metatate-governance

Metatate Governance

Use this skill when a user asks about governed data context, data meaning, policy rules, data-use authorization, query validation, decision explanation, or release readiness.

MCP Tools

Use the Snowflake-managed MCP server named metatate when available.

Tool contract:

• discover-context: optional database_name, schema_name, min_sensitivity, compliance_framework, has_pii, domain.
• get-decision-context: required table_name.
• inspect-data-meaning: required table_name, optional column_name.
• inspect-governance-rules: required table_name, optional column_name.
• authorize-use: required table_name, operation, intended_use; optional actor_role, columns_csv, destination_system, destination_jurisdiction, consumer_jurisdiction, context_json, raw_request_text, normalized_request_json, normalization_meta_json.
• validate-query-context: required sql_text; optional operation, intended_use, actor_role, destination_system, destination_jurisdiction, consumer_jurisdiction.
• explain-why: one of decision_id or validation_id.

Use JSON object strings for context_json, normalized_request_json, and normalization_meta_json. Use comma-separated uppercase column names for columns_csv.

Behavior

• Treat Metatate as the source of truth for policy, catalog, authorization, validation, and explanation.
• Do not infer authorization from policy text alone when authorize-use or validate-query-context can answer.
• Keep local repository analysis separate from Metatate facts. Use local files to identify candidate tables, SQL, dbt models, migrations, and code paths, then call Metatate tools for governed context.
• Do not request or expose raw row-level data by default.
• Include decision IDs and validation IDs in user-facing summaries when they are returned.