What problem does it solve? FutureFin's projection and FIRE engine can produce numbers that look plausible but are silently wrong, and ad-hoc fixes risk breaking determinism, Decimal money discipline, or client-server parity. This Skill provides a gated, evidence-driven campaign to audit, stress, and extend the economic model without regressions. ## Core Features & Use Cases - Baseline establishment (Phase 0): Runs engine unit tests, Postgres-backed integration tests including the fire-parity fixture, and frontend Vitest parity suites before any change. - Model fidelity inventory (Phase 1): Classifies every model simplification with code anchors and discriminating experiments whose predicted numbers are written down before running. - Ranked solution menu (Phase 2): Pre-registered candidates including property-based invariant testing with proptest, seeded Monte Carlo percentile bands, sequence-of-returns stress, tax-aware drawdown, and dynamic SWR. - Use Case: When asked to "add Monte Carlo simulation to the retirement projection" or "audit why the engine ignores loan interest", load this Skill to follow the fenced-off wrong paths, parity discipline, and promotion protocol instead of improvising. ## Quick Start Load this skill and run the Phase 0 baseline to audit the projection engine before proposing any realism improvement.
