Lido MCP has clear, domain-specific tool names (lido_stake, lido_unstake, etc.) and reasonable descriptions (avg ~120 chars). All 8 tools have input schemas with type definitions. However, parameter descriptions are minimal (often single phrases like 'Amount of ETH to stake'), lacking actionable constraints (min/max, format, examples). Output schemas are not documented, LLMs cannot predict return structure. Error handling is absent from tool definitions. No dry-run confirmation pattern despite WRITE operations on financial assets. Rate limiting exists in code but is not exposed to the LLM. Missing tool annotations (destructiveHint, idempotentHint) for financial operations.

• Add min/max constraints and format hints to all numeric parameters. E.g., 'amount (string, 0.001 - 10 ETH, decimal format, e.g., "1.5")'.
• Document output schemas for each tool. E.g., lido_stake returns {txHash: string, stETHAmount: string, gasUsed: string}. Include field types and meanings.
• Add destructiveHint and idempotentHint annotations to tool definitions. Mark lido_stake, lido_unstake, lido_wrap, lido_unwrap as destructive; mark lido_balance, lido_apy, lido_rewards as idempotent and read-only.
• Enhance error handling: return structured errors with category (retryable|user_fixable|fatal), actionable message, and recovery hint. E.g., {error: 'Insufficient balance', category: 'user_fixable', hint: 'Call lido_balance to check available stETH'}.
• Add scope declarations to tool definitions. E.g., lido_stake requires 'write:ethereum, sign:transactions'. This enables least-privilege agent configuration and audit trails.