Four tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. request_review has the most complete definition with 5 parameters and descriptions, but lacks output schema documentation. get_session_list and get_session_details have minimal parameter documentation. resolve_session lacks guidance on expected feedback format. No tool includes error handling guidance, recovery paths, or output structure documentation. Descriptions are present but generic (10-50 chars for some). No tool annotations (readOnlyHint/destructiveHint) despite clear semantic differences. Parameter descriptions are sparse or missing entirely.

• Document output schema for each tool. E.g., request_review should specify: returns {sessionId: string, status: string, createdAt: ISO8601, ...}. get_session_list should return {sessions: [{id, status, createdAt, ...}], total: number}.
• Add format/constraint descriptions to all parameters. E.g., 'sessionId: UUID v4 format' or 'feedback: 1-5000 characters, plain text or markdown'. Replace bare 'string' types with actionable guidance.
• Add error handling guidance to resolve_session and get_session_details. E.g., 'If sessionId not found, returns 404 with available session IDs. If feedback is empty, returns 400 with required format.'
• Implement tool annotations: add destructiveHint=true to resolve_session, readOnlyHint=true to get_session_* tools. This enables agents to reason about safety.
• Add pagination to get_session_list: accept limit (1-100, default 20) and offset/cursor parameters. Return total count and next_cursor. Document result cap in description.

Inferred effective spec: <=2025-11-25.

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