Use this skill when the task explicitly requires formal quantitative finance methods: VaR, CVaR, portfolio optimisation (mean-variance, Black-Litterman, risk parity), Monte Carlo simulation, backtesting, factor regression, GARCH volatility modelling, or statistical validation. Do NOT use for qualitative commentary, narrative valuation, report writing, or basic portfolio review without statistical requirements. This skill is a sub-engine that returns structured results for reintegration into investment-lens.
Specialized quantitative engine for the broader investment workflow. Runs formal quantitative analysis and returns structured results for interpretation by investment-lens.
Use only when the task explicitly requires one or more of: portfolio optimization, VaR/CVaR, Monte Carlo simulation, risk decomposition, factor regressions, volatility modeling, time-series analysis, backtesting, or statistical validation.
riskfolio-lib optimisation silently drops assets with zero variance. Check for zero-vol assets before running.reintegration_note field. investment-lens will not reintegrate without it.Expect or construct:
objective, tickers, portfolio_weights, base_currencybenchmark, lookback_period, risk_free_proxymodel_type, constraints, valid_as_ofIf required inputs are missing, ask for them or state assumptions clearly.
Load references/input-schema.md for full schema detail.
Historical volatility, drawdown, VaR/CVaR, stress framing, correlation and concentration diagnostics.
Mean-variance, Black-Litterman, risk parity, equal risk contribution, constraint-aware construction.
Factor exposure, rolling regression, event studies, GARCH-family volatility, econometric validation.
Monte Carlo, strategy comparison, parameter sensitivity, historical backtests with stated limitations.
Always: state assumptions before results; use explicit lookback periods; state data limitations; return structured outputs; distinguish descriptive from predictive statistics.
Never: present model output as certainty; hide parameter choices; extrapolate beyond model scope without warning.
Return results in this structure:
analysis_type | objective | inputs_used | assumptions
summary_statistics | model_output | risk_findings
limitations | valid_as_of | reintegration_note
Where relevant, include: optimized_weights, var, cvar, beta, factor_exposures, drawdown, scenario_results.
Load references/output-schema.md for full schema detail.
When used as sub-analysis engine:
reintegration_note.investment-lens.Load when needed:
references/input-schema.mdreferences/output-schema.mdreferences/model-selection.md