Server has 8 tools with basic schemas and descriptions in Russian. All tools are explicitly registered with name, description, and Zod schemas. However, descriptions are minimal (avg ~60 chars, well below the 194-char baseline), parameter descriptions are sparse or missing, and output schemas are not documented. No error handling guidance, no parameter constraints (enums, ranges), and no security annotations. Tool names follow verb_noun pattern (list_*, get_*, start_*, stop_*, invoke_*) which is good, but parameter documentation is weak. Schemas use Zod but are not fully visible in the provided code excerpt, only partial schema definitions shown. No evidence of pagination support, result limits, or chaining IDs in responses.

• Expand tool descriptions to 100 - 200 characters. Include WHAT the tool does, WHEN to use it (vs similar tools), and what it returns. Example: 'List all Compute instances in a folder. Returns instance IDs, names, statuses, and IP addresses. Use this to discover available VMs before calling get_instance or start_instance.'
• Add parameter descriptions with format, constraints, and examples. For folder_id: 'Yandex Cloud folder ID (UUID format, e.g. b1g4agip9gp6eeww1234). Defaults to YANDEX_CLOUD_FOLDER_ID env var if omitted.' For payload: 'JSON string (max 1MB) passed to the function. Must be valid JSON. Example: {"key": "value"}.'
• Document output schemas for each tool. Specify which fields are returned, their types, and which are required for chaining. Example: 'Returns {instances: [{id, name, status, ipv4_address, ...}], total_count}. Use instance.id with get_instance or start_instance.'
• Add pagination parameters (limit, offset/page_token) to list_* tools. Cap results at 50 items by default. Document in descriptions: 'Returns up to 50 instances per call. Use offset to fetch additional pages.'
• Enhance error messages with recovery guidance. Instead of 'Yandex Cloud HTTP 403: ...', return 'Permission denied. Verify YANDEX_CLOUD_TOKEN has compute.instances.get scope and YANDEX_CLOUD_FOLDER_ID is correct.' This guides the LLM to fix the issue.