Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

Version: 2026.3.31

Description:
When the agent executes shell commands through remote channels (WeChat/WebChat), the exec approval prompt appears. Clicking allow-always results in the same command requiring approval again on the next execution, behaving identically to allow-once.

agent 通过远程渠道（微信/WebChat）执行 shell 命令时弹出审批，点击 allow-always 后，同一命令下次执行时仍需审批，表现和 allow-once 一致。

Expected behavior:
Clicking allow-always should permanently approve the command (at least until Gateway restart), without repeatedly showing the approval prompt.

点击 allow-always 后，相同命令应被永久放行（至少在 Gateway 重启前），不再重复弹出审批请求。

Steps to reproduce:

Ask the agent to execute any shell command via WeChat/WebChat
Click allow-always in the approval prompt
Ask the agent to execute the same command again
Approval prompt still appears (behaves same as allow-once)

Steps to reproduce

1. From a remote channel (WeChat/WebChat), ask the agent to execute a shell command (e.g., openclaw --version)
2. The exec approval prompt appears with options: allow-once / allow-always / deny
3. Click allow-always
4. Ask the agent to execute the same command again
5. Approval prompt still appears — behaves same as allow-once

Expected behavior

After clicking allow-always, the same command should be permanently approved (at least until Gateway restart) and no longer trigger the approval prompt.

Actual behavior

Clicking allow-always only approves the current execution, same as allow-once. The next execution of the same command still shows the approval prompt.

OpenClaw version

2026.3.31

Operating system

Ubuntu 24.04

Install method

npm global

Model

openrouter/xiaomi/mimo-v2-pro

Provider / routing chain

Channel: openclaw-weixin (WeChat) > Agent: main > Model: openrouter/xiaomi/mimo-v2-pro

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response