[FEATURE:] Add configurable hooks for custom commands on task events (similar to Claude Code)

[Decade-qiu]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

I would like to request support for configurable hooks, similar to the Hooks feature in Claude Code: https://code.claude.com/docs/en/hooks.

Hooks would allow users to define custom commands/scripts that automatically execute at specific points during a task's lifecycle, such as:

• On successful task completion (e.g., run a notification script, trigger a deployment, or perform cleanup)
• On task error/failure (e.g., send an alert, log details to an external service, or attempt automatic recovery)
• When user confirmation is required (e.g., pause execution and run a custom command to prompt the user via another tool/integration)
• more ......

[github-actions]

This issue might be a duplicate of or closely related to existing issues. Please check:

• Hooks for commands (Plugin Commands) #2185: Hooks for commands (Plugin Commands) - An earlier request for hooks functionality focused on plugin command interception and lifecycle

While #2185 is more specific to plugin command hooks, #7640 appears to be a broader feature request for task lifecycle hooks (like Claude Code's implementation). These two issues could potentially be consolidated or addressed as part of a comprehensive hooks system.

[Decade-qiu]

I've implemented this feature using OpenCode's plugin system. My plugin opencode-message-notify (https://github.com/Decade-qiu/opencode-message-notify) provides hooks for:

• Task completion - Sends detailed Agent messages
• Permission requests - Notifies when confirmation is required
• Usage statistics - Tracks cost, tokens, cache usage
• iOS notifications - Via Bark app

Moreover, plugin docs lack detailed explanations of event types and input (output) types. I had to log events during development to understand what data was available. It would be helpful to document key events with their input/output types and field meanings more clearly.