feat: --review-output for blocking agent-in-the-loop reviews
I think it would be great if my skill could call hunk in a tmux popup, allow me to review it there, make my own annotations, reply to the llm annotations, then quit (which would ideally close the tmux popup) and return the annotations i made to the llm
(imeplemenation wise, i think i would expect the user to do the wiring of agent. tmux, named pipe, all that would be needed from hunk is the ability to add your own annotations and return them from stdout on exit
i personally want my agent to block while waiting for the output, but many may not want that behaviour)
if you implement persistent sessions you could also return the session id for the llm to read your annotations from
feat: --review-output for blocking agent-in-the-loop reviews
I think it would be great if my skill could call hunk in a tmux popup, allow me to review it there, make my own annotations, reply to the llm annotations, then quit (which would ideally close the tmux popup) and return the annotations i made to the llm
(imeplemenation wise, i think i would expect the user to do the wiring of agent. tmux, named pipe, all that would be needed from hunk is the ability to add your own annotations and return them from stdout on exit
i personally want my agent to block while waiting for the output, but many may not want that behaviour)
if you implement persistent sessions you could also return the session id for the llm to read your annotations from