Two tools with basic descriptions and minimal schemas. search_docs has a reasonable description (145 chars) explaining purpose and output format, but lacks output schema documentation. list_documents has a trivial description (71 chars) with no parameters and no output schema. Neither tool declares input validation constraints, error handling guidance, or pagination limits. Parameter descriptions are present but generic. No tool annotations (readOnlyHint, idempotentHint). Schemas are minimal JSON with type and description only, no constraints, enums, or ranges. This is typical of educational/demo MCP servers but falls short of production quality.

• Add output schema to search_docs: document that it returns an array of objects with fields like {source: string, content: string, score?: number}. Include total_results and next_cursor for pagination.
• Add output schema to list_documents: document the structure (e.g., array of {filename: string, size: number, indexed_at: string}).
• Add pagination to search_docs: accept limit (1-100, default 20) and offset or cursor parameters. Return total_results so agents know if more results exist.
• Add error handling descriptions: 'If no documents match the query, returns empty array. If corpus is empty, returns error with guidance to index documents first.'
• Declare tool annotations in MCP schema: add readOnlyHint=true to both tools to signal they are safe, idempotent read operations.