experimental-design

Design experiments and studies BEFORE data is collected — choosing a design, randomizing, blocking, and laying out treatment combinations so results are interpretable. Use whenever someone is planning a study, asks how to assign subjects/samples to groups, mentions randomization, blocking, stratification, controls, factorial or fractional-factorial designs, design of experiments (DOE), screening many factors, response-surface optimization, crossover or repeated-measures or split-plot designs, cluster/group randomization, Latin squares, plate layouts, batch/run-order effects, replication vs. pseudoreplication, or sequential/adaptive/group-sequential designs. Trigger even for informal phrasings like "how should I set up this experiment", "how do I avoid confounding", "what's the best way to test these 6 factors", or "assign these mice to conditions". For computing the sample size or power once the design is chosen, use statistical-power; for analyzing data already collected, use statistical-analysis.

Quality

93%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

86%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable with executable, seeded code examples and a clean one-level-deep reference structure, and it is largely concise. The main gap is a missing explicit validation/sanity-check step inside the batch workflow, which keeps workflow clarity at 4 rather than 5.

Suggestions

Add an explicit validation checkpoint to the Workflow (e.g., between steps 5 and 6): run arm_balance(sched) / inspect the generated layout and only proceed once balance and run-order randomization are confirmed — turning the existing arm_balance snippet into a stated workflow step.

Tighten the "mistakes that ruin studies" section: lead each item with the structural rule and keep the one-line consequence, trimming the explanatory prose to reduce tokens without losing the design-time guidance.

Consider a short "verify before running" feedback loop note (regenerate the schedule from the archived seed and diff, to confirm reproducibility) given the skill's emphasis on auditable, pre-registered layouts.

Dimension	Reasoning	Score
Conciseness	Mostly lean and information-dense assuming Claude's competence, but the 8-item "mistakes that ruin studies" list and Fisher's-principles bullets carry some explanatory prose that could be tightened; sits above 3 (efficient with minor trim opportunities) but below 5 (not every token is maximally lean).	4 / 5
Actionability	Two complete, copy-paste-ready script examples with real factor dictionaries, seeded calls, and CSV export cover the common cases (block/stratified/cluster randomization; factorial, Plackett-Burman, CCD), plus a concrete decision tree; matches the fully-executable anchor.	5 / 5
Workflow Clarity	The 8-step workflow is clearly sequenced with cross-skill handoffs (step 4 to statistical-power, step 8 to statistical-analysis), but for batch allocation/layout operations it lacks an explicit in-workflow validation checkpoint (the arm_balance sanity check appears only in a code snippet, not as a workflow step); below 5's explicit-validation/feedback-loop anchor, above 3's implicit-checkpoint level.	4 / 5
Progressive Disclosure	Clear overview with well-signaled one-level-deep references — all four reference files and two scripts exist as real bundle files, no nested references — and content appropriately split (decision tree and script usage inline, deep concepts externalized); matches the clear-overview, one-level-deep anchor.	5 / 5
	Total	18 / 20 Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is exemplary: concrete actions, exhaustive trigger terms covering both technical and informal phrasings, explicit what/when guidance, and clear boundary routing to sibling skills to avoid conflict. No meaningful improvements needed.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — "choosing a design, randomizing, blocking, and laying out treatment combinations" — with comprehensive coverage of the design-time task surface; not the level below (4) because no meaningful action gaps remain.	5 / 5
Completeness	Explicitly answers both what ("Design experiments and studies BEFORE data is collected") and when ("Use whenever someone is planning a study...") with concrete trigger phrases, plus negative boundaries routing sample-size to statistical-power and analysis to statistical-analysis.	5 / 5
Trigger Term Quality	Comprehensive coverage spanning technical terms (randomization, blocking, factorial, DOE, response-surface, Latin squares, pseudoreplication) and informal phrasings users actually say ("how should I set up this experiment", "assign these mice to conditions"); matches the anchor's synonym-and-variation breadth.	5 / 5
Distinctiveness Conflict Risk	Clear niche (pre-collection design) with distinct triggers and explicit boundary guidance to sibling skills, minimizing overlap; matches the clear-niche, minimal-conflict anchor.	5 / 5
	Total	20 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: K-Dense-AI/scientific-agent-skills
Path: skills/experimental-design/SKILL.md
Commit: b2a92ba

Reviewed: 3 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.