How to use the Data Analytics Router agent skill

What the Data Analytics Router skill does

The skill is a router. When a request involves structured records, metrics or an evidence-based business decision, it points the agent to a specialised workflow file to read. Those workflows cover metric changes, product and business decisions, data quality checks, KPI frameworks, KPI reporting, dashboards, reports, charts, notebooks, market sizing, validation and semantic-layer or data-context work.

The skill asks for business context and source quality checks before running the analysis. It then produces a Markdown answer or the artifact you requested. It tells the agent not to invent data and to use sample data only when it is clearly labelled as synthetic.

• Metric changed or missed a target: diagnostics workflow.
• Launch or experiment recommendation: product and business analysis workflow.
• Dashboard or chart from data you already have: dashboard and visualization workflows.
• Reproducible SQL or Python analysis: notebook workflow.
• TAM, SAM, SOM or segment sizing: market sizing workflow.

Install the skill from the ZIP download

Download the ZIP for this skill from this page and unzip it. The skill folder contains agents/openai.yaml and SKILL.md. Put the unzipped folder in one of two places, depending on how you want to use it.

For a personal install available across your projects, place it in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

Check the bundle is complete before you start

The router points to other SKILL.md files, such as metric-diagnostics, build-dashboard and validate-data. Those files must be present in the bundle. If they are missing, the router cannot load the workflow it needs for a given request.

The skill also refers to assets/demo-product-growth.csv as a synthetic demonstration file. Use it only to test the workflow, never to answer a question about real business data.

• Confirm the sibling workflow folders sit next to this skill.
• Do not expect the router to work alone for specialised tasks.

Trigger the skill with natural requests

You can invoke the skill explicitly when Data Analytics is named, or simply ask for something that needs data interpretation. The skill applies when the request involves structured records, quantitative evidence, metrics or an evidence-backed decision.

Give the agent the data in a form it can read, such as a CSV or XLSX file, a query result, a schema, a dashboard export, a screenshot or a pasted table. Name the metric and time window if you know them.

• "Why did weekly active users drop last month? Use users.csv."
• "Design a KPI set for our onboarding funnel, with a guardrail metric."
• "Build a dashboard from sales_export.xlsx for the regional team."
• "Size the SAM for our invoicing tool using the attached segment table."
• "Review this analysis for join errors before we share it."

What the skill needs and what it does not do

The skill needs access to the data the agent should analyse. That can be local files, pasted tables, or connectors that your host agent provides. The skill itself does not decide which tools exist; the host does.

According to the skill's metadata, it does not run scripts and does not use the network. It works from the files and data the agent can already reach.

• Requires: access to your data through local files, pasted tables or host connectors.
• Requires: the referenced sub-workflow SKILL.md files in the bundle.
• Runs scripts: no.
• Uses network: no.

Safety notes and sensible limits

The skill tells the agent to avoid exposing credentials, secrets or unnecessary personal data. It also says not to invent records, measurements, source access or query results. If the available sources conflict in a way that changes the answer, the agent should surface the conflict rather than pick silently.

Because the skill does not run scripts or use the network, it does not send your data elsewhere on its own. Any data access comes from the tools your host agent has already been given, so review those permissions in your host.

• Expect the agent to ask for an export rather than guess when no usable source exists.
• Check that any synthetic output is labelled as such before sharing it.
• Review the host agent's connector permissions for databases and BI tools.

Tips for better results

Give business context up front: the decision you need to make, the metric owner and any recent changes. The skill's first step is to gather missing definitions, so providing them saves a round trip.

Ask for the artifact you want. The skill produces Markdown answers by default, and uses HTML, PDF, DOCX or notebooks only when you request them and the host can create them.

• Name the grain, time window, filters and units of your data.
• Ask for a Markdown answer first, then request a dashboard or report if you need one.
• Ask for the conflicts or caveats section if sources disagree.