Server has 17 tools with schemas and descriptions present, but quality is inconsistent. Most descriptions are adequate (50-150 chars) but lack actionable context for LLM selection. Parameter descriptions are minimal or absent. No error handling guidance, no output schemas documented, no tool annotations. Many tools combine multiple concerns (e.g., send_transaction does gas estimation + dry-run). Naming is mostly verb-noun but some tools are vague (e.g., 'call_contract' could mean read or write). Critical security issue: PRIVATE_KEY exposed as environment variable with no mention of secret injection pattern. No pagination support despite tools returning lists. Schemas present but incomplete, many lack type constraints, enums, or ranges.

• Add output schema documentation for every tool. Specify return type (object/array), required fields, and field types. E.g., yield_farm returns [{pool_name: string, apy: number, tvl: string, risk_level: enum}].
• Implement secret injection pattern: move PRIVATE_KEY to secure vault (e.g., AWS Secrets Manager, HashiCorp Vault). Never expose in environment or tool parameters. Use server-side signing for write operations.
• Add error handling guidance to write tools (send_transaction, swap_tokens, call_contract, execute_flash_loan). Include recovery hints: 'If gas estimation fails, try base_gas_estimate() first' or 'If swap fails due to slippage, reduce amount or increase slippage tolerance.'
• Split call_contract into read_contract (read-only, no write param) and execute_contract (write operations). Add tool annotations: readOnlyHint=true for read_contract, destructiveHint=true for execute_contract.
• Add pagination to yield_farm and get_transaction_history: include limit (default 20, max 100), offset/cursor, and total_count in response. Document in descriptions: 'Returns up to 20 results; use offset to paginate.'