agentleFS
TrecekTrecek/useful-claude-skills/.claude/skills/make-experiment-diag/SKILL.md

make-experiment-diag

Skill1 starsChanged 8 months agoMITView on GitHub

Interactive selection of experimental design lens for visualizing experiment methodology. Routes to the appropriate exp-lens-* skill.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the Skill tool to load the selected /exp-lens-{slug} skill.

    line 89

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Experimental Design Diagram Selection

Select the right experimental design lens for your analysis. Each lens asks one primary question about an experiment's design, surfacing assumptions and failure modes specific to that epistemic dimension.

When to Use

  • Want to visualize or audit experimental design methodology
  • Need to select the right lens for a specific experimental concern
  • User says /make-experiment-diag
  • User wants to examine a specific aspect of an experiment

Critical Constraints

NEVER:

  • Skip the selection step — always confirm which lens to use
  • Modify any source code or experimental artifacts
  • Combine multiple lenses in a single invocation

ALWAYS:

  • Present the selection table to help the user choose
  • Match user description to the most relevant lens
  • Load the selected lens skill using the Skill tool

Core Lenses (always recommended)

Your experiment involves...Use this lensQuestion it answersPhilosophical Mode
Unclear or shifting claimsEstimand ClarityWhat exactly is the claim?Evidential
Causal attribution without explicit assumptionsCausal AssumptionsWhat causal assumptions support this design?Causal-Structural
Baseline quality or fairnessComparator ConstructionIs the comparator fair and relevant?Counterfactual
Data preprocessing, splits, or feature pipelinesPipeline IntegrityCould data handling create optimistic bias?Integrity
Nondeterminism, seed sensitivity, run-to-run noiseVariance & StabilityIs the signal larger than the noise?Stability
Asymmetric tuning, compute, or engineering effortFair ComparisonAre alternatives compared symmetrically?Fairness
Reproducing results from artifacts aloneReproducibility & ArtifactsCould an independent party reproduce this?Transparency
Metric choice, proxy validity, score interpretationMeasurement ValidityDo measurements justify the interpretation?Psychometric
Robustness to preprocessing or modeling choicesSensitivity & RobustnessWhich assumptions are load-bearing?Robustness
Generalization claims beyond the test suiteBenchmark RepresentativenessDoes this generalize beyond the test bed?Generalizability

Extended Lenses (enable by experiment class)

Your experiment involves...Use this lensQuestion it answersPhilosophical Mode
Shared resources, network effects, spilloversUnit & InterferenceWhat is the unit, and can treatments spill over?Causal-Structural
Power, multiplicity, sequential monitoringError BudgetAre error risks sized and controlled?Statistical
Theory claims needing adversarial stressSeverity TestingWould this design have caught the error?Falsificationist
Assignment mechanism, blocking, stratificationRandomization & BlockingWhere does comparability come from?Design-Structural
Confounds, history effects, co-interventionsValidity ThreatsWhat alternative explanations survive?Adversarial
Multi-step exploration, adaptive allocationIterative LearningHow does this maximize learning per cost?Decision-Theoretic
Mixing discovery and confirmation in one studyExploratory vs ConfirmatoryIs this discovery or test?Boundary
Deployment risk, fairness, stakeholder harmGovernance & RiskWhat risks arise from acting on this result?Governance

Workflow

Step 1: Prompt User

Ask the user:

What aspect of your experimental design would you like to examine?

Example prompts:

  • "I want to check if my causal claims are well-supported"
  • "I need to verify my data pipeline doesn't leak"
  • "Are my baselines fair comparisons?"
  • "Could someone reproduce my experiment?"
  • "Is my benchmark representative enough?"

Step 2: Match to Lens

Based on the user's description, match to the most appropriate lens using the selection tables above. If ambiguous, present the top 2-3 candidates and ask the user to choose.

Step 3: Load the Selected Lens

Use the Skill tool to load the selected /exp-lens-{slug} skill.

Step 4: Execute the Loaded Skill

The loaded lens skill takes over and runs its full analysis workflow.


Alias Table

AliasLens
estimandexp-lens-estimand-clarity
causalexp-lens-causal-assumptions
comparatorexp-lens-comparator-construction
pipelineexp-lens-pipeline-integrity
varianceexp-lens-variance-stability
fairnessexp-lens-fair-comparison
reproducibilityexp-lens-reproducibility-artifacts
measurementexp-lens-measurement-validity
sensitivityexp-lens-sensitivity-robustness
benchmarkexp-lens-benchmark-representativeness
unitexp-lens-unit-interference
errorexp-lens-error-budget
severityexp-lens-severity-testing
randomizationexp-lens-randomization-blocking
validityexp-lens-validity-threats
iterativeexp-lens-iterative-learning
exploratoryexp-lens-exploratory-confirmatory
governanceexp-lens-governance-risk

Related Skills

  • /mermaid - Shared diagram styling (loaded by individual lenses)
  • /make-arch-diag - Software architecture lens counterpart
  • /verify-diag - Verify diagram accuracy against codebase

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.