Single tool with complete JSON Schema and basic descriptions. Tool name 'video_to_markdown' is verb-noun and clear. Input schema properly typed with enum constraint on 'lang' parameter. However, descriptions are minimal (tool: 28 chars, params: 15-40 chars), falling below the 50-200 char LLM-optimized baseline. No output schema documented. Error handling returns user-facing messages but lacks recovery guidance. Parameter descriptions lack format/constraint details (e.g., URL validation rules, digest behavior specifics). No tool annotations (destructiveHint, idempotentHint). Validation logic exists in code but not exposed in schema.

• Expand tool description to 80-150 chars: 'Convert YouTube video audio and visuals to Markdown. Supports Japanese and English output. Requires GEMINI_API_KEY environment variable. Returns markdown content, file path, and format warnings.'
• Add output schema to tool definition: {type: 'object', properties: {markdown: {type: 'string', description: 'Generated markdown content'}, file_path: {type: 'string', description: 'Absolute path to saved file'}, warnings: {type: 'array', items: {type: 'string'}, description: 'Format validation warnings'}}}
• Enhance 'url' parameter description: 'YouTube video URL (e.g., https://youtube.com/watch?v=dQw4w9WgXcQ). Must be a valid, publicly accessible YouTube link.'
• Clarify 'digest' parameter: 'If true, generate timestamped summary of key points instead of full verbatim transcript. Reduces output length and processing time.'
• Improve error messages with recovery hints: 'GEMINI_API_KEY not set or invalid. Set the environment variable before starting the server, then retry.' and 'Invalid YouTube URL. Ensure the URL is a valid youtube.com/watch?v=... link and the video is publicly accessible.'