Weighted Decision Simulation

Weighted Decision Simulation

A Claude Code skill for working through decisions that involve comparing a few discrete options, multiple weighted criteria, and uncertainty about the future. Examples: choosing between job offers, deciding how to allocate savings across investment options, or picking between competing plans where cost, risk, and personal priorities all pull in different directions.

Instead of a one-off script or spreadsheet, the skill sets up a small project with a clear split: a generic simulation engine, and a set of config files that describe your specific decision. Ask a new question by editing config.

What it does

Given a decision with multiple options, weighted criteria, and a time horizon, the skill:

• builds an outcome grid across every option and every scenario (or Monte Carlo draw), instead of collapsing straight to one number
• sweeps decision variables you specify, such as a contribution level or a term length
• scores options against your weighted criteria (financial return, risk, convenience, or whatever applies to your decision)
• runs sensitivity analysis and reports switching thresholds, for example "B overtakes A if its growth exceeds 4.1%/yr"
• backtests any forward-looking prediction against historical data before reporting it as trustworthy, or marks it UNVALIDATED
• keeps a history of past runs so later questions can be compared against earlier findings

Requirements

• Claude Code
• Python 3 with pytest for the generated project's test suite

How to use it

1. Open this folder in Claude Code.
2. Give Claude a prompt describing the decision you want to work through. See RUN_PROMPT.md for an example.
3. Claude reads SKILL.md and follows its workflow: survey any existing config, gather missing facts, write config/question.json, build only the engine capability the question needs, validate, run, and report back using a fixed ou