[DOCS] Missing `model` parameter in Prompt-Based Hooks documentation
Documentation Type
Missing documentation (feature not documented)
Documentation Location
https://code.claude.com/docs/en/hooks
Section/Topic
Prompt-Based Hooks > Configuration section
Current Documentation
The documentation currently lists only three fields for configuring prompt-based hooks:
Fields:type: Must be"prompt"prompt: The prompt text to send to the LLM Use$ARGUMENTSas a placeholder for the hook input JSON If$ARGUMENTSis not present, input JSON is appended to the prompt *timeout: (Optional) Timeout in seconds (default: 30 seconds)
What's Wrong or Missing?
The documentation is missing the model parameter which was introduced in version 2.0.41. This parameter allows users to specify a custom model (e.g., switching from the default Haiku to Sonnet) for evaluating the hook logic. Without this documentation, users are unaware they can control the reasoning capability and cost/latency trade-offs for their intelligent hooks.
Suggested Improvement
Update the Configuration list under Prompt-Based Hooks to include the model field:
Fields:
type: Must be"prompt"prompt: The prompt text to send to the LLM...model: (Optional) The model alias to use for evaluation (e.g.,sonnet,haiku,opus). Defaults tohaiku.timeout: (Optional) Timeout in seconds (default: 30 seconds)
Additionally, update the example JSON to reflect this capability:
{
"hooks": {
"Stop": [
{
"hooks": [
{
"type": "prompt",
"model": "sonnet",
"prompt": "Evaluate if Claude should stop: $ARGUMENTS. Check if all tasks are complete."
}
]
}
]
}
}
Impact
High - Prevents users from using a feature
Additional Context
This feature was officially noted in the Claude Code version 2.0.41 release notes:
"Added model parameter to prompt-based stop hooks, allowing users to specify a custom model for hook evaluation."
Current documentation pages for reference:
- https://code.claude.com/docs/en/hooks
- https://code.claude.com/docs/en/model-config (for valid model aliases)
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗