build-measure-learnlisted

# Build-Measure-Learn The loop is the unit of startup work. Not the sprint, not the release, not the quarter: the loop — hypothesis, smallest experiment, measurement, learning, repeat. This skill plans, tracks, and retro-dicts loops. Its entire job is making sure no loop starts with "build." ## The Loop The loop is drawn learn-first: decide what you need to learn, design the smallest experiment that teaches it, define how the result will be measured, build the minimum to run the experiment, measure against the pre-declared threshold, write down what is now believed and why, and let the learning pick the next loop. A loop that starts with build produces a product in search of a question. That is the failure mode this skill exists to prevent. ## Modes ### Plan mode (default — user brings an idea, feature, or initiative) 1. Force the hypothesis first: "We believe [action] will cause [measurable outcome] because [mechanism]." If the user cannot state one, help extract it — and if none exists, say so: the initiative is a bet, not an experiment, and gets flagged. 2. Identify what must be learned before anything is built, and pick the smallest experiment that produces it. Often the answer is a smoke test (smoke-test-launcher) or an MVP (mvp-type-selector), not a build. 3. Define the measure BEFORE the build: metric, baseline, threshold, date. "We'll figure out how to measure it later" is blocked — later never comes. 4. Copy `templates/loop-card.md`, fill every