Trustworthy Experiments

pmprompt/claude-plugin-product-management

Design rigorous A/B tests and holdouts with clear hypotheses, sample sizing, guardrails, and analysis plans before shipping pricing, onboarding, or feature experiments.

Related skills

Paper Context ResolverResolve narrow paper-backed gaps (splits, preprocessing, eval protocol, checkpoints, runtime) when README and repo files are insufficient for faithful ML paper reproduction.140k412Repo Intake And PlanScan an ML research repository, extract README commands, and recommend the smallest trustworthy inference or evaluation reproduction target.140k412Minimal Run And AuditRun documented inference, evaluation, or smoke commands and normalize evidence into standardized repro_outputs with patch notes.140k412Env And Assets BootstrapBootstrap conda-first environments plus checkpoints, datasets, and cache paths before a documented reproduction run.140k412Analyze ProjectMap a deep learning repo read-only—training, inference, and eval entrypoints, configs, and suspicious patterns—before you let an agent patch model code.52.4k412Ai Research ReproductionOrchestrate README-first, minimal-trustworthy reproduction of a deep learning repository with auditable repro_outputs evidence.52.3k412