Feature request: "Yes, but..." during permissions prompts
Status Fixed / completed
Maintainer reply ✓ Yes — bcherny
Workaround ✓ Mentioned in description ↑
Activity 8 comments · opened May 11, 2025 · closed May 13, 2025
💡 Likely answer: A maintainer (bcherny, collaborator)
responded on this thread — see the highlighted reply below.
I frequently find myself in a situation where Claude proposes a file edit or a command execution that is very, very close to what I want it to do. So close, in fact, that I'd rather it execute that change and then take some direction from me to make a targeted follow-up change.
Right now, my options are:
Yes. Do it.
Yes. And just keep going without checking in.
No. Don't do it. Stop.
I'd really love a fourth option:
Yes. But then stop for a prompt.
I can work around this by hitting "Yes. Do it" and then hitting escape once Claude has started in on its next thing, but that often ends up with polluted context and feels clunky.
8 Comments
seconded
Same as my very common usage pattern. The wording I used the other way was "Yes, but tell Claude how to improve it". https://github.com/anthropics/claude-code/issues/296#issuecomment-2868626645
100% my main interaction with Claude
You can now do this! Say yes, then send a prompt and send it. This will steer Claude in real-time -- Claude will incorporate your new prompt into its work as it goes.
<3 <3 <3
Amazing. Thank you so much.
@bcherny - I'm not sure if I'm interpreting things correctly, but I'm not seeing the real-time steering. When I do this, the prompt I send sits as a queued prompt through multiple interactions. (And this is still current with 1.0)
Yeah this should be reopened.
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.