Server has 3 tools with complete schemas and descriptions. Tool names follow verb_noun pattern (generate_image, generate_ui_image, list_generated_images). Descriptions are detailed (150-250 chars) and explain WHAT/WHEN/WHY. All parameters have type definitions and descriptions. However, output schemas are not formally documented, responses are inferred from code. Error handling is minimal (no recovery guidance, no error categorization). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Parameters lack validation constraints (e.g., size format, quality enum mapping per provider). Missing dependency hints and chaining IDs in responses.

• Add formal output schema documentation for each tool. E.g., generate_image returns {content: [{type: 'text', text: string}, {type: 'image', data: base64, mimeType: 'image/png'}], metadata: {provider, size, quality, prompt}}. Document list_generated_images return as {images: [{path, size, createdAt}], total}.
• Add tool annotations: generate_image and generate_ui_image should have destructiveHint=true (write files). list_generated_images should have readOnlyHint=true. This enables agents to reason about side effects.
• Enhance error handling with recovery guidance. E.g., 'Image generation failed: API rate limit exceeded. Retry in 60 seconds or try a smaller size.' Categorize errors: retryable (rate limit, timeout), user-fixable (invalid prompt, unsupported size), fatal (auth failure).
• Add validation for 'size' parameter. Accept enum shortcuts (small, medium, large, portrait, landscape) and validate custom WxH format (must be integers, multiple of 32 for Flux). Return: 'Invalid size: 1025x768. Must be multiple of 32. Try 1024x768 instead.'
• Enhance list_generated_images description: 'List all previously generated images with metadata (path, size, creation date). Call this to review past generations before creating new ones.' Add optional parameters: limit (default 20), sortBy (date|size|name).