Shiplo MCP has well-structured tool definitions with complete input schemas and descriptions for all 8 tools. Naming follows verb_noun convention (platform_*). However, several parameter descriptions lack actionable detail (e.g., 'Site UUID override' without format guidance), and output schemas are generic (additionalProperties: true) rather than documented. Error handling exists but lacks recovery guidance. The platform_deploy_static tool has an exceptionally detailed description (280+ chars) that guides LLM behavior, but others are terse (30-50 chars). No tool annotations (readOnlyHint, destructiveHint) despite clear risk classifications in metadata.

• Document output schemas for all tools. Replace OBJECT_OUTPUT_SCHEMA with explicit field definitions (e.g., account_status returns {plan: string, usage_bytes: number, max_bytes: number}). This enables LLMs to extract and chain data correctly.
• Add tool annotations to all tool definitions: readOnlyHint for platform_account_status, platform_list_sites, platform_inspect_project, platform_deployment_status; destructiveHint for platform_delete_site; idempotentHint for platform_deploy_static (if resumable via deployment_id).
• Expand parameter descriptions with actionable constraints: 'preferred_subdomain: 3-63 lowercase alphanumeric and hyphens, e.g. my-site'; 'max_bytes: target size in bytes (e.g., 5242880 for 5MB); use plan.max_file_size_bytes from platform_account_status'; 'site_id: UUID format (8-4-4-4-12 hex digits)'.
• Enhance error handling in apiRequest() to return recovery guidance. E.g., if site not found, catch and return 'Site not found. Call platform_list_sites() to see available sites.' Wrap errors in a structured {error, recovery_hint} response.
• Expand terse descriptions (platform_account_status, platform_list_sites, platform_inspect_project, platform_deployment_status) to 100-150 chars, following the pattern of platform_deploy_static. Include when to call and what to do with the result.