DataMaster MCP exhibits significant definition quality gaps across the 14 tools. While tool names generally follow verb_noun conventions (connect_, execute_, list_, manage_, get_, analyze_, process_, export_, fetch_, create_, api_), parameter descriptions are present but often generic or incomplete. Most critically: (1) Input schemas lack proper JSON Schema type definitions beyond basic 'string', 'dict', 'int', 'list', no constraints, enums, minimum/maximum, or format specifications visible. (2) Tool descriptions, while present, are overly brief (average ~150 chars) and lack actionable guidance on WHEN to use each tool, dependencies between tools, or error recovery paths. (3) Output schemas are entirely undocumented, callers have no specification of what fields to expect from tools like analyze_data, process_data, or fetch_api_data. (4) Parameters like 'operation' (process_data), 'analysis_type' (analyze_data), and 'action' (manage_database_config, manage_api_config) accept undefined sets of values, no enums, no validation guidance, forcing LLMs to hallucinate. (5) Critical dependencies are unstated (e.g., connect_data_source must be called before query_external_database, but this is nowhere documented). (6) Error handling guidance is absent, tools provide no indication of which errors are retryable, what to do if a connection fails, or how to recover from invalid parameters. (7) Composite operations like 'create_api_storage_session' lack prerequisites documentation. The code shows import statements and function implementations exist, but parameter constraints and output shapes are not visible in the provided source.

• Document output schema for every tool. For example, execute_sql should specify: {"rows": [{"col1": <type>, "col2": <type>, ...}], "row_count": int, "total_count": int}. analyze_data should list available analysis types and return format. This is critical for LLM planning.
• Convert string parameters with constrained values into enums. For process_data, list valid operations: 'filter', 'sort', 'group', 'aggregate', 'deduplicate', 'normalize'. For manage_database_config, specify valid actions: 'list', 'test', 'add', 'remove', 'reload', 'list_temp', 'cleanup_temp'. Enums prevent hallucination.
• Expand parameter descriptions with examples and constraints. 'config' in connect_data_source should read: 'Configuration dict. For source_type=mysql: {"host": str, "port": int (default 3306), "user": str, "password": str, "database": str}. For source_type=excel: {"file_path": str, "sheet_name": str (default first sheet)}.' Provide similar detail for all dict params.
• Add error handling guidance to tool descriptions. E.g., execute_sql: 'On invalid SQL, returns error code with suggestion to check syntax. On connection failure, check database is running and credentials are valid (see manage_database_config test action).' Repeat for all tools that can fail.
• Document tool dependencies and prerequisites. Add to connect_data_source: 'Call this BEFORE query_external_database. Name the config in the return value and pass it to query_external_database as database_name parameter.' Add to manage_database_config: 'Call test action FIRST to verify credentials before add action.'

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

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