Guide · Agent skill
How to use the Experiment Readout agent skill for A/B tests
Updated · written with AI from the project's own documentation and checked against our catalogue data
Experiment Readout is an agent skill that takes raw experiment notes, such as A/B test results, growth experiments, onboarding changes or email tests, and turns them into a structured decision report. This guide explains what an agent skill is, how to install this one, how to ask for it in plain language, and where its limits are. By the end you will know how to get a report that recommends shipping, stopping, extending or iterating on a change.
Experiment Readoutnexu-io/html-anything · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Get the Experiment Readout ZIP from its page on this site.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder for one project.
- 3Or upload it in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Paste your experiment notes. Provide the hypothesis, variants, primary and guardrail metrics, and results as markdown, CSV, a table or a mix.
- 5Ask for a decision report. Request a readout or a ship, stop, extend or iterate recommendation, and choose a style if you have a preference.
What the Experiment Readout skill makes the agent do
An agent skill is a folder of instructions that an AI agent loads when a task matches it. With this skill installed, the agent does not return a plain data summary. It writes a decision report about whether a change should go live.
The report covers the hypothesis, the setup, primary and guardrail metrics, an interpretation that separates signal from noise, a clear recommendation to ship, stop, extend or iterate, and follow-up experiments. It also flags uncertainty, such as small samples or missing significance data.
- Header with experiment name, owner, date, status and decision
- Hypothesis rewritten as a testable statement
- Setup: audience, variants, duration, sample size, primary metric and guardrail metrics
- Result snapshot with primary metric lift, absolute change and caveats
- Control vs variant metric table
- Interpretation, Decision, follow-up experiments and instrumentation notes
Install the Experiment Readout skill
Download the ZIP file for the skill from its page on this site. The ZIP contains the skill folder, including SKILL.md, the three HTML style templates in assets/, and example files.
You can install it in one of two ways. For Claude Code, unzip it into ~/.claude/skills/ to make it available everywhere, or into a project's .claude/skills/ folder to limit it to that project. In the Claude apps, upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the skill
How to trigger the skill with natural requests
You do not need a special command. Describe the experiment and ask for a readout, decision or recommendation in ordinary language. The skill is meant for requests about A/B tests, growth experiments, pricing tests, onboarding changes, feature rollouts and email tests.
Paste your notes as markdown, CSV, a table, or a mix of these. Include the hypothesis, the variants, the primary metric, any guardrail metrics and the results for each group. The more complete the input, the more specific the report.
- Review this A/B test and tell me whether we should ship the new onboarding checklist.
- Turn these experiment notes into a decision report: ship, stop, extend or iterate.
- Write a readout for this email subject line test with the caveats kept visible.
- What follow-up experiments should we run after this pricing test?
Choose one of the three visual styles
The skill can produce an HTML report in one of three styles. Pick one that fits the context and do not mix them. If you do not specify a style, the default is product-readout.
product-readout is a light product readout for PMs, growth teams and leadership. lab-notebook suits early-stage experiments, mixed qualitative and quantitative work, and exploratory tests where caveats must stay visible. growth-console is a dark analytics layout for growth teams focused on live metrics, funnels, activation and conversion.
- product-readout: default, for PM, growth and leadership readouts
- lab-notebook: for exploratory or mixed qualitative and quantitative experiments
- growth-console: for funnel, activation and conversion monitoring
What the skill needs and what it does not do
The skill lists no required tools or setup. It does not run scripts and does not use the network, so it works only with the data you give it. It is designed to rely only on user-supplied numbers and does not invent sample sizes, p-values or confidence intervals.
It also does not perform statistical validation. It does not calculate p-values or confidence intervals. If you need a formal analysis, use a statistician or an analytics tool, and give the report's numbers to them for checking.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
- Not for formal significance testing or confidence intervals
- Not for experiments without real data
Tips for better readouts
Give the agent complete inputs. Missing sample sizes, unclear metric definitions or absent guardrail results lead to more caveats in the report, which is intended behaviour.
When the report says a result is directional, inconclusive or needs more data, treat that as the honest answer. Small samples and missing significance information are flagged on purpose. Use the instrumentation notes section to find tracking gaps before the next test.
- Define the primary metric precisely, including its time window
- Include guardrail metrics such as support tickets or deletions
- Provide sample sizes for every variant
- Use the follow-up experiments section to plan the next test
What people use it for
- A product manager reviews an onboarding A/B test and decides whether to launch the change.
- A growth team writes a readout of an email subject line test and lists follow-up experiments.
- A leader gets a concise, action-oriented summary of a pricing or feature rollout test.
- An early-stage team records a mixed qualitative and quantitative experiment and keeps its caveats visible.
Frequently asked questions
Is the Experiment Readout skill free to use?
The source lists the licence as Apache-2.0, which is the licence under which the skill is distributed. The source does not list prices, so check the licence text for the terms that apply to your use.
What do I need to install the Experiment Readout skill?
The skill lists no required tools. You download the ZIP, unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
Does the skill calculate p-values or confidence intervals?
No. The skill does not calculate p-values or confidence intervals and is instructed not to invent them. For formal statistical validation, use a statistician or an analytics tool.
Does Experiment Readout run scripts or use the internet?
According to the source, the skill runs no scripts and uses no network. It works only with the data you provide.
What if my experiment has a small sample or no significance data?
The skill flags uncertainty in the report and uses terms such as directional, inconclusive or needs more data. It is designed to keep these caveats visible rather than present results as certain.
How is this different from a normal data summary or dashboard?
The skill is aimed at a decision: it asks whether to ship, stop, extend or iterate, and lists follow-up experiments. The source describes it as not a standard data report or dashboard. The source does not compare it with other specific tools.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Experiment Readout suits product managers, growth teams and leaders who want a clear recommendation from an A/B test and caveats kept visible. It is not a replacement for formal statistical analysis, so teams that need p-values or confidence intervals should pair it with a statistician or an analytics tool.
Download the skill