Three tools with complete schemas and basic descriptions, but significant gaps in parameter documentation and error guidance. generate_image has a comprehensive schema with 8 parameters, all with types and descriptions, but lacks output schema documentation. check_credits and list_models have empty input schemas (no parameters), which is correct but minimal. Descriptions are present but brief (10-50 chars), below the 50-200 char baseline for LLM-optimized docs. No error handling guidance, no recovery paths, and no tool annotations. Parameter constraints exist (enums, min/max) but descriptions lack actionable format guidance. Output structures are inferred from handler code but not formally documented.

• Add formal output schemas to all tools. Document generate_image response as {success: boolean, generation_id: string, model: string, images: [{id, url, saved_to?}], credits_used: number}. This enables LLMs to extract and chain results.
• Expand descriptions to 80-150 chars. Example: 'Generate images using Leonardo AI. Supports 5 models (diffusion, vision, anime, dreamshaper, realistic). Returns URLs and optional local saves. Requires LEONARDO_API_KEY env var.'
• Add error classification to handlers. Return {success: false, error: string, recoverable: boolean, suggestion: string}. E.g., 'API key not set' → recoverable=false, suggestion='Set LEONARDO_API_KEY environment variable'.
• Document parameter constraints in descriptions. E.g., width/height: 'Multiple of 8 pixels (e.g., 512, 1024). Larger sizes consume more credits.' guidance_scale: 'Range 1-20; higher values follow prompt more strictly.'
• Add tool annotations (readOnlyHint, destructiveHint) to indicate side effects. generate_image should have destructiveHint=false (read-only from user perspective, but consumes credits). check_credits and list_models should have readOnlyHint=true.