This server has 13 tools with basic naming (verb_noun pattern mostly followed) but critical gaps in descriptions and schemas. Most tool descriptions are present but generic (10-50 chars), falling below the 50-200 char LLM-optimized baseline. Input schemas exist but lack depth: parameters have minimal descriptions, no enums for constrained values (e.g., 'mode', 'decision'), and no documented output schemas. Error handling is absent, no recovery guidance, no categorization of retryable vs fatal errors. The server manages cursor sessions but provides no guidance on when to use each tool or how they compose. Security concerns: no visible permission checks, audit logging, or input validation. Overall, definitions are skeletal, sufficient to invoke but insufficient for reliable LLM reasoning.

• Add output schemas to all tools. Document field names, types, and whether results are paginated. E.g., cursor_session_status should return {session_id, status, mode, turn_id, created_at}.
• Convert 'mode' and 'decision' parameters to enums with explicit allowed values. E.g., mode: enum ['agent', 'manual', ...], decision: enum ['accept', 'reject', 'allow-once', 'reject-once'].
• Expand descriptions to 50-150 chars. Include WHEN to call the tool and WHAT it returns. E.g., 'Resume a paused cursor session. Use after cursor_wait returns a pending event. Returns updated session status.'
• Add error handling guidance. Document common failures (session not found, invalid mode, timeout) and recovery steps. E.g., 'If session_id is invalid, call cursor_start_session to create a new one.'
• Document parameter dependencies. E.g., 'option_ids is required only if the turn has a question request. If the turn has a plan request, use cursor_answer_plan instead.'