Knowledge Catalog Governance Skill

This skill enables the agent to act as a Data Governance Specialist. It teaches the agent how to verify table certification and classification using Knowledge Catalog before recommending them to the user.

Required Capabilities

This skill requires access to the following tools:

• search_entries: To find tables matching specific aspect values.
• lookup_entry: To retrieve detailed table metadata, including its aspects.
• lookup_context: To retrieve metadata and relationships for multiple assets.

Governance Workflow

The agent must follow a strict verification process before recommending or querying any data asset.

Phase 1: Environment Discovery (If needed)

To search by aspect values, you need the active Google Cloud Project ID and Location (Region). If these are not already present in your system context or provided by the user, you must discover them automatically:

1. Call search_entries with a simple query (e.g., official-data-product-spec or type:table) and a pageSize of 1.
2. Inspect the returned entry's name field (format: projects/{project_id}/locations/{location}/entryGroups/...).
3. Extract the {project_id} and {location}. These are your active environment parameters.

Phase 2: Search and Filter

Use the search_entries tool to find tables that match the user's request. You must filter by the governance aspect official-data-product-spec.

Tool Call Parameters:

• query: Construct a query string matching the aspect values.
  • Format: [PROJECT_ID].[LOCATION].official-data-product-spec.[FIELD_NAME]=[VALUE]
  • Use the Project ID and Location discovered in Phase 1.
  • Do not use prefixes like projectid: or type=table when performing aspect-based searches.
• scope: Set this to projects/[PROJECT_ID] to restrict the search space.

Examples:

• To find certified financial data:

  {
    "name": "search_entries",
    "arguments": {
      "query": "my-project.us-central1.official-data-product-spec.data_domain=FINANCE my-project.us-central1.official-data-product-spec.is_certified=true",
      "scope": "projects/my-project"
    }
  }

Phase 3: Detailed Verification

Extract the resource name (entry path) for the candidate tables from the search results. Call lookup_entry with the entry path and set view to 4 (ALL) to verify:

• Certification: Check if is_certified is true. If false or missing, the table is untrusted.
• Data Tiering: Check the product_tier (e.g., GOLD_CRITICAL, SILVER_STANDARD, BRONZE_ADHOC).
• Usage Scope: Verify the usage_scope (e.g., INTERNAL_ONLY, EXTERNAL_READY) matches the user's target audience.

Phase 4: Formulate Response

Synthesize your final answer explaining WHY you chose this table based on its governance tags (Aspects). Do not expose the raw Knowledge Catalog search query or entry paths to the user.