Server exposes 4 tools with complete JSON schemas and basic descriptions. However, descriptions are generic and lack actionable context for LLM selection. Parameter descriptions are minimal (e.g., 'Descriptive text prompt' for prompt param). No output schema documentation. No error handling guidance. Tool names follow verb_noun pattern but lack specificity about async behavior and polling requirements. Schemas are present but descriptions fall short of production baseline (194 chars avg; these average ~80 chars).

• Expand tool descriptions to 150-250 chars. Example: 'Generate a Wan 3.0 Prime video from text. Returns a request_id; poll get_task_status() to retrieve the video URL when ready. Use this for text-only prompts; use image_to_video() to animate an existing image.'
• Add enum constraints to resolution and aspect_ratio parameters. Document valid values: resolution: ['480p', '720p', '1080p']; aspect_ratio: ['16:9', '9:16', '1:1'].
• Document output schema in tool descriptions. Example: 'Returns {request_id: string, status: string, video_url?: string, error?: string}.'
• Add parameter constraints to descriptions. Example for prompt: 'Text description of the video (1-2000 chars, no special characters).' For duration: 'Video length in seconds (1-60, default 5).'
• Document the async polling pattern. Add to all generation tools: 'This call is asynchronous. Use get_task_status(request_id) to poll for completion. Typical latency: 30-120 seconds.'