← ClaudeAtlas

experiment-card-generatorlisted

Turn claims, assumptions, draft excerpts, or roadmap items into Strategyzer Test Cards and Learning Cards: extract every risky assumption, rank by evidence risk and test cost, write each card in the exact we-believe / to-verify / confident-if format with a numeric threshold, apply the cheapest-test-first rule so waitlists, search-query checks, and fake-door links always precede builds and paid campaigns, and output a paste-ready experiment backlog grouped by evidence strength. Use when the user says test cards, experiment cards, turn claims into experiments, how would I test this assumption, or design experiments for this draft. Not for running the experiments, analytics instrumentation, or prioritizing already-validated work.
karozi/awesome-product-management-skills · ★ 2 · AI & Automation · score 56
Install: claude install-skill karozi/awesome-product-management-skills
# Experiment Card Generator Turn claims, assumptions, or draft excerpts into Strategyzer testing artifacts: Test Cards, Learning Cards, and a prioritized experiment backlog. ## When to Use This Skill Use when the user provides a draft, claim list, product idea, roadmap item, or claim ledger from a fact-check, and wants experiments, test cards, or a testing plan. Trigger phrases: "test cards", "experiment cards", "turn claims into experiments", "how would I test this assumption", "design experiments for this draft". ## Instructions 1. **Extract assumptions.** Read the input and list every risky assumption it depends on: what must be true for the idea, claim, or roadmap item to work. If the input is a claim ledger from a fact-checking skill, carry its claims over directly. Label each assumption as desirability (do they want it), feasibility (can we build it), or viability (will it pay off). - Done when: every load-bearing claim in the input maps to at least one assumption, and none restates a fact already evidenced. 2. **Rank by evidence risk x test cost.** Score each assumption 1 to 5 on evidence risk (how little real evidence supports it) and 1 to 5 on test cost (time and money to get a signal). Sort descending by risk, then ascending by cost. Output the ranking as a table with both scores. 3. **Read `references/test-card-templates.md`.** Follow it to generate one Test Card per ranked assumption in the exact format: "We believe that [hypothesis]. To verify that, we