Server has 5 tools with basic descriptions and partial schemas. Tool names follow verb_noun convention (lister_*, consulter_*) but descriptions are minimal (10-60 chars, below the 50-200 char LLM-optimized baseline). Only 3 of 5 tools have input schemas; 2 tools (lister_eleves, consulter_messages) have empty input objects with no schema validation. Parameter descriptions exist but are sparse. No output schemas documented. Error handling returns generic JSON stringification without field guidance. No tool annotations (readOnlyHint, etc.). Composition is sound, each tool does one thing, but parameter documentation and output structure fall short of production baseline.

• Add explicit inputSchema to lister_eleves and consulter_messages (even if empty, use { type: 'object', properties: {}, required: [] }) to satisfy schema validation.
• Expand tool descriptions to 80-150 chars. Example: 'consulter_notes' → 'Retrieve all grades and scores for a student. Call lister_eleves first to get valid studentId values. Returns grades by subject with dates and teacher comments.'
• Document output schemas for each tool. Example for consulter_notes: { type: 'array', items: { type: 'object', properties: { subject: { type: 'string' }, grade: { type: 'number' }, date: { type: 'string', format: 'date' } } } }.
• Add readOnlyHint: true to all tool definitions to signal safety and enable agent optimizations for read-only operations.
• Enhance error messages with recovery guidance. Example: 'User not found. Call lister_eleves() to see available students.' or 'Network timeout. Retry in 5 seconds or check your internet connection.'