Both tools lack explicit schema definitions, parameter descriptions, and structured output documentation. Tool definitions are inferred from Composio framework calls rather than directly visible in the source. The supabase_toolkit and github_toolkit are instantiated via composio.create(toolkits=[...]) but their actual tool names, input schemas, and output formats are not defined in the provided code. Descriptions are generic and do not explain WHEN to use each tool, what parameters it accepts, or what it returns. No error handling guidance, no parameter constraints, no pagination support documented. This is a framework-delegated pattern where the actual tool definitions live inside the Composio library, not in the MCP server code itself.

• Define explicit tool schemas in the MCP server code (not delegated to Composio). Each tool must have a name, description (50 - 200 chars), input schema with typed parameters, and documented output structure.
• Add parameter descriptions for all inputs. Example: 'user_id (string, required): The Composio user identifier for session management. Use the same user_id across connect() and query() calls.'
• Document output schemas. Example: 'Returns {projects: [{id, name, region}], total_count: number}' so agents know what fields to extract.
• Add error handling guidance. Example: 'If query fails with "connection_error", call connect() first. If timeout, retry with a smaller limit parameter.'
• Implement pagination for list operations. Document limit and offset/cursor parameters and return total_count so agents can iterate over large result sets.