brainstorm-experiments-existing

brainstorm-experiments-existing helps a product team design low-effort experiments to validate assumptions for an existing product before committing to full implementation. It is used when validating assumptions, testing feature ideas cheaply, or planning product experiments, with the aim of maximum validated learning at minimal effort. The user describes their feature idea and the assumptions that need validation, and if they provide files such as PRDs, assumption lists, or designs, the skill reads them first.

The skill works through a clear sequence. It first clarifies the idea and the assumptions to confirm what the team wants to build and validate. It then suggests experiments for each assumption, drawing on methods including first-click testing or task completion with a prototype, feature stubs or fake door tests, technical spikes, A/B tests on production with risk mitigation, Wizard of Oz approaches, and behavioral (not opinion-based) survey validation. It follows key principles: measure actual behavior rather than opinions, test responsibly without putting users or the business at risk, explain risk mitigation for production tests, and aim for maximum validated learning with minimal effort. For each experiment it specifies the Assumption (what is believed), the Experiment (exactly what will be done), the Metric (what will be measured), and the Success threshold (the expected value if the belief is right), presenting the experiments in a clear table or structured format and saving as markdown if substantial.

The skill targets product managers and discovery teams who already have a feature idea and a set of assumptions and want cheap, responsible validation methods. Documented in a single SKILL.md, it takes the product context via $ARGUMENTS and points to further reading including an experiments library, an assumption prioritization canvas, and product discovery guides. It emphasizes behavior over opinion and responsible testing as its guardrails rather than depending on any specific tooling.