ImageReader has clear, well-structured tool definitions with good descriptions and complete input schemas. Both tools follow verb_noun naming (read_image, get_image_info) and include parameter descriptions. However, output schemas are not explicitly documented in the visible code, the return type is inferred as list[TextContent | ImageContent] but no structured schema is shown. Error handling is present (TimeoutError, FileNotFoundError, ValueError) but lacks recovery guidance for LLMs. The await_for_seconds parameter with a sensible default (3.0) is well-designed. Overall solid foundation but missing output documentation and error recovery hints.

• Add explicit output schema documentation to both tools. For read_image, document that it returns [TextContent with description, ImageContent with mime_type and data]. For get_image_info, document the metadata structure (source, path/url, mime_type, size_bytes, resolution).
• Enhance error messages with recovery guidance. E.g., TimeoutError: 'File did not appear within 3.0 seconds. Try increasing await_for_seconds or verify the file path exists.' FileNotFoundError: 'File not found at {path}. Check the path is absolute or relative to the working directory.'
• Add tool annotations via FastMCP decorator. Mark both tools with readOnlyHint=true since they do not modify state. This aligns with MCP 2026-07-28 spec.
• Document the MIME type constraint more formally. Add a note in parameter descriptions: 'Supported formats: PNG, JPEG, WebP, GIF, BMP, TIFF, SVG, ICO. Unsupported formats will raise ValueError.'
• Add a note about maximum file size or timeout behavior. E.g., 'Remote images are fetched with a 30-second timeout. Very large files may fail; consider checking file size with get_image_info first.'