Three tools with complete schemas and descriptions in Russian. Naming follows verb_noun pattern (find_projects, find_people, utilization_summary). Descriptions are detailed and context-aware (194 - 250 chars), explaining WHEN to use each tool. Parameters have types and descriptions with constraints (e.g., min_free_percent: 0 - 100). However, output schemas are not formally documented, responses are JSON strings without declared field structures. Error messages are generic ('not found', 'no match'). No error recovery guidance. No tool annotations (readOnlyHint present in metadata but not in schema). Composition is sound: three distinct concerns, no overlapping responsibilities.

• Add formal output schema documentation to each tool. Define field names, types, and required fields. Example: find_projects returns [{code: string, name: string, status: string, manager: string, rate_per_hour: number, note: string}].
• Enhance error messages with recovery hints. Replace 'По запросу ... проектов не найдено' with 'Project not found. Try: (1) search by partial name, (2) call find_projects() with empty query to list all, or (3) check project code format (e.g., ALFA, BETA).'
• Add readOnlyHint: true to all three tool schemas to signal to LLM that these are safe, non-destructive operations.
• For find_people and find_projects, replace free-form 'query' and 'role' parameters with enums or provide validation. Example: role enum: ['backend', 'frontend', 'аналитик', 'дизайнер']. Return 'Invalid role. Valid options: backend, frontend, аналитик, дизайнер' on mismatch.
• Add optional limit and offset parameters to utilization_summary to support future scaling. Document that default limit is 100 people; if exceeded, pagination is required.