SystemPrompt MCP Core has moderate definition quality with inconsistent tool design. Of the 3 tools, only systemprompt_config has a detailed schema (9 properties with descriptions). The other two tools (systemprompt_introduction, systemprompt_heartbeat) have empty input schemas with minimal descriptions. Tool naming follows verb_noun convention adequately, but descriptions lack clarity on when/why to use each tool. No output schemas are documented. Error handling is absent from visible tool definitions. The codebase shows TypeScript + SDK integration with proper structure, but tool definitions in src/constants/tools.ts lack the depth needed for production LLM agents.

• Add output schema documentation to all tools. Example for systemprompt_heartbeat: 'Returns {user: {name, email, ...}, billing: {...}, subscription: {...}}' with field types.
• Expand systemprompt_introduction description: 'Returns a comprehensive introduction to systemprompt including core features, use cases, and quick-start guidance. Call this when the user is new to the platform or asks what systemprompt does.'
• Convert systemprompt_config parameter descriptions to use enum constraints instead of inline examples. E.g., communication_formality: {type: 'string', enum: ['casual', 'formal'], description: '...'}
• Add maxLength constraints to unbounded string parameters: expertise_background, expertise_history, personal_background, personal_history should each have maxLength (e.g., 1000).
• Add error recovery guidance to tool descriptions. E.g., 'If user is not authenticated, call systemprompt_introduction first or check API key configuration.'
• Clarify systemprompt_config side effects: does it persist to a backend? Immediate effect or queued? Return confirmation of which fields were updated.
• Add example invocation to systemprompt_heartbeat description to clarify it takes no parameters and returns a JSON object with user/billing/subscription details.
• Document idempotence for each tool. E.g., is systemprompt_config idempotent? Can it be safely retried with identical params without duplicating side effects?

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

• Relies on Sampling (deprecated) - integrate directly with the LLM provider API