You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Feature Request] Add Label Option to Comment dialog (combine) #932
I would be nice if the Label feature integrated in to the comment feature so we could make a comment as a question or some other thing. This combine with tweaking the wording of the default agent instructions when feedback is sent would make for more precise and streamlined communication with between the user and the agent.
Some times I want a answer, before it updated the plan, I am just asking "WHAT if questions"
So part of the fix to my problem of OPus 4.8, always interpreting comments as a direct request to reword, was simply changing feedback instructions from the plannotator provided by the agent. However, there is still a UX gap here. (atleast I feel like there is)
Combining the Label feature (which provides some instructions to the agent for each one) and the comment feature would be a nice UX improvement. Then we could replace some of the label types with out own flavor to be clearer about what we are wanting.
Agents/Models very easily conflate types of comments. Are they
a direct request to change the wording of a selected section
a comment about the agent seeming to misunderstand something about the subject as a whole (meta)
are they a comment about the framing?
are they simply a user asking a question about something, and not wanting anything changed?
Activity
BenNewman100 commentedon Jul 20, 2026
So part of the fix to my problem of OPus 4.8, always interpreting comments as a direct request to reword, was simply changing feedback instructions from the plannotator provided by the agent. However, there is still a UX gap here. (atleast I feel like there is)
Combining the Label feature (which provides some instructions to the agent for each one) and the comment feature would be a nice UX improvement. Then we could replace some of the label types with out own flavor to be clearer about what we are wanting.
Agents/Models very easily conflate types of comments. Are they
@backnotprop