Guide · Agent skill
How to use the Statistical Problem Formulation agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Statistical Problem Formulation skill is a planning checklist for an AI agent. It makes the agent turn a broad research topic into a precise statistical problem before any method design, theory or report writing begins. It is aimed at researchers, graduate students and teams running AI research pipelines who need a consistent problem specification. After this guide you will know what the skill does, how to install it, how to trigger it, and what to check in its output.
Statistical Problem Formulationaiming-lab/AutoResearchClaw · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from this page.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4State your research topic. Describe the topic and the data you have, using words like estimand, assumptions or hypothesis.
- 5Review the formulation. Check the target, assumptions, claims and evaluation criteria against your actual study.
- 6Use the handoff or template. Pass the YAML handoff to later stages or write up the Markdown template.
What the skill makes the agent do
The skill tells the agent to work through a fixed set of elements before doing anything else in a study. It covers the observed data and how it was sampled (iid, dependent, clustered, time series, censored or selected), the data-generating model, the target parameter, decision, prediction or risk, the assumptions needed for identifiability or for the method to work, the claims being supported or refuted, the success metrics, and the theoretical property to derive, such as bias, variance, consistency, rate, coverage, error bound, robustness or impossibility.
- Observed data and sample size
- Data model: the family of distributions or processes considered
- Target: estimand, decision, prediction or risk
- Assumptions, hypotheses and evaluation criteria
- Theory target to derive
Install the skill from the ZIP download
Download the ZIP for this skill from this page. Unzip it into your personal skills folder ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill available only in that project. The package contains a single file, SKILL.md, which holds the instructions the agent reads.
You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/
- Claude apps: Settings → Capabilities → Skills, then upload
Trigger the skill with natural requests
You do not need a special command. The skill's description covers formulating statistical research problems, and its trigger keywords include problem formulation, statistical formulation, estimand, assumptions, data model, hypothesis and theory target. Ask in plain terms that use one of these words, and state the topic you are studying.
Ask for the formulation before the method or report, since the skill is meant to run first.
- "Formulate the statistical problem for my study on whether a new tutoring program raises test scores."
- "Before we design methods, list the estimand, assumptions and hypotheses for this survey analysis."
- "Write a problem formulation with a theory target for this clustered-data question."
What the output looks like
The skill provides a YAML handoff structure with fields such as observed_data, data_model, target, assumptions (structural, sampling, regularity, identifiability), claims, evaluation_criteria, theory_targets and blocking_ambiguities. It also provides a Markdown template with sections for the research question, observed data, data-generating model, target or estimand, candidate procedure class, assumptions, claims, evaluation criteria, and theoretical and experimental questions.
The quality bar is that another researcher could implement or analyze the problem without guessing the target, assumptions or success criteria. If the output leaves those vague, the formulation has not passed.
- Use the YAML block when a later pipeline stage needs structured input
- Use the Markdown template when writing up for people
topic_id: TXX
title: ""
research_question: ""Requirements, scripts and network use
The listing states that the skill requires nothing beyond the agent it runs in. It does not run scripts and does not use the network. The package holds only SKILL.md, so there is no code to review beyond the instructions themselves.
Because the skill is a planning template, its output is only as good as your inputs. Check that the assumptions and target it proposes match what your data actually allows.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
Tips for better formulations
Give the agent real details: the sample size, how the data were collected, and the question you actually want answered. Vague topics produce vague targets. Review the assumptions list closely, because surfacing hidden assumptions is the main value of the checklist.
Run the skill before method design so the later stages inherit a fixed specification rather than a loose topic.
- Name the data source and sampling scheme up front
- Ask which assumptions are doubtful, not only which are listed
- Keep blocking_ambiguities open until you resolve them
What people use it for
- Framing a clustered-data study before choosing an estimator
- Listing the assumptions a causal or observational analysis depends on
- Writing a theory target such as consistency or coverage before running experiments
- Producing a structured problem specification for an AI research pipeline's later stages
- Auditing an existing study design for hidden assumptions
Frequently asked questions
How do I install the statistical-problem-formulation skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.
Does the statistical problem formulation skill require anything else?
The listing states that it requires nothing and does not run scripts or use the network. The package contains only SKILL.md.
Is the statistical-problem-formulation skill free?
The source material does not state a price. It is published under the MIT licence, as listed on this page.
What topics is this skill not suited for?
It is not meant for quick data cleaning or descriptive charts, or for non-statistical topics. The skill adds planning overhead and its vocabulary of estimands and identifiability may mislead outside statistics.
How do I trigger the skill?
Use plain requests with trigger words such as problem formulation, estimand, assumptions, data model, hypothesis or theory target, and name your topic. Ask for the formulation before method design or writing.
What is the difference between this skill and a data analysis tool?
This skill produces a planning specification: the data, model, target, assumptions and criteria. It does not analyse data or run scripts.
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
Use this skill if you are a researcher or graduate student who needs a rigorous statistical framing before experiments, or a team whose AI pipeline needs a consistent problem specification. Pick something else for quick cleaning, descriptive charts or non-statistical topics, where its checklist adds overhead without benefit.
Download the skill