agentleFS
45ck45ck/marketing-product-skills/.agents/skills/experiment-design-readout/SKILL.md

experiment-design-readout

Skill3 starsChanged 8 months agoMITView on GitHub

Turn a product or growth idea into a rigorous experiment brief and post-test readout format.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user wants to design, prioritize, or review product/growth experiments.

    line 9
  2. Distinctive:

    Prefer minimum viable tests.

    line 26
  3. Distinctive:

    Use one primary metric.

    line 29
  4. Distinctive:

    Call out when a proposal is not suitable for an A/B test.

    line 32
  5. Distinctive:

    Create the post-test structure before the test starts.

    line 77
  6. Distinctive:

    Use this exact structure after the test finishes:

    line 143
  7. Distinctive:

    If the requested test is pricing-related, state plainly when classic A/B testing is a bad fit due to low sample size, customer fairness …

    line 156
  8. Distinctive:

    If the requested test has major irreversible downside, recommend a smaller-risk pretest or staged rollout.

    line 163

≈ “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

Experiment Design Readout

Use this skill when the user wants to design, prioritize, or review product/growth experiments.

What this skill must produce

Always produce:

  1. Experiment brief
  2. Hypothesis
  3. Primary and secondary metrics
  4. Sample / ramp plan
  5. Implementation notes
  6. Decision rules
  7. Risks
  8. Post-test readout template

Working rules

  • Prefer minimum viable tests.
  • State the causal claim explicitly.
  • Define what success and failure mean before the test launches.
  • Use one primary metric.
  • Guard against underpowered or low-quality tests.
  • Distinguish operational guardrails from outcome metrics.
  • Call out when a proposal is not suitable for an A/B test.

Step-by-step method

Step 1: define the problem

State the specific product or growth problem.

Step 2: define the hypothesis

Use this format:

If [change]
then [primary metric] will move by [expected direction / magnitude]
because [behavioral or product assumption].

Step 3: define the minimum viable test

Specify the smallest intervention that could validate or invalidate the hypothesis.

Step 4: define metrics

Choose:

  • one primary metric
  • 2-4 secondary metrics
  • 2-4 guardrail metrics

Step 5: define the audience and segmentation

Specify:

  • eligible population
  • exclusions
  • key segments to inspect

Step 6: define ramp and duration

Provide:

  • initial ramp
  • escalation path
  • minimum duration
  • sample size considerations
  • stop conditions

Step 7: define interpretation rules

State what outcome would mean:

  • launch
  • iterate
  • kill
  • rerun with higher power / cleaner implementation

Step 8: define readout

Create the post-test structure before the test starts.

Output structure

Experiment brief

  • Problem:
  • User / segment:
  • Proposed change:
  • Why now:

Hypothesis

...

Test design

  • Test type:
  • Control:
  • Treatment:
  • Minimum viable version:
  • Why this is the minimum viable test:

Metrics

Primary metric

  • Name:
  • Why it matters:
  • Expected movement:

Secondary metrics

MetricWhy it matters

Guardrail metrics

MetricWhy it matters

Audience and segmentation

  • Eligible audience:
  • Exclusions:
  • Segments to inspect:

Ramp / duration / sample plan

  • Ramp plan:
  • Target split:
  • Duration:
  • Sample / power caveats:
  • Abort conditions:

Implementation notes

  • Product / eng tasks:
  • Analytics tasks:
  • QA tasks:
  • Dependencies:

Decision rules

  • Launch if:
  • Iterate if:
  • Kill if:
  • Rerun if:

Risks

  • Underpowered test:
  • Confounded implementation:
  • Novelty effects:
  • Long-term effects not captured:
  • Maintenance cost vs gain:

Post-test readout template

Use this exact structure after the test finishes:

  1. Experiment summary
  2. What happened on the primary metric
  3. What happened on secondary and guardrail metrics
  4. Segment-level observations
  5. Did the result support the hypothesis?
  6. Most likely explanation
  7. Decision
  8. Follow-up actions
  9. New questions created

Special instructions

If the requested test is pricing-related, state plainly when classic A/B testing is a bad fit due to low sample size, customer fairness concerns, or poor statistical power. Offer an alternative such as:

  • pricing research
  • segmented rollout
  • minimum viable pricing test
  • sales-assisted validation
  • willingness-to-pay study

If the requested test has major irreversible downside, recommend a smaller-risk pretest or staged rollout.

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.