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
I have verified this feature I'm about to request hasn't been suggested before.
Describe the enhancement you want to request
Hi,
I'd like to request that OpenCode's yolo mode isn't just blindly accepting everything, but asks a decision model like Jev about if some tool call/bash run is safe. ofc it doesn't guarantee perfect safety but will stop stuff like an LLM potentially running something like rm -rf /*
OpenCode Zen already includes Jev (wen Go???), so it would be a usecase for it in OpenCode.
Thanks for the request, and for sharing your plugin! I think this is a duplicate of #37564, which asks for the same thing: when auto-approve is on, a model checks each tool call first, and unsafe ones fall back to asking you. Someone there has already suggested using the Jev/Kev/Leya models, and a draft implementation (#39015) is linked in that thread.
Related issues, ordered from most to least likely to match:
It would help to add your use case and a link to your plugin on #37564. The bot will close this issue as a duplicate in 1 day unless you reply explaining how your request is different.
We didn't hear back, so I'm closing this as a duplicate of #37564. That issue asks for the same thing: when auto-approve is on, a model checks each tool call and unsafe ones fall back to asking you. Someone there has already suggested the Jev/Kev/Leya models. Please follow and comment on #37564; adding your use case and a link to your plugin there would help. If your request is actually different, reply here and we can take another look.
Activity
opencode-agent commentedon Oct 15, 2026
Thanks for the request, and for sharing your plugin! I think this is a duplicate of #37564, which asks for the same thing: when auto-approve is on, a model checks each tool call first, and unsafe ones fall back to asking you. Someone there has already suggested using the Jev/Kev/Leya models, and a draft implementation (#39015) is linked in that thread.
Related issues, ordered from most to least likely to match:
It would help to add your use case and a link to your plugin on #37564. The bot will close this issue as a duplicate in 1 day unless you reply explaining how your request is different.
surfaceflinger commentedon Oct 15, 2026
oh well
opencode-agent commentedon Oct 16, 2026
We didn't hear back, so I'm closing this as a duplicate of #37564. That issue asks for the same thing: when auto-approve is on, a model checks each tool call and unsafe ones fall back to asking you. Someone there has already suggested the Jev/Kev/Leya models. Please follow and comment on #37564; adding your use case and a link to your plugin there would help. If your request is actually different, reply here and we can take another look.