What problem does it solve? Teams often start projects without agreeing on what success looks like, then cherry-pick favorable metrics after seeing results. This Skill locks in a primary metric, guardrails, baselines, and decision rules before execution so outcomes can be judged objectively. ## Core Features & Use Cases - Primary Metric Definition: Select one deciding metric with baseline, target, source, and an explicit decision rule instead of a dashboard of competing numbers. - Guardrails and Anti-Metric-Shopping Lock: Add safety, cost, latency, and quality constraints, and require human approval before any post-result metric change. - Measurement Plan and Gate Status: Specify instrumentation, evaluation windows, and confidence tolerances, then emit a ready/blocked gate verdict with telemetry. - Use Case: Before launching an ML experiment, use this Skill to lock an offline/online primary metric, set guardrails for latency and fairness, and define the pass/fail threshold so the launch decision cannot be relitigated after results arrive. ## Quick Start Define success criteria and guardrail metrics for my upcoming feature launch, including a baseline, target, and decision rule.
