How to use the Create Data Context agent skill

What an agent skill is and what this one does

An agent skill is a folder containing a SKILL.md file with instructions that a compatible AI agent (such as Claude Code or Codex) reads when a request matches the skill's description. The agent then follows those instructions for that task.

This skill creates, updates, inspects, or repairs Data Analytics semantic layers. A semantic layer here is an explicit artifact you can inspect and cite later. It captures how future analyses should interpret the data, for example which metric definition is canonical, which dashboard or table is the best source of truth, and what caveats should be checked before answering.

• Keeps the generated skill compact and puts detailed metric definitions, tables, query patterns, caveats and evidence into linked reference files.
• Favors durable sources such as transformation code, tests, maintained metric docs and verified dashboards over looser signals like query history or team discussion.
• Keeps source conflicts visible instead of silently choosing one.
• Writes a source inventory listing what was checked, what is missing and which sources are lower-confidence.

Install the create-data-context skill

Download the ZIP for this skill from this page and unzip it. For Claude Code, place the unzipped skill folder in your personal skills directory ~/.claude/skills/, or in a project's .claude/skills/ folder if you want it available only inside that project.

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

• Personal: ~/.claude/skills/ (available across your projects).
• Project: .claude/skills/ inside the project folder (available only in that project).
• Claude apps: Settings → Capabilities → Skills, then upload.

Trigger the skill with natural requests

The skill is selected when your request matches its description. The description covers creating or maintaining a semantic layer and saving data context. Ask in plain language, and say what you want the layer to cover.

Use this skill only for saving data context or creating, updating, inspecting or repairing a semantic layer. For ordinary data work such as answering a metric question, building a report or checking a dataset, the skill routes you to the Data Analytics index skill instead.

• "Save data context for our weekly revenue metrics."
• "Create a semantic layer for the subscriptions reporting area."
• "Inspect the existing churn semantic layer and tell me where its sources conflict."
• "Repair the orders semantic layer; the dashboard link it cites no longer exists."

Where the saved layer goes

The skill chooses one destination based on the runtime. In a local Codex setup it writes to $CODEX_HOME/skills/<area>-semantic-layer, or to ~/.codex/skills/<area>-semantic-layer if $CODEX_HOME is unavailable and the default location is clearly in use. In ChatGPT with a personal Skills library, it installs into that library and returns a link. If no persistent destination is available, it returns a portable skill package and states that it has not been installed persistently.

After creating or updating a layer, the skill offers a weekly refresh automation. It does not create one without your explicit approval unless you asked for recurring refresh.

What to provide so the layer is useful

The skill works from the sources you let it read. Give it access to the places where your metric logic lives, such as repositories, dashboards or docs. Before crawling, it builds a list of what it checked and what is missing. Where a source is thin, it returns a practical plan for what to collect next instead of a full layer.

• Repositories containing transformation code and tests.
• Maintained metric documentation and verified dashboards.
• Table metadata the agent can read.
• Requires Python 3 to run the bundled helper scripts.

Review the output before sharing it and check safety notes

This skill runs bundled scripts (scripts/data_analytics_preflight.py, scripts/validate_data_context_contract.py and scripts/record_plugin_install_suppression.py) and is not network-free in every workflow, so review what it runs in your environment. The skill tries to avoid copying raw sensitive data, credentials, row-level examples or long private messages into generated files, but you should still review outputs before sharing them, because generated files could capture sensitive details.

• Do not use it on sensitive datasets where generated files might capture row-level data or credentials.
• Check every generated definition and caveat against the source it cites.
• Keep the generated files in a location you control if they describe private systems.