Server has 5 well-named tools with clear descriptions and complete input schemas. All tools are READ_ONLY and follow verb_noun naming (get_, search_, extract_, clip_). Descriptions are concise (50 - 150 chars) and explain purpose clearly. Input schemas include types, defaults, and constraints (e.g., context_segments 0 - 5, max_results 1 - 100). However, output schemas are not documented, LLMs cannot predict response structure. Parameter descriptions are minimal (e.g., 'video' lacks format guidance; 'lang' lacks language code examples). Error handling is present but generic ('Failed to fetch transcript...') without recovery guidance. No tool annotations (readOnlyHint, idempotentHint) despite all being read-only and idempotent.

• Document output schemas for all tools. Add a 'returns' field to each tool definition specifying the response structure (e.g., {video_id: string, segments: [{text, offset, duration, timestamp}]}). This enables LLMs to plan multi-step operations.
• Enhance 'video' parameter description: 'YouTube URL (youtube.com/watch?v=..., youtu.be/..., shorts, embed) or 11-character video ID (e.g., dQw4w9WgXcQ)'. Provide format examples so LLMs pass valid inputs.
• Enhance 'lang' parameter description: 'Two-letter ISO 639-1 language code (e.g., en, es, fr, de, ja, zh). Defaults to en. Not all videos have transcripts in all languages.'
• Improve error messages with recovery guidance. When transcript fetch fails, return: 'Transcript not available for video {videoId} in language {lang}. Try: (1) Verify the video ID with extract_video_id(), (2) Try a different language code, (3) Check if the video has captions enabled.'
• Add tool annotations to the TOOLS array. Mark all tools with readOnlyHint: true and idempotentHint: true to signal safety to agents. Example: {name: 'get_transcript', ..., readOnlyHint: true, idempotentHint: true}.