builder-product

Use before running any experiment that will inform a product decision. All six elements — hypothesis, metric, sample size, duration, stopping rule, and decision rule — must be defined before the test starts. Blocks "we'll stop when we see something" completions.

Use before any AI feature is approved for build. Covers four AI-specific risks that standard PRD review misses: hallucination UX, trust calibration, scope creep, and reversibility. Blocks "we validated it like any other feature" completions.

Use before any engineering estimate is made. Requires a written scope before an estimate, and a written estimate before a commit. Blocks "we'll scope as we go" and "just give me a rough number" completions.

Use before writing any implementation code for a new feature. North star metric, guardrail metrics, and diagnostic metrics must be defined — with baselines — before the build begins. Blocks "we'll figure out the metrics from the data" completions.

Use before committing scope, estimate, or engineering time to any feature. Requires user problem, success metric, scope boundary, and anti-goals defined before any estimate is made. Blocks "we'll define success once we see the data" completions.