pre-registering-eval-study

What problem does it solve?

This Skill prevents HARKing, p-hacking, optional stopping, and outcome switching by forcing a study design to be specified before any data is observed.

Core Features & Use Cases

• Hypothesis Locking: Turns vague ideas into a directional, falsifiable primary hypothesis with an operationalized outcome.
• Analysis Planning: Fixes the statistical test, minimum effect of interest, power justification, stopping rule, and falsification criterion in advance.
• Confirmatory Study Workflows: Supports LLM evals, jailbreak studies, A/B tests, clinical-style studies, observational analyses, and model comparisons that need defensible pre-registration.

Quick Start

Ask the skill to pre-register your study before any data is collected, specifying the hypothesis, primary outcome, effect threshold, stopping rule, and commit location.