experiment-design

Experiment Design Skill

Purpose

Break an approved research design into a fully executable experiment plan. Every step must be runnable by someone with zero prior context on the project. No ambiguity, no placeholders, no "check results" without specifying exactly what metric and what threshold.

This is the research equivalent of writing-plans: just as that skill breaks a feature into bite-sized TDD tasks with exact file paths and code, this skill breaks a research design into discrete experiment steps with exact data paths, commands, seeds, and success criteria.

Announce at Start

At the beginning of every session using this skill, state:

"I'm using the experiment-design skill to create the experiment plan."

Trigger Conditions

Use this skill when:

• A research design or hypothesis document has been approved (exists in docs/eureka/designs/)
• Someone asks to "plan the experiments," "operationalize the design," or "turn this into runnable steps"
• A hypothesis-first output needs to be translated into concrete execution

Do NOT use this skill to:

• Generate hypotheses (use hypothesis-first)
• Review completed experiments (use requesting-research-review)
• Produce analysis or interpret results

---

Plan Document Header

Every experiment plan MUST begin with this header, fully filled in — no blanks, no TBDs:

# [Study Name] Experiment Plan
**Goal:** [one sentence describing what this plan will determine]
**Design Reference:** [relative path or URL to the approved design doc]
**Registered Hypothesis:** [exact H1 statement, copied verbatim from the design]
**Primary Outcome:** [exact measure, units, and dataset]
**Analysis Plan:** [exact statistical test, correction method, significance threshold]
---

If any of these fields cannot be filled in, stop and ask the user to provide the missing information before continuing. An incomplete header is a hard blocker.

---

Scope Check (Do This First)

Before writing any tasks, examine the approved design for scope: