Describe the feature or problem you'd like to solve
In claude-code, you can ask the agent anytime, regardless of actively running/executing/sleeping or just idle, using his accumulated context to instantly ask a question about something, without bloating/corrupting/writing to the context/session.
Proposed solution
/btw <question prompt> command should allow to ask simple questions as a quick sidecommand, thinking/reasoning supported, without writing to the session
Example prompts or workflows
Example scenario (happening frequently)
- Agent is working on a feature, kicking off a training run
- Sleeping the console output for 600 secs
- I am contemplating whether to stop him because something was misconfigured, and i cant ask him about it either
Should consume only max a half request, since no agentic work can be executed, or use a free/smaller/configurable model by default, ingesting the same context
Additional context
No response
Describe the feature or problem you'd like to solve
In claude-code, you can ask the agent anytime, regardless of actively running/executing/sleeping or just idle, using his accumulated context to instantly ask a question about something, without bloating/corrupting/writing to the context/session.
Proposed solution
/btw <question prompt>command should allow to ask simple questions as a quick sidecommand, thinking/reasoning supported, without writing to the sessionExample prompts or workflows
Example scenario (happening frequently)
Should consume only max a half request, since no agentic work can be executed, or use a free/smaller/configurable model by default, ingesting the same context
Additional context
No response