How to use the Market Sizing agent skill: TAM, SAM, SOM

What is an agent skill, and what does market-sizing make the agent do?

An agent skill is a folder of instructions, usually centred on a SKILL.md file, that an AI agent loads when a task matches the skill's description. This one, market-sizing, is described as a way to estimate market, segment or opportunity size with transparent assumptions and uncertainty.

When installed, the agent approaches market-size questions in a structured way. It defines the market boundary, picks a sizing method, gathers evidence for each input and separates sourced facts from assumptions. It shows its calculation chain, tests which assumptions move the answer most and reports a range along with the next checks that would improve confidence.

• Defines what is being sized, where and when it applies, who counts as part of the market and the unit of measurement.
• Chooses a top-down, bottom-up or value-based approach, and explains which one it trusts most and why.
• Labels each major input as structured data, company context, public source, user input or proxy assumption.
• Runs sensitivity checks on the assumptions that move the estimate most.
• Hands the result to other skills for charts, notebooks, spreadsheets or a written report when relevant.

Install the market-sizing skill from the ZIP

Download the ZIP for the market-sizing skill from this directory page. The skill is MIT licensed. Unzip it into one of these locations, depending on whether you want it for all projects or one project only: ~/.claude/skills/ for your user account, or .claude/skills/ inside a project folder.

If you use the Claude apps instead of the file system, upload the skill under Settings → Capabilities → Skills. The package contains agents/openai.yaml and SKILL.md. The instructions live in SKILL.md.

• 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 do you trigger the skill with a natural request?

You do not need a special command. Ask for a market or opportunity estimate in plain language, using the kinds of terms the skill's description names: TAM, SAM, SOM, sizing scenarios, or comparing the scale of possible opportunities. The agent should load the skill when your request matches.

Say what you want measured and why. The more concrete the question, the better the boundary can be defined. If you want a specific output, name it, for example a workbook or a chart.

• How big is the TAM, SAM and SOM for a B2B invoicing tool in my region?
• Compare the size of the pet food market in two countries, using a bottom-up model.
• Size the opportunity for an on-site EV charger service across our three cities.
• Which of these two adjacent segments is larger in revenue terms? Show your assumptions.

What does the skill need before it can work well?

The skill itself needs no fixed data source. Its inputs are optional, but they improve the result. Connected structured data sources, such as a data warehouse or BI tools, let the estimate use your own numbers. Context sources such as company documents help define the business meaning of those numbers and the assumptions behind them.

Some companion skills are optional: visualize-data for charts, build-report for a written report, gather-business-context for context sources, jupyter-notebooks for code-based models and Spreadsheets for workbooks. Code-based models need a notebook environment that can run Python or a similar language. The skill uses the network, so it can reach public sources for benchmarks and comparable markets.

• Optional: connected data warehouse or BI sources
• Optional: company docs and other context sources
• Optional companion skills: visualize-data, build-report, gather-business-context, jupyter-notebooks, Spreadsheets
• Optional for code models: a notebook environment that runs Python or similar
• Runs scripts: no. Uses network: yes.

Safety notes: network access and required sources

The skill uses the network, so it may query public sources and any sources you have connected. Review which sources are connected before you start, especially if they contain confidential data.

The skill is built to stop rather than substitute weaker data. If a source it needs is unavailable, it will tell you what is missing and ask you to provide it or approve a fallback. This is deliberate, but it means the agent may pause often.

• Check which data sources and context sources are connected before asking for an estimate.
• Expect pauses when a required source is missing; supply the source or approve a labelled fallback.
• Treat the output as an estimate that depends on the quality of inputs and assumptions, not verified market data.

Tips for a defensible estimate

Give the boundary first. Tell the agent the geography, time horizon, customer type and unit of measurement. Ask it to show the calculation chain and sensitivity table, and ask which assumption you should validate first. Note that the skill is designed to hand the final estimate to build-report unless you explicitly decline a report, so say so if you want only a brief chat answer.

• Start with a clear market boundary and unit of measurement.
• Ask for a range and the assumptions that move it most.
• Use the validation priorities the skill lists to decide what to check next.
• If you only need a quick guess, say so; a full model is overkill for that.