ESPN server has 12 tools with basic Zod schemas and descriptions, but significant gaps prevent higher scores. All tools are READ_ONLY and properly named with action verbs (espn_search, espn_news, etc.). Descriptions exist but are terse (avg ~60 chars, below the 194-char baseline). Critical issue: NO output schemas documented anywhere, LLMs cannot plan downstream calls or extract structured data. Parameters lack enums for constrained values (sport, league accept free-form strings despite known valid sets). No error handling guidance, no pagination support despite tools returning lists, no per-item success/failure for batch operations. Tool composition is reasonable (each does one thing), but missing chaining IDs in responses and no recovery guidance.

• Document output schemas for all 12 tools. Specify return type (object/array), required fields, and field types. E.g. espn_search returns {results: [{id, name, sport, league}], total_count}.
• Convert sport and league parameters to enums. Define valid values in Zod: z.enum(['football', 'basketball', 'baseball', 'hockey', 'soccer', 'mma', 'racing', 'tennis', 'golf']) instead of free-form z.string().
• Add pagination to list-returning tools (espn_news, espn_teams, espn_team_roster, espn_rankings). Include limit (default 20, max 100), offset/page, and return total_count or next_cursor.
• Expand descriptions to 100-200 chars. Add WHEN to use each tool and dependencies. E.g. 'Get live scores for a specific sport/league. Call espn_scoreboard for recent games, espn_live_scoreboard for real-time updates.'
• Add error handling guidance. Include recovery hints in descriptions: 'If team not found, call espn_search() with team name first.' Categorize errors as retryable (network) vs user-fixable (invalid sport).