The server exposes a single tool (quip_read_spreadsheet) with a well-structured JSON schema and input parameters defined. The tool name follows verb_noun convention and the description is detailed, though somewhat verbose and technical. The input schema declares required and optional parameters with types and descriptions. However, the description does not explicitly state what the tool returns beyond a vague reference to 'JSON object' and metadata structure. The tool lacks explicit error handling guidance, recovery instructions, and the description at 560+ characters exceeds the 200-char baseline for optimal LLM consumption. Output schema structure is mentioned in the description but not formally documented as a separate schema definition visible in the code. The server supports resources (reading via alternative URIs) but does not declare tool annotations (readOnlyHint, destructiveHint, idempotentHint) that would clarify this is a read-only, idempotent operation.

• Add a formal outputSchema property to the tool definition: { type: 'object', properties: { csv_content: { type: 'string', description: 'CSV data (may be truncated to 10KB)' }, metadata: { type: 'object', properties: { rows: { type: 'number' }, columns: { type: 'number' }, is_truncated: { type: 'boolean' }, resource_uri: { type: 'string' } } } } }
• Refactor the description to lead with a concise functional summary (50 - 100 chars): 'Read a Quip spreadsheet by thread ID and return CSV content with metadata. Large files are truncated; use the resource interface to access complete data.' Move URI format examples and storage details to implementation comments or a separate technical reference.
• Add tool annotations during registration: { readOnlyHint: true, idempotentHint: true } to signal to agents that repeated calls are safe.
• Add error handling guidance to the description: 'Throws InvalidParamsError if threadId is missing. Returns truncated CSV (10KB max); consult the resource URI in metadata to access complete content. If sheetName is omitted, the first sheet is read.'
• Document the default behavior of sheetName in its parameter description: 'Optional sheet or tab name. If omitted, the first sheet is read.'
• Consider splitting tool description into two parts: (1) concise functional description for LLM selection, (2) detailed implementation notes for developers. This reduces token waste and improves clarity.

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

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