← ClaudeAtlas

venture-experimentlisted

Designs a concrete low-cost market experiment for the current blocking assumption in a Venture OS venture. Use before substantial product implementation or when a gate returns TEST.
nacorga/venture-os · ★ 0 · AI & Automation · score 75
Install: claude install-skill nacorga/venture-os
# Venture Experiment Target venture directory: `$ARGUMENTS` Read the target venture's latest gate decision and `framework/experiment-principles.md`. Invoke `experiment-designer`. Create a new experiment under `experiments/<id>-<slug>/experiment.yaml` using `templates/experiment.yaml`, plus any small supporting brief needed. The experiment must define: - one primary assumption; - target participant; - procedure; - assets; - success/failure/ambiguous signals; - time and cost cap; - downstream decision rules, one per result branch. Write `decision_rules.on_success` and `decision_rules.on_failure` as `{ outcome, instruction }`, where `outcome` is the gate outcome that branch leads to — `PROCEED`, `TEST` or `PARK` — and `instruction` says what happens next. `on_ambiguous` takes the same form; its `outcome` may stay `null` when an ambiguous result routes to more than one outcome, and then its `instruction` must say how the outcome is chosen. The lock refuses success and failure rules that do not name their outcome, because a later decision can only be checked against a preregistration that says where each result leads. Prefer behavioral evidence over stated preference. Do not build production software if a smaller test can resolve the same uncertainty. Before execution, run `npm run experiment:lock -- <path-to-experiment.yaml>` while the experiment is still `status: designed`. Locking is the readiness gate: it must fail if the procedure, observable signals, decision rules,