TubePilot has 32 tools with consistent naming patterns (verb_noun style: get_*, search_*, extract_*, etc.) and descriptions for all tools. However, schema quality is inconsistent: most tools have parameter schemas with descriptions, but several lack proper constraints (enums missing for fields like 'order', 'duration', 'quality'). Output schemas are not documented in tool definitions. Parameter descriptions vary from good (specific format hints) to minimal. Error handling is basic (try/catch returns generic error text). The server is STDIO-only, which limits protocol readiness to 50 max regardless of definition quality.

• Document full output schemas for all tools. For example, get_video_details should declare: {videoId: string, title: string, viewCount: number, likeCount: number, publishedAt: ISO8601 string, channelId: string, ...}. This enables LLMs to reliably compose multi-step workflows.
• Convert informal enum descriptions ('default: relevance') to explicit JSON Schema enums in parameter definitions. For search_videos, declare order as enum ['date', 'rating', 'relevance', 'title', 'viewCount'] to prevent hallucinated values.
• Enhance error responses with recovery guidance. Instead of returning isError: true with generic text, categorize errors and provide actionable next steps. E.g., 'Video not found (videoId: dQw4w9WgXcQ). Try searching with search_videos first. Available channels: <list>.'
• Add min/max constraints to numeric parameters. Use JSON Schema properties like 'minimum': 1, 'maximum': 10 for count, maxResults, and similar fields. Document in descriptions too: 'Number of frames to extract (1-10, default: 5).'
• Implement tool annotations in handler definitions. Mark all tools with readOnlyHint: true since they are all READ_ONLY. This enables clients to optimize caching and safety warnings.
• Document pagination strategy for list-returning tools. Specify whether result sets are capped, whether offset/cursor is supported, and how to fetch next page. Add to tool descriptions: 'Returns up to maxResults items. For additional results, call again with offset parameter.' (if supported) or 'Results capped at maxResults per API rate limits.'