BigQuery Asset Impact Analysis
Map the downstream blast radius of BigQuery table or view changes.
The evidence shows a primarily read-only lineage workflow, explicit bypass restrictions, no guessing of asset names, and parameter transparency; however, it does not specify least-privilege IAM, sensitive-data handling, user confirmation, remote data flows, recovery, or rollback, so points are deducted.
The workflow from FQN resolution through location discovery, lineage traversal, and formatted output is mostly coherent, and it gives some permission/tool failure guidance; however, there are no tests or execution evidence, region configuration depends on external MCP definitions, and the instruction to treat an empty response immediately as no dependencies is unsafe, so points are deducted.
The description, use cases, exclusions, FQN input, direction, and search parameters are clear; however, the skill depends on Google Cloud Data Lineage MCP, bq, and credentials, with no evidence of Chinese-language support or mainland-China network reachability and limited boundary documentation, so points are deducted.
The documentation is readable and layered, with prerequisites, a reference file, parameter details, output examples, Apache-2.0 licensing, installation information, official provenance, and a support path; however, it lacks skill versioning, a changelog, named maintenance responsibility, a clear update path, FAQs, and systematic troubleshooting, so points are deducted.
The required executive summary, critical path, complete dependency table, and analysis metadata are directly useful for blast-radius assessment; however, the source files provide no verification of actual outputs, and results depend on external lineage coverage, permissions, and regional configuration. The empty-response rule can also produce false conclusions, so points are deducted.
The supplied material is revision-scoped and includes source files, documentation references, and explicit parameters, providing limited auditability; however, it contains no committed tests, CI coverage, execution records, or corroborating operational evidence, so key paths are not independently reproducible from static review.
- Verify bq and MCP permissions, project/user-project configuration, and regional consistency first; least-privilege IAM is not defined.
- Do not treat an empty or permission-filtered lineage response as proof of no dependencies; distinguish no results, insufficient permissions, wrong region, and service failure.
- The skill provides no sensitive-data redaction guidance, remote MCP data-flow disclosure, user-confirmation step, or rollback plan.
- Core functionality depends on Google Cloud Data Lineage MCP, Google credentials, and external services; Chinese support and mainland-China reachability are not evidenced.
What does this skill do, and when should you use it?
This skill assesses downstream impact when a BigQuery table or view is broken, stale, missing, or scheduled for maintenance. It uses the Google Cloud Data Lineage MCP Server to discover affected downstream tables, dashboards, and processes. The workflow distinguishes direct from indirect impact and inspects the ETL pipelines, views, or scheduled queries that propagate data. It is a good fit for teams already using Google Cloud lineage services and needing change-risk visibility.
It resolves the asset's fully qualified name in the form bigquery:{project_id}.{dataset_id}.{table_or_view_id}. It runs bq show --format=json {project_id}:{dataset_id} to obtain the dataset location, then builds the Data Lineage API parent path using the MCP server location. It calls DataLineageServer:search_lineage with DOWNSTREAM, max_depth = 10, and max_process_per_link = 5. It then reports affected assets, transform processes, depth, impact type, critical paths, and search metadata.
- A data engineer is changing a production BigQuery table and needs to identify affected downstream tables and pipelines.
- A data platform team finds stale BigQuery data and needs to locate dashboards and reports exposed to cascading issues.
- An operator plans to pause updates to a BigQuery asset and needs a blast-radius assessment before maintenance.
- A data governance or analytics user needs the direct and indirect downstream dependencies of a BigQuery table or view.
What are this skill's strengths and limitations?
- Focused specifically on downstream impact analysis for BigQuery tables and views.
- Distinguishes direct, indirect, and EXACT_COPY-related stale-copy impact.
- Includes assets, propagation processes, depth, and search boundaries for review.
- Defines clear guardrails for empty results, missing assets, and lineage-query bypasses.
- Requires the Google Cloud Data Lineage API, Data Lineage MCP Server, and bq CLI.
- The supplied source does not document MCP configuration details or provide a test suite or verification results.
- It is not intended for general BigQuery querying, non-BigQuery assets, or direct lineage-link editing.
- Results are bounded by queried locations, maximum depth, and processes-per-link limits.
How do you install this skill?
The skill is located at skills/cloud/datalineage-bigquery-asset-impact-analysis/SKILL.md. Install the repository and select this skill during installation with npx skills add google/skills. The supplied source does not document a standalone installation command or complete Data Lineage MCP Server configuration; use also requires Google Cloud Data Lineage API access and an active MCP client connection.
How do you use this skill?
Give an enabled agent a BigQuery asset, for example: Analyze the downstream impact of modifying bigquery:my-prod-project.analytics.orders. The agent needs the correct project, dataset, and table or view name. If bq show cannot determine the location, provide the dataset location. The resulting analysis should include the affected-asset count, maximum depth, critical path, complete dependency table, and search metadata.