Guide · Agent skill
How to use the Data Lineage Summary agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Data Lineage Summary skill (datalineage-summary) is a set of instructions that lets an AI agent investigate where a specific Google Cloud table or column gets its data and where that data goes next. It then writes a plain-English Markdown report that walks through the flow from left to right. This guide is for data engineers, analysts and documentation owners who want that report without reading raw lineage graphs. By the end you will know how to install the skill, what it needs, how to ask for a summary and how to read the result.
Data Lineage Summarygoogle/skills · Apache-2.0 Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from this page and unzip it so you have a folder containing SKILL.md.
- 2Place the folder. Move it to ~/.claude/skills/ for personal use, or to .claude/skills/ in your project. Alternatively, upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Connect the MCP server. Set up the Data Lineage (Knowledge Catalog) MCP server using Google's guide and add its configuration to your agent's MCP config file.
- 4Confirm permissions. Check that you can read data lineage in the Google Cloud projects that hold your asset.
- 5Ask for a summary. Name the asset with its fully qualified name, such as bigquery:project.dataset.table, and ask for its upstream and downstream lineage.
- 6Review the metadata. Read the Analysis Metadata section to see which locations and limits were used.
- 7Rerun if needed. If coverage looks incomplete, ask the agent to rerun with more locations or a greater depth.
What an agent skill is and what this one does
An agent skill is a folder containing a SKILL.md file and supporting reference files. When a request matches the skill's description, the agent reads these instructions and follows them. This skill covers BigQuery and Cloud Storage assets on Google Cloud.
Once installed, the agent can investigate a specific table or column and trace its upstream sources and downstream consumers using Google's data lineage service. It then writes a report covering the systems involved, the data flow and the search limits it used.
- Table-level lineage: the agent looks at the whole table as the focal asset.
- Column-level lineage: the agent limits the analysis to the fields you name, and states that scope at the top of the report.
- Output: a Markdown report with an overall flow type, a systems overview, upstream and downstream sections, and analysis metadata.
Install the skill from the ZIP download
Download the ZIP for the skill from this page and unzip it. The folder you get contains SKILL.md, references/attribution.md and references/mcp-usage.md. Put the unzipped skill folder in one of the locations below.
For a personal install available across your projects, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder so it applies only there.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills → upload
Set up the Data Lineage MCP server
The skill cannot work without the Data Lineage (Knowledge Catalog) MCP server. The skill's own instructions say it relies on that server for graph traversal and that you must be able to run search_lineage queries in both directions.
Follow Google's guide to connect the server, then add the client configuration block to your agent's MCP configuration file, such as mcp_config.json. The skill's reference page on MCP usage contains the tool schema and connection details.
You also need access to the gcloud CLI and/or REST API, and bq or gcloud storage so the agent can locate the asset's region. You need permission to read data lineage in the relevant Google Cloud projects.
- A Google Cloud project with the Data Lineage MCP server available
- Permission to read data lineage in those projects
- gcloud CLI and/or REST API access
- bq or gcloud storage for finding the asset's region
Trigger the skill with a natural request
You do not need a special command. Describe the asset and what you want to know, and include the fully qualified name of the table or column where you have it. The skill's description covers summarising upstream and downstream data flows, debugging data quality issues and understanding provenance.
Requests that will not trigger it include generic BigQuery queries, edits to lineage relationships, and downstream impact analysis for planned changes. For impact analysis, the skill says to use the separate datalineage-bigquery-asset-impact-analysis skill instead.
- Good: "Summarise the upstream and downstream lineage for bigquery:project.dataset.table."
- Good: "Trace where the efficiency_score column in this table comes from and where it is used."
- Not this skill: "What breaks if I rename this table?" (use the impact analysis skill).
Read the report the agent returns
The agent starts directly with the summary. It states the inferred workflow type and data domain, then lists the primary systems. Upstream and downstream narratives follow under the bold headers Upstream Lineage and Downstream Lineage.
Fewer than five assets are named explicitly. Five or more are grouped by count. The final section lists analysis metadata: locations searched, parent location, depth limit and process-per-link limit. Check this section to see the boundaries of the summary. If not all regions were searched or the depth was limited, the report suggests asking for a rerun with expanded settings.
- Overall Flow Type and Systems Overview come first.
- Column-level reports state the field scope up front.
- Analysis Metadata shows the search boundaries.
Safety notes: scripts and network use
This skill does not run scripts. It does use the network, because it calls Google Cloud lineage services through MCP, gcloud, REST API and bq. It also reads Google's list of Knowledge Catalog locations from a documentation link before each search.
Every gcloud, REST or MCP command the agent runs directly in the terminal must carry a usage attribution tag. The skill's attribution reference explains that the tag identifies only the skill and its version and carries no user data. The tag is set inline on each command and is not saved with gcloud config set.
- Runs scripts: no
- Uses network: yes
- Commands run in the terminal are tagged inline, not persisted
What people use it for
- Debugging a BigQuery table whose figures look wrong by tracing which upstream tables and processing tasks feed it.
- Giving an analyst a readable overview of how a dataset is produced and which dashboards or tables consume it.
- Documenting data provenance for a specific table or column in a team wiki or handover note.
- Checking which column-level sources feed a single field such as efficiency_score.
Frequently asked questions
Is the Data Lineage Summary skill free to use?
The source does not state pricing or licence fees for the skill beyond its licence, which is Apache-2.0. Google Cloud services and the Data Lineage MCP server are used as part of your own Google Cloud setup, so check your project's billing and the service documentation for costs.
What do I need before installing the skill?
You need a Google Cloud project with the Data Lineage (Knowledge Catalog) MCP server available and able to run search_lineage, gcloud CLI and/or REST API access, and bq or gcloud storage. You also need permission to read data lineage in the relevant projects.
Where do I put the skill folder?
Unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder for one project. In the Claude apps you can upload it under Settings → Capabilities → Skills.
Does the skill run scripts or access the internet?
It does not run scripts, according to its listing. It does use the network, because it calls Google Cloud lineage services and reads the Knowledge Catalog locations list from a documentation link.
Can I use this skill to check what breaks if I change a table?
No. The skill says downstream blast-radius impact analysis is out of scope. For that, use the separate datalineage-bigquery-asset-impact-analysis skill.
Why did my summary only cover some regions?
The report lists the locations it searched in its Analysis Metadata. If coverage is partial, ask the agent to rerun with expanded locations or a greater depth limit, as the report itself suggests.
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 data engineer, analyst or documentation owner who needs a readable explanation of one BigQuery table or column's lineage, and you already have the Data Lineage MCP server working. Pick something else if you need impact analysis for planned changes, want to query BigQuery or edit lineage relationships, or cannot set up the MCP server.
Download the skill