task-intent-skill

What it does

Detects user intent from natural language related to Todo tasks (add, list, update, delete, complete, tag, sort), maps intents to MCP tools, handles multi-step workflows, and generates friendly confirmation messages for operations.

How it works

• Uses IntentDetector to classify messages into intents such as add_task, list_tasks, complete_task, delete_task, update_task, add_tag, list_by_tag, sort_tasks or general.
• For each detected intent, it calls corresponding handlers in ReasoningEngine (e.g., _handle_add_task, _handle_list_tasks, _handle_complete_task, _handle_delete_task, _handle_update_task).
• Each handler invokes MCP tools via methods like add_task, list_tasks, complete_task, etc., passing user_id and extracted parameters (title, description, task_id, status, etc.).
• Confirmation messages are generated using the CONFIRMATIONS mapping and can be formatted with format_confirmation or errors with format_error when needed.
• Example flows include: extracting title/description for add_task, extracting task_id for complete_task, deriving list filters (status, priority, tag) for list_tasks, and handling ambiguous matches by returning a prompt to disambiguate.

When to use it

• When parsing natural language commands for task operations
• When detecting user intent (add, list, update, delete, complete)
• When mapping intents to MCP tools
• When handling multi-step reasoning workflows
• When generating confirmation messages
• When managing ambiguous or error cases

What it can touch

• Tools: the skill references MCP tools via methods like add_task, list_tasks, complete_task, etc., and returns tool_calls with tool names and parameters.
• File/code structure: intents and handlers are wired through IntentDetector in app/agents/intent_detector.py and ReasoningEngine in app/agents/reasoning_engine.py, with confirmation logic in app/agents/confirmations.py.

Caveats

• The _handle_delete_task and _handle_update_task methods are placeholders with pass, indicating incomplete implementation for those multi-step flows.
• The skill relies on the MCP tools being available (e.g., add_task, list_tasks, complete_task) and the user_id being provided.
• Error and confirmation messages are defined in CONFIRMATIONS and ERROR_MESSAGES and will format accordingly if invoked.