Three tools with basic verb-noun naming (fetch_image, process_local_image, get_image_info). All have short descriptions (20-50 chars) that lack context for LLM selection. Input schemas present but minimal: only 'url' and 'file_path' parameters with basic string types and descriptions. No output schemas documented. No error handling guidance, no parameter constraints (enums, ranges), no indication of idempotency or side effects. Missing: pagination, batch operations, recovery guidance, and structured output documentation. Descriptions are too terse to guide LLM reasoning about when to use each tool vs. alternatives.

• Expand tool descriptions to 100-200 chars. Example: 'fetch_image: Download an image from a URL and return it as an MCP Image object. Supports JPEG, PNG, GIF, WebP. Max 1024px dimension, 800KB after compression. Use this for remote URLs; use process_local_image for local files.'
• Document output schema for all tools. Specify that Image objects contain 'data' (bytes), 'format' (string: jpeg|png|gif|webp), and implicit size constraints. For get_image_info, document returned fields (width, height, format, file_size).
• Add parameter constraints: 'url' must be a valid HTTP(S) URL; 'file_path' must be an absolute or relative path to an existing file. Include format hints: 'Supported formats: JPEG, PNG, GIF, WebP. Images larger than 1024px will be resized.'
• Add error handling guidance in descriptions. E.g., 'If the image is too large to process, the tool will compress it. If compression fails, check that the file is a valid image format.'
• Add idempotency hints. Clarify whether repeated calls with the same input produce identical results (important for agent retry logic).

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

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