Server declares 25 tools with consistent verb_noun naming (lol_get_*, lol_analyze_*) and Zod schemas. All tools have descriptions (10-100 chars) and typed parameters with descriptions. However, descriptions are uniformly terse (avg ~50 chars vs. baseline 194 chars), lacking context on WHEN to use each tool, dependencies, or return structure. No output schemas documented. No error handling guidance. Parameter constraints (e.g., tier enum values) are mentioned in descriptions but not enforced as enums in schemas. Tools like lol_get_league_entries and lol_get_league_entries_exp are near-duplicates, signaling unclear composition. No pagination, rate-limit, or result-limit guidance despite 25 tools that likely return lists.

• Expand tool descriptions to 100-200 chars. Include WHEN to use (e.g., 'Call this after lol_get_account to fetch ranked tier and LP'), what it returns (e.g., 'Returns tier, rank, LP, and win/loss counts'), and any prerequisites.
• Document output schemas for all tools. Specify which fields are returned (e.g., lol_get_account returns {puuid, gameName, tagLine, accountId}) so agents can chain calls correctly.
• Convert tier and division parameters to Zod enums: z.enum(['IRON', 'BRONZE', 'SILVER', 'GOLD', 'PLATINUM', 'EMERALD', 'DIAMOND']). Remove enum values from descriptions; let the schema enforce constraints.
• Merge lol_get_league_entries and lol_get_league_entries_exp into a single tool. If 'experimental' refers to pagination or filtering, add optional parameters (page, limit) instead of duplicating the tool.
• Add pagination guidance: specify max count/limit values (e.g., 'count: 1-100, default 20'), document cursor/offset behavior, and return total_count or next_cursor in responses so agents can iterate safely.