portfolio-hedge

Portfolio & position hedging (ERM3 cascade + Lstar)

Report the ETF hedge legs the decomposition implies for a position or portfolio. This skill wraps the hosted RiskModels MCP and reports tool outputs; it computes nothing itself.

What to call

• riskmodels_get_my_positions — the caller's linked book (Alpaca Connect, Connect Trade, or Plaid). Call this first when the user means their account and has not pasted tickers. Pass for_analysis and for_hedge to the long-only portfolio tools below; when has_shorts is true or the book has more than 100 names, pass for_exposure to riskmodels_portfolio_exposure. This server does not place orders.
• riskmodels_decompose with dollars — one ticker: scale the L-level ETF hedge ratios to a dollar position.
• riskmodels_analyze_portfolio — a long-only book: holdings-weighted L1/L2/L3 hedge-level aggregate; with dollar positions it also scales the hedge ratios at the chosen cascade level (level L1/L2/L3) into ETF USD hedge legs.
• riskmodels_portfolio_exposure — a signed long/short book (dollar values, short < 0, up to 1000 names): beta-dollars by sector, the ETF trades that neutralise the book at L1/L2/L3, and the systematic vs residual risk split. Use it whenever any position is short or the book has more than 100 names; riskmodels_analyze_portfolio takes long positions only (up to 100). Report any coverage.dropped names or warnings to the user.
• riskmodels_get_residual_returns (1-100 tickers) — dispatch the simplest cascade level that clears the marginal-ER threshold for a name (or basket) and return the residual-return series after that hedge.

For an ETF or 13F filer named instead of pasted tickers, resolve real holdings first (riskmodels_search → riskmodels_get_etf with holdings_top, or riskmodels_get_filer with holdings_top / riskmodels_get_fund with holdings_limit) and treat the result as the portfolio. Never fabricate or approximate holdings — if no tool covers the portfolio, say so and ask the user to paste the positions.