<p align="center"> <img src="assets/readme-hero.svg" alt="High-Stakes Analytics & Decision Lab routes a question to an Evidence Intelligence Report and adds a Decision Intelligence Brief only when justified" width="100%"> </p>

# High-Stakes Analytics & Decision Lab

A platform-neutral, evidence-constrained Skill for moving from an ambiguous question to a defensible analysis—and only then, when justified, to action.

## What the Skill does

This Skill turns an ambiguous, consequential question into a reproducible evidence product and, only when justified, a bounded decision product. It can start from a question alone or from uploaded row-level data. Instead of assuming the data are clean, the model is useful, or a recommendation must be produced, it makes each transition conditional on visible evidence.

| Highlight | Instead of… | The Skill… | |---|---|---| | **Evidence first** | Starting with a favorite method | Declares the question, estimand, population, grain, horizon, lineage, and claim boundary | | **Data readiness** | Silently cleaning until a model runs | Preserves the source, profiles quality and privacy, and pauses on material choices | | **Adaptive analysis** | Filling one fixed report template | Selects only the descriptive, diagnostic, predictive, or prescriptive work supported | | **Honest endpoints** | Treating a recommendation as mandatory | Accepts an evidence request, negative validation, `do_not_deploy`, or bounded action | | **Shared uncertainty** | Disturbing alternatives independently | Retains common time, market, participant, operational, campaign, or spatial shocks | | **Reproducible evidence** | Separating prose from analysis | Links claims and accessible figures to JSON, CSV, hashes, and rerunnable code |

Its contribution is an adaptive orchestration system for evidence gating, method selection, claim control, dependent-risk handling, and case-specific analytical communication.

The public portfolio contains 15 primary reports and 10