Critical gaps across all dimensions. Tool names lack clear action verbs (op_alive, op_twofa, op_privacy use 'op_' prefix that obscures intent). Descriptions are present but written in Chinese with domain-specific jargon, making them inaccessible to English-speaking LLMs. Parameter schemas exist but lack proper type definitions and descriptions in many cases. No documented output schemas. Error handling and recovery guidance absent. Security concerns: tools manipulate Telegram accounts with no visible permission gates or audit trails. The server is STDIO-only, capping protocol readiness at 50.

• Rename tools to follow verb_noun pattern: op_alive → check_account_alive, op_twofa → update_2fa_password, op_privacy → set_privacy_settings, etc. This makes intent immediately clear to LLMs.
• Translate all descriptions to English and expand to 50-200 characters. Include WHAT the tool does, WHEN to use it, and what it returns. Example: 'Check if a Telegram account is active by sending a lightweight GetAppConfigRequest. Returns true if account responds, false if dead. Use before batch operations to filter inactive accounts.'
• Document output schemas for every tool. Specify field names, types, and what they represent. Example for op_alive: {"alive": boolean, "response_time_ms": integer, "error": string|null}.
• Add confirmation parameters to destructive tools. Example: op_twofa should require {"confirm_2fa_change": true} to prevent accidental execution. Document that confirmation is mandatory.
• Expand parameter descriptions with constraints and examples. For op_privacy 'items' param: 'Object with keys from [last_seen, phone, invite, avatar, call, forward]. Values must be one of: everybody, contacts, nobody. Example: {"phone": "nobody", "last_seen": "contacts"}'.