Single tool with basic schema and description. Tool name 'search_movies' follows verb_noun convention (+). Description is present but minimal (47 chars, below 50-200 char baseline for LLM optimization). Input schema is properly typed with required/optional params, but lacks critical details: no enum constraints, no format specifications, no output schema documented, no error recovery guidance. Parameters lack depth, 'query' description is generic, 'year' and 'page' descriptions are terse. No documentation of pagination behavior, result limits, or what fields the response contains. Error handling returns raw API errors without actionable recovery steps. Missing: output schema, per-parameter validation rules, result limits, chaining IDs for downstream tools.

• Add output schema documenting response structure: { results: [{id, title, release_date, overview, poster_path}], total_results, total_pages, page }. Include field descriptions so LLM knows what data is available.
• Expand tool description to 100-150 chars: 'Search TMDB for movies by title, year, or keywords. Returns up to 20 results per page. Use pagination (page param) for large result sets. Requires valid TMDB API token.'
• Add parameter constraints: year (1800-2100, optional), page (1-1000, default 1), query (1-255 chars, required). Document in descriptions: 'year: Release year filter (1800-2100, optional)'.
• Implement actionable error messages: 'Invalid query: must be 1-255 characters' or 'TMDB API rate limit exceeded. Retry after 60 seconds.' Include classification (retryable vs user-fixable).
• Document result limits and pagination in description: 'Returns max 20 results per page. Use page param to iterate. Call with page=1 first, then increment page until total_pages is reached.'