Let user describe work in natural language, Ava creates the task.
User: 'I need a function to validate email addresses' Ava: 'I'll create a task for that. Which namespace? (e.g., Omni/Utils)' User: 'Omni/Validation' Ava: 'Created t-290: Add email validation function Namespace: Omni/Validation [Start now] [Edit first] [Cancel]'
This requires LLM integration for NL understanding, so more complex than simple command parsing.
Closing as 'already works'. The agent already has create_task and create_task_from_description tools. When a user says 'fix the login bug', the LLM understands the intent and calls the tool. The 'natural language' part is already handled by the LLM itself. Action buttons would be a nice UX enhancement but not essential for the core functionality.