mcp-sim has 10 well-named tools with clear verb-noun patterns (list_devices, boot_device, stop_device, etc.). All tools have descriptions (avg ~80 chars, within baseline 34-392). Input schemas are present and typed for all tools. However, parameter descriptions are minimal (1-2 lines), lacking actionable constraints, format hints, and dependency documentation. Output schemas are inferred from code but not explicitly documented in tool definitions. Error handling is basic (ToolError with code/msg) but lacks recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Missing pagination for list_devices despite potentially large result sets.

• Add enum constraints to 'platform' parameter across all tools: {"enum": ["ios", "android"]} with description 'Platform identifier (ios or android).'
• Declare tool annotations in mcp.Tool definitions: readOnlyHint=true for list_devices/get_state/await_ready/controller_status; destructiveHint=true for wipe_device; idempotentHint=true for get_state/controller_status.
• Document output schemas explicitly in tool definitions. E.g., list_devices returns {"devices": [{"id": string, "platform": string, "state": string}]}. Include this in the tool's schema field or description.
• Add pagination to list_devices: accept optional 'limit' (1-100, default 20) and 'offset' (default 0) parameters. Return {"devices": [...], "total": number, "offset": number}.
• Enhance error messages with recovery guidance. E.g., ToolError for unsupported platform should suggest 'Call list_devices() to see available platforms.' Errors should be actionable for LLM retry logic.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry