosystem gap of ZERO futures factor libraries (vs 10 for equities). Emits factor values for the factor toolchain (factor-evaluate / ic-analysis / backtest), NOT human-readable reports.

1 skill · 1 agent

Turns futures DeepView natural-language requests into data-call plans and fact/inference-separated reports.

1 skill · 1 agent

Futures research Skill for analyzing, comparing, and screening futures markets with technical indicators, futures structure, and committee-style reports. Use when Codex needs to analyze a futures symbol, compare futures symbols, screen futures candidates, explain signal changes, or generate a structured futures research report without stock analysis, auto-trading, or deterministic buy/sell promises.

1 skill · 1 agent

Audits continuous-contract selection, roll gaps, and adjustment factors and produces a roll ledger.

1 skill · 1 agent

Build an auditable futures factor from contract transitions and crowding shifts

1 skill · 1 agent

Uses public data to study global commodity-futures term structure, roll yield, and spreads.

1 skill · 1 agent

Produces sourced global macro briefs from public rates, central-bank, and FX data.

1 skill · 1 agent

Turns global signals and public daily prices into backtestable research strategies, positions, and risk rules.

1 skill · 1 agent

Build auditable factor diffusion and residuals on point-in-time equity graphs

1 skill · 1 agent

Summarizes HK/US sell-side ratings, target prices, growth expectations, and changes.

1 skill · 1 agent

Organizes cross-period HK/US target-price and rating revisions into a research report.

1 skill · 1 agent

Generates HK and US equity dividend-event reports using Pandadata overseas-market interfaces.

1 skill · 1 agent

Scans HK and US insider transactions, net direction, trading clusters, and holding changes.

1 skill · 1 agent

Builds HK and US equity snapshots covering quotes, liquidity, valuation, and industry-relative position.

1 skill · 1 agent

Tracks A-share holder counts, top-holder concentration, and free float to assess ownership concentration.

1 skill · 1 agent

Evaluates quantitative factors through IC, grouped performance, and predictive effectiveness.

1 skill · 1 agent

Research-only RL lab for intraday timing on minute bars: Gym env, baseline policies, leakage-aware walk-forward train/eval. No live orders.

1 skill · 1 agent

Generates an event-ranking factor from Dragon-Tiger seat history, win rate, payoff, and next-session premium.

1 skill

Computes a futures top-20-seat position-change factor and signal from net-position data.

1 skill

Multi-factor portfolio health guardian producing a health matrix, crowding alerts, retire/rebuild candidates, IC decay curves, and a research-only effectiveness backtest L4 page.

1 agent

Assesses A-share financial health from audit opinions, statements, and industry benchmarks with risk checks.

1 skill

Applies entry-date and open-price rules for take-profit, stop-loss, forced exits, and single-name position caps.

1 skill

Manages intraday multi-instrument positions using sellable and locked quantity, price, and cash inputs.

1 skill

Evaluates backtest overfitting and multiple-testing risk with DSR, PBO, purged cross-validation, and Harvey-Liu haircut.

1 skill

Corporate governance scoring agent with 9-dimension risk scoring and evidence chains

1 skill

Cross-market corporate event radar: aggregates A-share placement/unlock and earnings events with HK/US dividend, earnings, meeting, and IR events into a standardized daily dashboard.

1 agent

Provides an all-weather allocation and backtest workflow for A-share assets, bonds, gold, and commodities.

1 skill

深度自编码器无监督异常检测 —— 用户问「今天哪些股票走势最异常」「沪深300 异常股」「AE 跑一下」「找不像自己往常样子的股票」类问题时触发。对沪深300成分股用最近60日窗口重训一个MLP自编码器，输出T日重建误差Top-10异常股，按「样式② 结构化播报」呈现。

1 skill

使用因果扩张 Temporal Convolutional Network 对沪深 A 股分钟线执行未来 1、2、3、5 个交易日的横截面收益排序研究，并以同数据、切分和预算的 LSTM 作为基准，生成可审计的数据、训练、速度和预测效果证据。用于运行或诊断 TCN 短线预测、核验感受野与因果卷积、PIT/walk-forward/purge/embargo、防止未来泄漏、比较 RankIC/Top 区域指标与训练吞吐，或判断研究模型是否达到冻结候选条件；不用于组合换手优化、券商连接、实盘交易或收益承诺。

1 skill

财报季 Surprise/暴雷猎手 Agent。获取财报预告、一致预期、审计意见，计算偏离度并生成分析报告。支持A股/港股/美股。

1 skill

每日盘后 ETF 资金流雷达 —— 用户问「今天/最近 ETF 有什么异动」「ETF 资金流看下」「哪些 ETF 在被大量申购/赎回」「ETF 净申赎异动」类问题时触发。扫描主流权益 ETF，输出三类信号（净申赎异动 / 折溢价背离 / 连续多日大额同向），以「样式② 结构化播报」呈现给用户。

1 skill

Computes and validates A-share valuation, quality, and growth factors from quarterly financial reports.

1 skill

Uses public sources to structure research evidence and risks for photonics, optical-network, and AI-infrastructure companies.

1 skill

Organizes, retrieves, and studies Gao Shanwen's public writings and articles.

1 skill

Generates nine-dimension Hong Kong equity due-diligence reports from Pandadata interfaces.

1 skill

Build, standardize, compare, and validate multi-factor fundamental panels for Hong Kong and US equities from PandaAI operating, market-financial, industry-median, price-volume, and financial-statement APIs. Use for quality, value, growth, momentum, low-risk, composite-score, percentile-rank, or cross-market stock-screening tasks.

1 skill

Build, compare, and validate institutional ownership structure panels for Hong Kong and US equities with PandaAI investor concentration, ranking, and shareholder-report APIs. Use for ownership breadth, top-holder dominance, HHI, controlling-holder risk, evidence confidence, or within-market institutional ownership screening.

1 skill

Generates macro-driven industry-rotation research materials from clock phases, nowcasts, and valuation filters.

1 agent

SSQuant Agent workflow for futures strategies, data services, CTP gates, and Chinese backtest reports.

1 skill

Canonical template for portable QuantSkills agent projects.

1 agent

Uses Pandadata price-series correlation changes to identify style shifts, diversification stress, and structural market moves.

Monitors crowded-trade risk from Pandadata price, turnover, margin, and Dragon-Tiger heat evidence.

Monitors derivatives sentiment from option implied volatility and underlying historical volatility.

Multi-agent futures research, strategy generation, backtesting, and research feedback workflow

Runs reproducible event studies for index additions, deletions, and weight changes.

Analyzes A-share index valuation percentiles, relative industry valuation, and rotation signals.

Monitors A-share institutional research activity, attention, and changes over time.

Audits normalized intraday OHLCV data for timestamp, gap, price, volume, and trading-date defects.

Combines research evidence, valuation, and risk information into an auditable investment decision report.

Converts JoinQuant strategy code into PandaAI JSON configurations runnable in backtests.

Uses long-term expectations and contrarian analysis to identify optimism, pessimism, and value traps.

Researches private placements, restructurings, spin-offs, and distressed turnarounds as special situations.

用截图或只读 PandaData 行情识别股票/期货K线趋势结构、蜡烛线和候选图表形态，支持日线与1/5/10/15/30/60分钟线，输出证据、确认条件、失效条件和不确定性。

Mine formulaic alpha factors end to end: LLM 主导生成候选公式 → 三层校验（白名单/量纲/前视）→ warm-start 遗传编程精修 → AlphaEval 五维打分 → LLM 经济解释 → 自包含 HTML 报告，返回结构化因子结果。Use when the user wants to mine/discover alpha factors, generate formulaic (expression-tree) trading factors, run LLM+GP factor search, or evaluate factor predictive power (RankIC) on stocks or futures. 只挖因子、不做回测（回测归另一 skill）。

QA over A-share financial reports and announcements with exact numeric computation, BM25 retrieval and disclosure-sourced citations; refuses when the corpus lacks coverage. Research only.

Uses high-frequency alternative macro data for industry nowcasts and trend monitoring.

基于 PandaData 的宏观事件—期货预期分析：读取宏观经济日历的实际值、市场预期与前值，结合期货价格、成交量、持仓量、基差、期限结构、库存与仓单，对沪金、沪铜、原油及用户指定品种输出基准、偏强、偏弱情景。当用户询问「美国CPI对期货影响」「宏观事件期货预期」「美联储对商品影响」「期货供需与宏观共振」「事件公布前情景」时触发。仅作条件化研究，不承诺涨跌或生成自动交易指令。

Monitors macro data, industry conditions, economic calendars, and recurring macro changes.

Computes moving-average crossover signals and reports trend state, latest cross, MA gap, and price bias.

Generates Pandadata-based A-share after-close daily market review reports.

Classifies A-share market regimes using index, macro, futures term-structure, and volatility features.

Monitors market regimes from Pandadata index breadth, volatility, and funding evidence.

A-share material contract alpha factor: cross-sectional ranking based on log-sum of contract amounts.

Minute-bar VWAP deviation strategy research and auditable SSQuant futures validation. Use for rolling VWAP deviation signals, confirmed mean reversion, trend guards, next-bar execution, execution-cost modeling, frozen-data backtests, and trade-result review.

Generate alpha factor expressions from minute-level market data using PandaData (panda_data). Takes a research document or natural language query as input, fetches intraday OHLCV bars via get_market_min_data, extracts rich minute-level features (VWAP divergence, intraday momentum decay, order flow proxy, volatility curve shape, trade concentration, lead-lag signatures), and produces validated alpha expressions backed by a formula contract. Use when an agent needs to generate alpha ideas from academic papers, market commentary, or research notes that leverage intraday microstructure signals invisible in daily data.

Ensembles machine-learning models into factor meta-signals with leakage-aware rolling validation.

审计任意金融特征、可训练时序模型或候选策略收益，并执行防泄漏的 Purged K-Fold、Embargo、CPCV、Causal Walk-Forward、PBO、DSR、Governed Holdout 与 Temporal Forward Evidence。用于检查未来函数、信息区间重叠、特征可用时间和血缘、Fold-Local 预处理、CPCV Path 稳健性、策略选择过拟合、预测是否在标签成熟前登记，以及在接受金融模型或策略前生成结构化验证证据。

Optimizes quantitative multi-factor model hyperparameters with fixed validation and trial-level evidence.

Applies a multidisciplinary mental-model framework to company investment research and judgment reports.

Collects, verifies, and analyzes A-share financial-news sentiment for research reports.

Monitors northbound flows, margin trading, and futures conditions with multiple risk signals.

Detects lookahead and data leakage in quantitative research workflows through numerical checks.

Combines futures, EIA, OPEC, and market data into Chinese crude-oil briefs.

Build auditable factors from point-in-time cross-sectional transport

Builds option strategies with legs, payoff charts, breakevens, Greeks, and margin analysis.

Analyzes option chains, implied and historical volatility, term structure, skew, and volatility premium.

Discovers and validates HK and US cross-sectional alpha factors by IC, decay, and turnover.

Identifies A-share oversold-rebound conditions and screens candidate stocks.

Computes and interprets asset-pair correlations, rolling relationships, and research uses.

Supports PandaAI factor onboarding, online mining, batch backtests, and cost review.

Audits PandaAI workflow graphs, code, timing, parameters, and backtest-validation evidence.

Generates importable PandaAI workflow JSON and embedded strategy or factor code from quant ideas.

Provides Pandadata market and research API calls and contract lookup across agent runtimes.

Manages local Pandadata DuckDB and Parquet quantitative data warehouses, caches, and queries.

Supports paper search, data extraction, experiment reproduction, and research-result reporting.

Use when an agent needs to attribute the returns of an A-share quant strategy or portfolio — decompose performance into Alpha/Beta/timing, Brinson allocation/selection/interaction effects by industry, and factor-return contributions (style factors, industry, residual Alpha). Outputs a unified AttributionReport (Markdown + JSON) with reconciliation checks.

Attributes active portfolio returns to industry allocation, stock selection, interaction, and factor contributions.

Black-Litterman 组合优化 —— 用户问「跑一下 BL 组合」「基于视图的组合权重」「相对沪深300 的主动配置」「动量/反转/换手视图对权重的影响」类问题时触发。以沪深300 指数权重为先验，用动量/反转/换手率三条因子视图更新，输出长权重组合，按「样式② 结构化播报」呈现给用户。

Aggregates concentration, benchmark deviation, valuation, quality, and risk exposures into a portfolio health report.

当需要开发、计算、验证 CVaR 尾部风险最小化组合时，使用此 skill。在预期收益约束下最小化组合 95% CVaR，支持极值理论(EVT/GPD)尾部补样、组合权重求解、样本外验证。

Estimates portfolio liquidation days, horizon cash, redemption shortfalls, and impact costs under volume stress.

Turns alpha signals into optimized weights under weight, sector, exposure, and turnover constraints.

Attributes realized portfolio returns by security and sector while reconciling fees, benchmarks, and input quality.

当需要开发、计算、验证风险平价（等风险贡献 ERC）组合时使用。手写 Ledoit-Wolf 收缩协方差稳定相关性，scipy SLSQP 求 ERC 权重，支持指数/期货/ETF 三类资产与月度 rebalance。

Screens A-share stocks after market close using technical patterns and capital-flow evidence.

Translates a five-part user questionnaire into factor directions and strategy parameters.

Converts approved trade targets into observable, cost-aware execution plans.

Provides an OHLCV directional-factor library for trend, breakout, and reversal research.

Provides an OHLCV factor library for volatility, chart-pattern, and drawdown-pressure research.

Batch-generates, validates, and packages framework-neutral OHLCV factor skills.

Analyzes portfolio risk exposures, constraints, and stress scenarios.

Guides quantitative research, backtest design, and statistical validation workflows.

Registers quantitative experiments and audits their reproducibility evidence.

Check recent strategy or factor deterioration using returns, IC, drawdown, turnover, benchmarks, and market regimes.

Tracks A-share refinancing lifecycles, pricing, and dilution risk.

Aggregates and grades A-share regulatory and compliance risk events.

Guides conversion of research reports into reproducible analysis workflows.

Selects factor combinations using residual IC and out-of-sample evaluation.

Builds a multifactor risk model and performs risk attribution.

Calculates risk-return metrics for portfolios or strategies.

skill rl portfolio allocator

Estimates rolling beta exposure of assets or portfolios relative to a benchmark.

当需要分析市场状态、行业轮动、ETF 强弱排序、市场宽度、风格切换或战术配置信号时，使用此 skill。适用于 A 股、港股、美股和 ETF 池的行业轮动分析、risk-on/risk-off 状态识别、相对强弱排名、宽度确认、假突破过滤和轮动失效条件设计。

Reconstructs Serenity-style research logic from public X/Twitter evidence.

Cost-aware A-share five-session cross-sectional return reversal research.

Converts one stock signal into benchmark-relative target weights with auditable risk, exposure, turnover, and cost controls.

Audits quantitative-signal stability across time and samples.

Studies A-share pairs trading with cointegration, spreads, and execution constraints.

Analyzes LHB seats, northbound activity, and capital-flow consensus or divergence.

Detects self-reinforcing reflexive regimes in A-shares with a fast/slow dual-loop model and outputs an eight-stage classification plus position discipline; no top or bottom forecasts. Research only.

Orchestrates SSQuant strategy research, paper trading, and runtime checks.

Turns natural-language trading ideas into a reusable Trader Skill and delegates simulated deployment.

Builds statistical-arbitrage time-series research and produces traceable reports.

Performs multidimensional research analysis of US memory-chip stocks.

skill stock score

Screens A-share stocks from natural-language criteria and Pandadata evidence.

Estimate strategy capacity and transaction costs from trades and ADV — participation, square-root/linear impact, capacity curve and breakeven vs alpha. Estimates only, no trade advice.

Use when an agent needs to generate a periodic performance report for a LIVE A-share quant strategy — 日报/周报/月报/半年报/年报 or custom period — covering returns, risk, trade analysis, and position/turnover detail, with embedded visualizations. Outputs a self-contained offline HTML dashboard (interactive ECharts) plus Markdown + JSON.

Generates strategy-performance tearsheets with risk-adjusted metrics.

Audits point-in-time universe membership, identities, and delisting returns before backtests.

当需要开发、计算、验证 Templeton 全球价值多因子 V2 时，使用此 skill。适用于 A 股/港股/美股跨市场价值筛选，基于 EP/BP/SP/股息/ROE/杠杆/动量 七子因子截面打分，生成 buy/sell/hold 信号。

Diagnoses financial time series and produces analysis reports.

skill tqx research

Trade review skill that turns trade records into attribution, pattern detection, per-trade review, and actionable advice.

Decomposes fills against VWAP/TWAP into transaction-cost components.

Calibrates commission, spread, slippage, and market-impact assumptions from execution and market data.

计算A股日线下跌突破划线压力位反转因子，识别下破下降压力线后反转确认的个股；当用户提供交易日期、要求自动获取当日全市场A股行情、量化选股、因子研究或回测时使用。

Harvests and structures public US SEC EDGAR filings.

Generates factual reports on US sector performance, valuation, and rotation.

Validates cross-sectional signals out of sample with purged and embargoed rolling windows.

Retrieves similar A-share and futures K-line patterns from text or image descriptions.

Builds trader-specific research-model skills from public X/Twitter post data.