artherahq
OrganizationReusable Agent Skills for quantitative finance research, extracted from the Aria toolchain.
Categories
Indexed Skills (10)
backtest-validation
Validate whether a backtest result is trustworthy before drawing any conclusion from it. Trigger for "回测这个策略", "这个策略过拟合吗", "夏普这么高可信吗", "加上手续费还赚钱吗", "样本外表现", "validate this backtest", "is this strategy overfit", or whenever the user (1) presents or produces backtest results and wants a judgement, (2) asks whether an edge is real, (3) compares strategy variants and picks the best one, or (4) is about to deploy/paper-trade a strategy based on historical performance. Fire even when the user only asks for the metrics ("just show me the Sharpe") — the metrics are not a conclusion until the gauntlet passes. Pair with point-in-time-research: that skill guards the data going in; this one guards the claim coming out. Do NOT trigger for pure data fetching or for live trading questions with no historical simulation involved.
equity-research-report
Produce comprehensive, evidence-backed public-equity research reports with market, filing, fundamental, valuation, technical, risk, catalyst, scenario, and source analysis. Use when the user asks for "全面分析报告", "股票研报", "深度分析公司", "完整个股分析", "comprehensive equity report", or requests analysis of a stock, company, or equity thesis with current auditable data.
execution-position
Convert a trading signal into a sized, risk-gated PAPER order intent — or a refusal with reasons. Trigger for "这个信号该买多少", "帮我算仓位", "凯利公式下注多少", "现在要不要加仓", "position sizing", "how much should I buy", "simulate this trade", or whenever the user (1) has a signal/decision and asks for the size, (2) wants a trade checked against limits before acting, (3) asks about stops or exposure, or (4) another skill hands over a strategy/rebalance that needs execution shape. Fire even for casual "梭哈吗". Do NOT trigger for portfolio-wide weight construction (portfolio-optimization) or post-hoc risk analysis (risk-assessment). NEVER place live orders.
factor-research
Evaluate whether a cross-sectional factor genuinely predicts returns. Trigger for "这个因子有效吗", "算一下IC", "动量因子在A股还有效吗", "帮我评估这个选股信号", "factor IC", "is this signal predictive", "compare momentum vs value factors", or whenever the user (1) proposes or computes a ranking/score across assets and asks if it works, (2) asks which factor explains recent moves, (3) wants factors screened/ranked before building a strategy, or (4) hands a signal to strategy construction. Fire even for informal phrasing ("这个指标选股靠谱吗"). Do NOT trigger for single-asset technical indicator questions (no cross-section) or for validating a finished strategy's returns (that is backtest-validation).
multiple-testing-correction
Correct for multiple hypothesis testing before calling any factor, sub-period, or parameter sweep "significant" — and check whether a claimed set of "independent" signals or bets is actually a smaller number of correlated ones wearing a disguise. Trigger for "测了8个因子哪个显著", "筛选因子", "factor screen", "parameter sweep", "which of these signals is real", "p-hacking", "data snooping", "统计显著性", "breadth", "多重检验", "这些信号真的独立吗", or whenever the user has run more than one significance test in a batch and is about to act on which ones "passed." Also trigger when a Sharpe/IR target is being justified by citing a number of "independent" bets, signals, or symbols — that count needs a correlation check before it goes into any Information Ratio math. Pair with backtest-validation: that skill's Deflated Sharpe Ratio is this same correction applied to the single case of "best strategy variant out of N backtested"; this skill is the general case for any batch of hypothesis tests, plus the separate breadth-illusion check neither s
point-in-time-research
Enforce point-in-time data discipline for any quant research task. Trigger for requests such as "回测策略", "因子IC", "检查前视偏差", "量化选股", "验证夏普比率", or "point-in-time backtest". Also trigger this skill whenever the user: (1) backtests a strategy or factor, (2) asks for a Sharpe / alpha / IC / return — even as a quick one-liner ("just give me the Sharpe", "run this backtest"), (3) reviews or asks you to confirm / audit / validate quant or backtest code ("is my code PIT clean?", "confirm this is correct", "check my backtest for look-ahead"), (4) evaluates whether a signal or edge is real, or (5) screens on fundamentals. Fire even when the user claims they already fixed PIT issues — check specifically for whichever of the three silent leaks (period-end dating, latest-value overwrite, same-session execution) may remain. Do NOT trigger for generic finance questions that involve no simulation or data join.
strategy-generation
Turn a natural-language trading idea into a disciplined, testable strategy spec and implementation. Trigger for "帮我生成一个动量策略", "写一个BTC均值 回归策略", "设计多因子选股", "做一个ETF轮动", "generate a momentum strategy", "build me a trading strategy", "turn this idea into code", or whenever the user (1) describes a trading idea and wants it made concrete, (2) asks for strategy code, (3) wants strategy variants compared, or (4) asks to improve an existing strategy. Fire even for vague asks ("有什么好策略推荐吗" — the answer is a disciplined spec, not a tip). Do NOT trigger for validating an already-backtested result (backtest-validation) or for pure factor evaluation (factor-research).
ui-design-system
Help the user build and enforce *their own* design system — capture their brand/aesthetic decisions into a portable design-tokens.json, then keep every UI they generate consistent with it. Trigger for "帮我建立设计系统", "把我的品牌色定成规范", "make my UI consistent", "set up design tokens", "check this screen matches my design system", "为我的 app 定一套配色和圆角规范", "review my components against my tokens", or whenever the user is building UI and wants coherence across screens rather than one-off styling. Do NOT trigger to invent an aesthetic from a blank slate with no user input (ask them, or use ui-ux-pro-max for exploration first) — this skill operationalizes the user's OWN choices, it does not impose a house style. Works across targets (CSS/Tailwind, SwiftUI, React Native, Flutter) because tokens are platform-neutral.
gamma-exposure
Compute or review a Gamma Exposure (GEX) / dealer-hedging-behavior estimate from options open interest before it gets reported as if it were an observed fact. Trigger for "GEX", "gamma exposure", "gamma squeeze", "gamma wall", "zero gamma flip", "dealer positioning", "做市商对冲", "伽马敞口", "波动率压制/放大", "0DTE flows", "净伽马", or whenever the user is inferring options-dealer hedging pressure from open interest and implied volatility, building a GEX chart/panel, or about to state a GEX regime call ("dealers are long/short gamma at X") as if it were measured rather than estimated under an assumption. Also trigger when reviewing GEX-computation code for a silent sign or scaling bug — the failure mode this skill exists to catch produces a plausible-looking wrong answer, not a crash. Do NOT trigger for a plain options-chain display (bid/ask/volume/greeks per contract, no hedging inference) or for risk exposure of options the user actually holds — that is `risk-assessment`.
your-skill-name
Replace this with one or two sentences that say exactly when to trigger the skill, in the words a user would actually use. Name the concrete situations. Add a "Do NOT trigger for …" clause if the skill is easy to over-fire. This field is the only thing the model sees when deciding whether to load the skill.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.