Feature: Prompt injection hooks for model

## Description

Add support for hooks that can inject content into prompts before they're sent to the model.

## Use Case

Currently, hooks like user prompt submit hooks can only execute scripts. This feature request is for hooks that can:
- Inject custom commands into user prompts before submission
- Not just execute side-effect scripts, but actively modify/enhance the user's prompt
- Inject context, instructions, or commands into the model request

## Example

A hook that could automatically inject environment-specific instructions or context into every prompt sent to the model, without requiring the user to manually type it each time.

## Benefits

- Automate prompt enhancement and context injection
- Create reusable prompt templates and patterns
- Enforce prompt standards across teams
- Enhance model responses with injected context