Three well-named, read-only tools with clear descriptions and complete input schemas. ask_lore and search_lore have strong, action-oriented names and detailed descriptions (120 - 180 chars). All parameters are typed and described. However, output schemas are not documented in the visible code, LLMs cannot predict response structure. Error handling is present (isError flag) but lacks recovery guidance. No tool annotations (readOnlyHint) despite all being read-only. Composition is sound: ask_lore generates answers; search_lore retrieves without generation; list_sources discovers corpus structure. Parameter validation (str/int helpers) is defensive. Missing: documented output schemas, error categorization, and tool annotations.

• Document output schemas for all three tools. E.g., ask_lore returns {content: [{type: 'text', text: string}], citations: [{source: string, excerpt: string}], isError?: boolean}. Include field types and nesting.
• Enhance error messages with recovery guidance. Instead of 'Question refused', return 'Question refused: exceeds player knowledge ceiling. Try search_lore to find relevant documents instead.'
• Add readOnlyHint: true to all three tool definitions. This signals to agents that these tools are safe to call without side effects.
• Expand list_sources description to ~100 chars: 'Discover what lore documents and DM clarifications are loaded. Call this first to understand corpus structure before asking or searching.'
• Add rationale to search_lore limit parameter: 'Maximum clarifications to return (1 - 25, default 5). Higher limits return more context but increase token cost; lower limits are faster for quick lookups.'

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

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained