Server exposes 11 tools with basic schemas and descriptions. All tools have input schemas with type definitions and descriptions present. However, descriptions are minimal (10-50 chars), lacking context on WHEN to use each tool, prerequisites, or recovery guidance. Parameter descriptions are present but terse. No output schemas documented. Error handling returns JSON errors but lacks actionable recovery guidance. Tools like start_agent and start_orchestrator expose dangerous capabilities (dangerousMode flag) without permission gates or confirmation patterns. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Composition is reasonable, tools chain via projectId/taskId, but agent-related tools (start_agent, write_to_agent, read_agent_output) lack clear sequencing guidance.

• Expand tool descriptions to 50-150 chars. Include WHEN to use (e.g., 'list_tasks: Fetch all tasks in a project. Use before create_task to check for duplicates or before get_board_state to see Kanban layout.')
• Document output schemas for all tools. E.g., 'Returns {tasks: [{id, title, status, priority, ...}], total: number}' so LLMs know what fields to extract.
• Add tool annotations: mark delete_task, start_agent, start_orchestrator as destructiveHint=true; mark list_tasks, read_agent_output, get_board_state as readOnlyHint=true.
• Implement confirmation pattern for destructive ops: add a 'confirm' parameter or require a separate confirm_delete_task tool that must be called before delete_task succeeds.
• Enhance error responses with recovery guidance: 'Task not found (ID: xyz). Try list_tasks(projectId) to see available tasks.' Include error classification (retryable, user-fixable, fatal).