Bug: 使用 switch 禁用插件后仍会执行部分逻辑

[MirrorCY]

Describe the bug

使用 switch 禁用一个插件后，如果该插件有介于 apply() 和 ctx.command() 之间的逻辑，仍会在匹配到 command 关键词时被执行。例如 novelai-bot 会在定义命令范围外进行参数检查，使用 switch novelai 禁用此插件后输入不合法的分辨率仍会报错。

Steps to reproduce

• 全新安装 koishi bp
• 安装 switch 和 novleai 插件并完成相关配置
• 进入 QQ 群输入 switch novelai
• 输入 nai -r 9999x9999 koishi
• 查看返回的报错

Expected behavior

当插件的所有已注册指令都被禁用时，不应该触发该插件其余逻辑；
或当一个已注册指令被禁用时，不应该在发现被禁用指令关键词时调用该插件。

Screenshots

Versions

• Koishi version: 4.10.5

Additional context

No response

[MaikoTan]

https://github.com/koishijs/koishi/blob/219ab235ef2be6e9aebb62a6a4dd8286faf9e292/plugins/a11y/switch/src/index.ts#L38-L50

So you could see here we block an invoke of the command using the command/execute event, which is triggered before executing the cmd.action(callback) part. That's why the validations or hooks still running even we disable it with the switch command.
But I don't know if there is an event that triggered AFTER parsing a command. (Because if you register a middleware BEFORE the parsing part, you could only parse the messages by yourself, generally we don't want to make a plugin complicated and unmaintainable.)

[MaikoTan]

cc @shigma

[Hieuzest]

closed via 8f353a9