ArthurLegal exposes 12 tools with significant definition gaps. Most tools have descriptions (good), but parameter schemas are incomplete or missing type information. Tool names are action-oriented (ara=search, getir=retrieve, oku=read) but in Turkish, limiting LLM discoverability. Critical issues: (1) Tools 5-8 (tr_*, az_*, eu_*, uk_*) are forwarded from a remote bridge with no visible schema definitions in the source, these are inferred, capping them at 50. (2) parsel_sorgula, konumdan_parsel, yer_bul, and tapu_kaydi_oku have parameter descriptions but lack explicit type declarations in the visible schema. (3) No output schemas documented for any tool. (4) Error handling is minimal, no recovery guidance or actionable error messages visible. (5) No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being READ_ONLY.

• Add explicit JSON Schema type declarations to all parameters. E.g., parsel_sorgula.il should be {"type": "string", "description": "Province (il) name, e.g. Istanbul, Ankara"}. Currently types are inferred from descriptions only.
• Document output schemas for all tools. E.g., arthurlegal_bilgi_ara returns an array of objects with fields: [{ path: string, score: number, snippet: string }]. Include this in the tool definition or a separate schema reference.
• Add tool annotations (readOnlyHint: true) to all 12 tools since they are all READ_ONLY. This signals to agents that these tools are safe to call without confirmation and can be retried on failure.
• Enhance error messages with recovery guidance. Instead of 'Dosya yok: {yol}', return 'File not found: {yol}. Try arthurlegal_bilgi_ara(profil, sorgu) to search for the file by keyword.' This guides the LLM to the next step.
• Consider English aliases or a dual-naming strategy for remote research tools. E.g., expose both 'tr_legal_research' and 'tr_hukuk_arastirmasi' so LLMs can discover them via English intent matching. Alternatively, add a discovery tool that lists available research tools with English descriptions.