rebalance

name	rebalance
description	Analyze portfolio allocation drift and generate rebalancing trade recommendations across accounts. Considers tax implications, transaction costs, and wash sale rules.
user-invocable	false

Portfolio Rebalance

This skill is the corrective-action engine that portfolio-review should call whenever a hard gate or threshold trigger requires action.

Internal MCP Tool Map

• Current state: portfolio-analytics.get_condensed_portfolio_state, portfolio-analytics.validate_account_taxonomy
• Risk: portfolio-analytics.analyze_portfolio_risk, portfolio-analytics.analyze_hypothetical_portfolio_risk
• Drift: portfolio-analytics.analyze_allocation_drift, portfolio-analytics.analyze_bucket_allocation_drift
• Barbell gap math: portfolio-analytics.classify_barbell_buckets
• TLH and tax overlay: portfolio-analytics.find_tax_loss_harvesting_candidates; use household-tax.assess_exact_support only for supported exact household-tax cases
• Convex candidate ranking: market-intel-direct.rank_convex_candidates

Workflow

Step 1: Establish current state

• Run portfolio-analytics.get_condensed_portfolio_state
• Run portfolio-analytics.validate_account_taxonomy
• Run portfolio-analytics.analyze_portfolio_risk
• Run portfolio-analytics.classify_barbell_buckets

Step 2: Determine the repair objective

Always solve in this order:

1. get ES below 2.5%
2. raise hyper-safe to at least 15%
3. raise convex to at least 10%
4. reduce fragile-middle toward 70%
5. clean up residual IPS drift and tax inefficiency

If ES is critical, do not let drift cleanup override the de-risking objective.

Step 3: Funding order

Rank sells in this order:

• tax-deferred high-risk trims first
• taxable TLH-eligible losers second
• taxable gain realizations third

Use include_decomposition=true when needed to identify the biggest component-VaR contributors.

Step 4: ES-driven repair

When risk.status == "critical" or illiquid_overlay.adjusted_es_975_1d > 0.025:

• estimate a first-pass trim set from the highest-risk symbols
• prefer reducing overlapping equity beta before cutting diversifiers
• do not recommend new risk-adding trades until the verified post-plan ES is back below limit

Step 5: Barbell repair

Use portfolio-analytics.classify_barbell_buckets gap outputs directly:

• safe_gap_pct/value
• convex_gap_pct/value
• fragile_excess_pct/value

Then:

• fill the safe gap first when ES is still binding
• fill the convex gap second using ranked convex candidates
• use remaining trims to reduce fragile-middle

Step 6: Convex implementation

When convex is below target:

• run market-intel-direct.rank_convex_candidates
• prefer the primary_path_shortlist for the verified recommendation
• keep options-based ideas in Advanced Alternatives unless and until overlay verification is supported for options structures

Important:

• TLT is conditional, not default
• gold / managed futures / tail-risk ETFs should usually rank better in inflationary or stagflationary setups
• options candidates are valid ideas only when account capability supports them

Step 7: Verification

Do not pretend to verify by mentally adjusting weights.

• Use portfolio-analytics.analyze_hypothetical_portfolio_risk(target_allocations=...)
• Require the primary path to show:
  • verification_pass == true
  • proposed ES(97.5%) < 2.5%
  • post-plan barbell closer to policy targets

Step 8: Tax overlay

• run portfolio-analytics.find_tax_loss_harvesting_candidates
• avoid wash-sale conflicts
• if taxable gains are material, assess whether the case is inside the exact household-tax scope before using that server

Output Contract

The rebalance output should include:

• current constraint summary
• proposed trade list by wrapper/account
• primary path
• lower-tax alternative
• verification block from analyze_hypothetical_portfolio_risk
• post-plan barbell summary
• estimated tax impact
• unresolved caveats

Constraints

• ES <= 2.5% is binding
• the verified recommendation must be wrapper-aware and tax-aware
• options-based convex ideas can be shown, but do not make them the primary recommended path until the risk-verification layer supports them directly