Financial Planning

Produce a financial plan a careful skeptic would sign: every fact verified, every assumption named with its sensitivity, risk expressed as the spending floor the person would actually live on, and the whole thing operable through written decision rules rather than a one-time forecast.

Why this skill exists

Self-built retirement models, including sophisticated ones with Monte Carlo engines and dynamic withdrawal rules, fail in recurring ways: spending targets that quietly contradict the person's actual spending, a single return assumption at the optimistic edge with no sensitivity shown, "100% success" claims that are true by construction because the model cuts spending instead of depleting, healthcare bridges and their tax interactions left unmodeled, horizons that stop at average life expectancy, and tax facts that were stale the year after they were typed. Professional plans fail differently: generic assumptions, no connection to the life actually being planned. This skill exists to produce the plan neither of those produces, and every stage below traces to one of those failures.

Core principles

• Verified beats recalled. Any fact that changes or varies by jurisdiction (contribution limits, tax brackets, benefit ages and formulas, healthcare rules) gets checked by web search at plan time and logged with source and date. A plan resting on remembered numbers is wrong on arrival or wrong within a year; the verified-facts register (see references/verification.md) is what separates this plan from the one it replaces.
• Actuals beat estimates. Spending comes from statements and exports wherever they exist. The gap between what people say they spend and what they spend is the single most common silent error, and it propagates into every headline number.
• Sensitivity is a first-class result. Never present one future. The deliverable's headline is a range with the assumptions that move it, and the return assumption always appears at the user's value and at least two more conservative values.
• The floor is the risk metric. For any plan with adaptive spending, report the 10th-percentile lifetime spending and the worst pre-benefit stretch, judged against the person's actual lifestyle. Survival percentages are reported but never headlined.
• A plan is a document plus decision rules. The forecast will be wrong; the plan should say, in observable if-then form, what to do when it is. Deliverable structure in references/deliverable.md.
• The model serves a life. Income phases, exits, sabbaticals, part-time years, and lumpy expenses come from the life being planned. When a life-paths workspace exists, consume its chosen path and finances files as input; when standalone, elicit the life shape in intake. If the person has not actually chosen what life they are financing and the conversation keeps sliding into "but what should I do," that is the life-paths skill's job; suggest it once and continue with the shape they give you.

Code policy