Agent-task-board defines 6 tools with complete input schemas and descriptions. Naming follows verb_noun convention (add_task, list_tasks, claim_next, report_result, move_task, get_board). Descriptions are present but brief (avg ~80 chars, below the 194-char baseline). All parameters have type definitions and most have descriptions, but several lack detail on constraints, valid ranges, or error recovery guidance. Output schemas are not documented, LLMs cannot predict response structure. Error handling returns generic messages without recovery hints. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear read/write semantics.

• Add outputSchema to each tool definition documenting the response structure. E.g., add_task should declare: {type: 'object', properties: {queued: {type: 'string'}, title: {type: 'string'}, agent: {type: 'string'}, status: {enum: ['queued', 'running', 'review', 'done']}}}
• Add tool annotations: destructiveHint: true for add_task, claim_next, report_result, move_task; readOnlyHint: true for list_tasks, get_board. This signals to LLMs which operations are safe to retry.
• Expand descriptions to 100 - 200 chars. E.g., move_task: 'Move a task to a different lane (queued → running → review → done). Use this to manually advance or revert task status. Returns the updated task with new status.'
• Add constraint details to parameters. E.g., agent: 'Target agent label (e.g., Claude Code, Cursor, or custom agent name)'; tags: 'Array of string tags (max 10 items, each 1-50 chars)'; worker: 'Worker identifier claiming the task (default: mcp)'.
• Enhance error responses with recovery hints. E.g., if claim_next returns {claimed: null}, add guidance: 'No queued tasks found. Try list_tasks(status: queued) to check available work, or add_task() to queue new work.'