Moanete exposes 7 tools with minimal parameter descriptions and no documented output schemas. Tool names follow verb_noun convention (get_*, list_*, call_*), which is good. However, 4 of 7 tools have empty input schemas ({}), and descriptions are generic (10-50 chars), well below the 194-char baseline. The ask_question tool has one parameter with a description, but others lack parameter-level guidance. No error handling guidance, no output schema documentation, and no indication of idempotency or side effects. The call_external_tool tool accepts a JSON-encoded string for args, which is error-prone and forces the LLM to construct JSON manually. Resources are defined but lack descriptions. Overall, this is a minimal implementation suitable for a prototype but not production-grade.

• Expand tool descriptions to 100 - 200 characters. Example: 'get_transcript' → 'Retrieve the live meeting transcript as a formatted speaker-by-speaker dialogue. Returns empty string if no session is running. Use this to provide context for follow-up questions or summaries.'
• Document output schemas for all tools. Add a comment or JSDoc block showing the return type: { content: [{ type: 'text', text: string }] } or { content: [{ type: 'text', text: string }], metadata?: { ... } }.
• Replace the 'args' string parameter in call_external_tool with a structured object. Either accept args: Record<string, unknown> (requires JSON Schema additionalProperties) or enumerate common parameters (e.g., arg1, arg2, arg3) with descriptions.
• Add parameter descriptions to ask_question and list_external_tools. Example: 'server' → 'Name of the external MCP server to query. Call list_external_servers() first to see available servers.'
• Add error handling guidance to tool descriptions. Example: 'If the session is not running, returns "No transcript yet, session may not be running." Call the UI to start a session, then retry.'