Subagent does not receive user rejection reason from bash permission prompt

[Onehand-Coding]

Description

Steps to reproduce:

1. Configure a primary agent with bash: "deny" and a subagent executor with bash access
2. Trigger a bash command via the primary agent so it delegates to the subagent
3. When the UI permission prompt appears, reject the command and type a reason in the provided input field
4. Observe the subagent's response and subsequent behavior

Expected behavior:
The rejection reason typed in the UI prompt is injected into the subagent's
conversation context, allowing it to relay the reason back to the calling agent.
The calling agent receives the reason and adjusts its approach without retrying
the rejected command.

Actual behavior:
The rejection reason never reaches the subagent's context. The subagent only
sees a failed tool call with no explanation. The calling agent receives no reason,
treats the task as incomplete, and re-delegates the same command to the subagent
creating an infinite loop until the user manually interrupts the session.

Workaround:
Type a placeholder in the UI reason field, then immediately interrupt the session
before the subagent processes the rejection. Then type the actual reason as a
regular chat message in the main conversation so it reaches the primary agent's
context directly.

Plugins

superpowers

OpenCode version

1.15.7

Steps to reproduce

See Description above

Screenshot and/or share link

No response

Operating System

Fedora 43 XFCE, Linux (ThinkPad T480)

Terminal

xfce4-terminal