18 tools with consistent verb-noun naming (get_, create_, list_, fund_, freeze_, etc.). Descriptions present for all tools (avg ~120 chars), but many lack depth on WHEN to use them or prerequisites. Input schemas visible for all tools with proper type definitions (string, object). Critical gap: NO output schemas documented anywhere, LLMs cannot plan downstream calls or extract return values. Error handling descriptions missing entirely. Security-sensitive tools (create_card, get_card_details) lack permission/scope declarations. Composition is reasonable (single responsibility per tool), but no pagination guidance for list_cards or get_transactions despite potential for large result sets.

• Document output schema for every tool. Include field names, types, and whether fields are always present or conditional. Example: create_card returns {cardId: string, pan: string, cvv: string, expiry: string, status: 'active'|'pending'}.
• Add permission/scope declarations to sensitive tools. Example: 'Requires: write:card, audit:card-creation. Logs: caller_id, timestamp, amount, card_id.'
• Enhance descriptions with error recovery guidance. Example for create_card: 'If insufficient balance, call get_wallet_status to check USDC balance, then call get_fund_link to top up wallet.'
• Add pagination to list_cards and get_transactions. Include limit (default 20, max 100) and offset/cursor parameters. Return total_count in response.
• Clarify telegram tool prerequisites and sequencing. Example: 'Call connect_telegram first to register wallet. Then call telegram_link to get the deep-link for user to scan.'