marvin-mcp has well-structured tool definitions with clear naming (verb_noun pattern), documented schemas using Zod, and descriptions for all tools. However, parameter descriptions are sparse or missing for several tools (marvin_brief, marvin_hierarchy, marvin_tasks has minimal param docs), and output schemas are not explicitly documented. The apply tool is complex with discriminated unions but lacks guidance on error recovery. No tool annotations visible in source despite errorReporting=true claim. Descriptions average ~80 chars, below the 194-char baseline for A+ tools.

• Add explicit output schema documentation for each tool. Example: marvin_tasks should document that it returns {tasks: [{id, title, scheduledFor, dueBy, ...}], total: number}. Use JSDoc @returns or TypeScript type exports.
• Expand parameter descriptions in marvin_tasks to clarify filter semantics: 'query: substring match on task title (case-insensitive)', 'parent: container ID or name; if omitted, searches all containers', 'done: include completed tasks (default: false)', etc.
• Add error recovery guidance to marvin_apply. Document which errors are retryable (network timeout) vs user-fixable (invalid task ID) vs fatal (permission denied). Example: 'If task ID not found, call marvin_tasks to verify the ID exists.'
• Implement and expose tool annotations. Add readOnlyHint to marvin_brief, marvin_tasks, marvin_hierarchy. Add destructiveHint to marvin_apply (delete action). Add idempotentHint to marvin_complete (idempotent if task already done).
• Document the marvin_apply change set format more explicitly in the tool description. Clarify that changes are applied in order, atomically, and that partial failures will halt the sequence. Provide an example change set.

Inferred effective spec: 2025-06-18+.

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