Three tools with clear names and reasonable descriptions, but critical gaps in schema completeness and parameter documentation. setup_status and list_models have minimal input schemas (empty/single param), while generate_video lacks critical safety parameters (duration, quality, model selection). Descriptions are adequate (60-150 chars) but lack actionable guidance on prerequisites, error recovery, and billing implications. No output schemas documented. Missing parameter descriptions for 'filter' in list_models. No error handling guidance visible in definitions.

• Add output schema documentation for all three tools. Example for generate_video: {type: 'object', properties: {operation_id: {type: 'string', description: 'Unique operation ID to poll with check_video_operation'}, status: {type: 'string', enum: ['PENDING', 'RUNNING', 'COMPLETED', 'FAILED']}, estimated_duration_seconds: {type: 'number'}}}
• Add safety parameters to generate_video: max_duration_seconds (enum: [6, 10, 15], default 6), quality (enum: ['standard', 'high'], default 'standard'), model_id (optional override, defaults to VEO_MODEL). Document each in description.
• Implement confirmation pattern for generate_video: add a 'dry_run' boolean parameter (default false). When true, return estimated cost and duration without charging. Requires user confirmation before actual generation.
• Expand list_models description to explain filter behavior: 'Text substring match (case-insensitive) against model name and description. E.g., filter="veo" returns veo-3.1-generate-preview, veo-2-generate-preview. Omit to list all available models.'
• Add error recovery guidance to setup_status description: 'If GEMINI_API_KEY is missing, visit https://aistudio.google.com/apikey to create a free key. If keys are set but tools fail with 404, call list_models to verify the account can reach the configured model IDs.'