Server implements 6 tools with mostly adequate naming (verb-first convention) and descriptions. Schemas are present and typed for all parameters. However, several critical issues reduce quality: (1) Missing output schemas, tool responses are undocumented, making it impossible for LLMs to plan downstream operations. (2) Parameter descriptions are minimal (10-30 chars on average) and lack actionable constraint details. (3) No error handling guidance, no recovery suggestions for common failure modes. (4) Missing validation constraints on string parameters (e.g., 'url' has no format or pattern; 'limit' has no min/max). (5) Composition issues, page_search and page_get overlap in intent and naming clarity. Overall: functional but below production baseline for agentic tools.

• Document output schema for every tool. Example for page_search: 'Returns array of {page_id, title, snippet, url, last_modified}'. Example for page_get: 'Returns {page_id, title, content, url, last_modified_by, created_at}'.
• Expand parameter descriptions to 50-100 chars with constraints. Example: 'url: The wiki page URL (format: https://wiki.yandex.../pages/<id>) or numeric page ID (1-9999999). Page IDs are resolved to their canonical URLs.'
• Add min/max constraints to 'limit' parameter in page_search: 'limit: Maximum results to return. Must be 1-100 (default: 10). Capped at 100 to prevent response bloat.'
• Add format/constraint to 'content' parameter: 'content: Page body in Yandex Wiki markup (see wiki.yandex.com/markup). Supports headers, bold/italic, lists, tables, code blocks. Max 1MB.'
• Implement error recovery guidance. Example for page_update: 'If update fails with 409 Conflict, the page was modified by another user. Call page_get() to fetch latest content and retry with merged changes.'
• Add confirmation mechanism for page_delete. Implement a two-step pattern: delete_page_preview (shows what will be deleted, requires approval) → delete_page_confirm (executes only with preview ID).
• Clarify page_search vs page_get in descriptions. Example: 'page_search: Find pages by keyword across all wiki content (full-text search). Returns summaries. Use this for discovery.' and 'page_get: Retrieve full page content by exact URL or ID. Use after page_search to get complete text.'
• Add validation constraints to authentication/credentials. Ensure Yandex API tokens are injected server-side via environment variables, not exposed as tool parameters.

Inferred effective spec: 2025-06-18+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry