Currently when you add output from a command to the chat, the request is handled by the LLM suggesting changes and then you are prompted to make the edits, but it allows you to provide additional instructions at that point. However, when using the architect with a simple chat message, the prompt to edit files only accepts y/n. It would be ideal to be able to: also be able to provide additional instructions that get sent to the editor model along with the architect's instructions.
Use case: you've asked the architect to design an implementation for xyz feature, most of it is good except one point and you want to tell the editor not to do that particular thing but everything else, or slightly modify the architects instructions without going through another turn of chat with the architect (loss of time and money).
Issue
Currently when you add output from a command to the chat, the request is handled by the LLM suggesting changes and then you are prompted to make the edits, but it allows you to provide additional instructions at that point. However, when using the architect with a simple chat message, the prompt to edit files only accepts y/n. It would be ideal to be able to: also be able to provide additional instructions that get sent to the editor model along with the architect's instructions.
Use case: you've asked the architect to design an implementation for xyz feature, most of it is good except one point and you want to tell the editor not to do that particular thing but everything else, or slightly modify the architects instructions without going through another turn of chat with the architect (loss of time and money).
Version and model info
latest o1-mini (architect), sonnet (editor)