Skip to content
Get packs

Guide · Agent skill

How to use the BigQuery Impact Analysis agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The BigQuery Impact Analysis skill is an agent skill that helps an AI agent work out what breaks downstream when a BigQuery table or view is broken, stale, modified, deleted or paused. It uses Google Cloud data lineage to list the affected assets, their transform processes and how far away they are. This guide explains what an agent skill is, how to install this one, how to ask for an analysis in plain language, and what to check before you rely on the results.

BigQuery Impact Analysisgoogle/skills · Apache-2.0

Quick start

  1. 1Check prerequisites. Confirm a Google Cloud project has the Data Lineage API (Knowledge Catalog) enabled and that you can read BigQuery metadata and lineage.
  2. 2Connect the Data Lineage MCP server. Follow the MCP usage reference to add the Data Lineage remote MCP server to your agent's MCP configuration.
  3. 3Install the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  4. 4Ask with the full asset name. Describe the planned change and give the table or view as project.dataset.table.
  5. 5Let the agent find the location. The agent runs bq show to read the dataset location, or asks you for it if that fails.
  6. 6Review the blast radius table. Check the executive summary, critical path, and table of downstream assets with depth and impact type.
  7. 7Widen the search if needed. Ask for a rerun with expanded locations or a larger depth limit if coverage looks incomplete.

What is an agent skill and what does this one do?

An agent skill is a folder of instructions (a SKILL.md file plus optional reference files and scripts) that an AI agent loads when a request matches the skill's description. It tells the agent how to approach one kind of task.

This skill covers downstream impact analysis, sometimes called blast radius analysis. When you ask about a BigQuery asset, the agent resolves its full name, finds the dataset location, queries the Data Lineage MCP server for downstream relationships, and reports every affected asset in a table.

  • Builds a table of affected assets with the transform process that connects each one and its depth.
  • Flags high-priority items whose names contain prod, dashboard, reporting or master.
  • Reports the search boundaries (locations, depth limit, processes per link) so you can ask for a wider search.

Install the skill

You can install the skill in one of two ways. Download the ZIP file from this page and unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it limited to one project.

Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

  • 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.

Set up what the skill needs

The skill needs a Google Cloud project with the Data Lineage API (Knowledge Catalog) enabled, access to the Data Lineage remote MCP server, and IAM permissions to read BigQuery metadata and lineage. For connection details, the skill points to the MCP usage reference, which includes a client configuration block for your agent's MCP configuration file.

The gcloud CLI and bq command-line tool are optional. They are used to find the dataset location with bq show. If they are missing or you lack permissions, the agent will ask you for the dataset's location instead.

  • Google Cloud project with the Data Lineage API (Knowledge Catalog) enabled.
  • Access to the Data Lineage remote MCP server.
  • IAM permissions to read BigQuery metadata and lineage.
  • Optional: gcloud CLI and bq tool, or simply tell the agent the dataset location.

Trigger the skill with natural requests

You do not need a special command. Describe the situation in plain words that include the table or view name in full, using the project.dataset.table form. The skill's description covers blast radius analysis, modifying, deleting or pausing updates, and finding downstream dependencies.

The agent expects the asset as bigquery:{project_id}.{dataset_id}.{table_or_view_id}. If you give a shorter name, the agent will build the full name for you.

  • What breaks downstream if I delete my-prod-project.analytics.orders?
  • Run an impact analysis on the orders table before I pause its updates.
  • Our revenue dashboard is stale. Which upstream-derived assets could be the cause, and what depends on this view?
  • List the downstream dependencies of my-prod-project.analytics.orders.

Read the results

The agent reports an executive summary with the total number of downstream assets and the maximum depth, then a critical path section, then the blast radius table. Each row shows the downstream asset, its transform process, its depth and whether the impact is Direct or Indirect. A link marked EXACT_COPY is labelled Directly Stale / Identical Copy.

The final section lists the locations searched, the parent path, the depth limit (max_depth 10 by default) and the process-per-link limit (max_process_per_link 5 by default). If the lineage response is empty, the agent reports that no dependencies exist in the queried locations.

  • Direct (depth 1): assets directly consuming the source.
  • Indirect (depth greater than 1): assets further down that may get stale data or failures.
  • Empty result: no dependencies in the queried locations, not proof that none exist elsewhere.

Safety notes: scripts and network access

This skill runs scripts and uses the network. It runs bq show to find a dataset's location and calls the Data Lineage MCP server. Before running any gcloud, REST or MCP command, the agent must read the attribution reference. For commands typed directly in the terminal, it prefixes gcloud with CLOUDSDK_METRICS_REQUEST_ATTRIBUTION set to a skill tag and adds that tag to User-Agent headers on direct HTTP requests.

The skill is designed to only read lineage. It does not create or edit lineage links. Review the commands the agent runs, and make sure the Google Cloud account it uses has only the read permissions you intend.

  • Skill is licensed Apache-2.0.
  • It runs scripts and makes network calls to Google Cloud.
  • Lineage is read-only: the skill does not create or change links.

Tips for better results

Give the full FQN and, if you know it, the dataset location, so the agent does not need to look it up. If you expect dependencies in other regions, say so, because the search only covers the locations it queried.

Ask for a rerun with expanded locations or a larger depth limit when the first result looks incomplete. Remember that lineage is limited to what the Data Lineage service records, so treat results as a starting point for review, not a guarantee.

  • Include the project, dataset and table in one name.
  • Name the locations you know are involved, such as us and us-central1.
  • Ask for a deeper search if the first table seems too short.

What people use it for

  • Checking downstream dashboards and reports before deleting a production BigQuery table.
  • Finding which views and scheduled queries are affected before pausing updates to a source table.
  • Tracing why a reporting dashboard is stale by mapping its upstream chain back to the source asset.
  • Producing a quick list of downstream dependencies for a platform review.

Frequently asked questions

How do I install the BigQuery Impact Analysis 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.

What do I need to use this skill?

You need a Google Cloud project with the Data Lineage API (Knowledge Catalog) enabled, access to the Data Lineage remote MCP server, and IAM permissions to read BigQuery metadata and lineage. The gcloud CLI and bq tool are optional for finding the dataset location.

Is the BigQuery Impact Analysis skill free?

The skill is licensed under Apache-2.0. The source does not state pricing for the Google Cloud services it uses, so check your Google Cloud account for any charges.

Does the skill change my lineage or BigQuery tables?

No. The skill only maps downstream lineage. It does not create or edit lineage links, and it is not meant for answering general data questions.

Does it work with Cloud Storage files?

Only when those files appear in BigQuery lineage. Otherwise results may be incomplete, because the skill is designed for BigQuery assets.

Why did the agent return no dependencies?

An empty lineage response means no dependencies were found in the locations queried. Check that the dataset location is correct and widen the search locations or depth limit before concluding nothing depends on the asset.

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

This skill suits data engineers planning a delete, rename or pause on a BigQuery table, analytics teams tracing stale dashboards, and platform owners who need a quick downstream list. It is not a tool for general BigQuery querying or for editing lineage, so use BigQuery tools for those tasks.

Download the skill