Agents now use the interactive question interface throughout discovery, brief, and blueprint flows whenever they need more context before continuing. Rather than guessing or stalling, the agent pauses and presents a focused set of options you can click through or answer with free text.
An intro message is written to the thread each time the agent enters a question flow, so the conversation history reflects what was collected and why.
When your answer to a question is ambiguous or too brief, the agent now reprompts you with a new set of structured questions rather than proceeding on a guess. This clarification loop is capped to prevent the agent from cycling indefinitely — once the limit is reached it moves forward with the best available interpretation.
Attaching a document to a conversation now gives the agent access to its full content. Previously, the agent could see that a file was attached but could not read it.