Standalone MCP server bridging Claude Code (or any MCP client) to Cognitive Core Rooms.
Single tool 'reply' has critical definition gaps. Description is in Russian and lacks English context for LLM selection. Input schema present but minimal (2 params, both strings). No output schema documented. No error handling guidance. Tool name 'reply' is generic and does not follow verb_noun pattern (should be 'send_message' or 'post_reply'). Parameter descriptions reference external context ('room_id брать из meta') rather than being self-contained. No indication of idempotency, retry safety, or side effects. Tool appears to be a write operation but lacks confirmation/dry-run pattern.
Ответить в комнату Cognitive Core. room_id брать из meta пришедшего события (отвечать нужно в ТУ ЖЕ комнату, откуда пришло сообщение).
Tool name 'reply' is generic and does not follow verb_noun convention. Should be 'send_message', 'post_reply', or 'send_room_message' to clearly signal the action.
Description is in Russian only ('Ответить в комнату...') and references external context ('room_id брать из meta пришедшего события'). LLMs cannot reliably select or invoke this tool without English descriptions and self-contained parameter guidance.
No output schema documented. LLMs cannot plan downstream actions or extract response data. Must specify return type (e.g., success boolean, message_id, timestamp).
No error handling guidance. Tool is marked WRITE risk but lacks confirmation pattern, dry-run option, or recovery instructions. Agents cannot distinguish retryable vs. fatal failures.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
Parameter 'room_id' description references external meta context instead of being self-contained. Should state: 'UUID of the room where the message will be posted. Typically extracted from the incoming event metadata.'