Spark MCP has 7 tools with basic naming and descriptions, but significant gaps in schema completeness and parameter documentation. Tool names follow verb_noun convention (spark_sql, spark_describe, spark_cancel), which is good. However, input schemas are visible but lack depth: most parameters have only type and description, with no enums, constraints, ranges, or validation guidance. Descriptions are present but brief (avg ~80 chars), missing context on when to use each tool and what happens on failure. No output schemas documented. Error handling is not evident in tool definitions. The spark_cancel tool accepts an optional 'tag' parameter but provides no guidance on valid tag formats or what happens if the tag doesn't exist. spark_submit_transform and spark_rerun_transform are write operations but lack confirmation/dry-run patterns. Overall, definitions are functional but lack the depth needed for confident LLM tool selection and error recovery.

• Document output schemas for all tools. Specify field names, types, and whether results are paginated. E.g., spark_sql should document that it returns {rows: [{col1: value, col2: value, ...}], total_rows: int}.
• Add parameter constraints: spark_sql row_limit should specify min=1, max=10000; spark_cancel tag should document valid format (alphanumeric, length limits); spark_submit_transform version should clarify semantic versioning or tag format.
• Implement dry-run or confirmation pattern for write operations. Add optional 'dry_run: bool' parameter to spark_submit_transform and spark_rerun_transform, or create separate preview tools.
• Expand tool descriptions to 100-200 chars, including when to use each tool and what it returns. E.g., spark_describe: 'Retrieve the schema and column metadata for a fully-qualified Iceberg table. Use this before spark_sql to understand available columns and types.'
• Add error handling guidance to descriptions. E.g., spark_sql: 'Returns error if Spark Connect is unreachable (connection failure) or if the query is invalid (syntax/semantic error). Distinguish between infrastructure and query issues.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry