Audit any eval, metric, experiment, or benchmark for leakage — does external ground-truth enter independently, or are the model, scorer, and designer just confirming a result no outside truth ever produced? Use before trusting any 'how we'll know it worked' — an A/B, a holdout, a score, a validation — and whenever a result feels too clean or self-confirming.
Install
The individual skill folder is not confirmed. This command opens the repository’s skill selector; follow the upstream instructions.
skills CLI
npx skills add LilMGenius/paperthin