Claude Code Hook Development FAQs

Question 1

Can I run multiple hooks for the same event?

Accepted Answer

Yes, Claude Code executes all matching hooks for an event in parallel. They are designed to be independent and do not see each other's output.

Question 2

What are prompt-based hooks in Claude Code?

Accepted Answer

Prompt-based hooks use natural language instructions to let the LLM make context-aware decisions, such as evaluating if a specific tool use is safe or appropriate based on the current transcript.

Question 3

What is the difference between the hooks.json and settings.json format?

Accepted Answer

Plugin hooks in hooks/hooks.json require a 'hooks' wrapper object and can include an optional description, while user settings in .claude/settings.json use a direct top-level format for events.

Question 4

How do I ensure my plugin hooks are portable?

Accepted Answer

Always use the ${CLAUDE_PLUGIN_ROOT} environment variable in your hooks configuration to reference scripts within your plugin directory, avoiding hardcoded absolute paths.

Question 5

How do I block a tool execution using a hook?

Accepted Answer

In a PreToolUse hook, you can return a JSON object with 'permissionDecision': 'deny' or, in a command hook, exit with code 2 to feed an error back to Claude and stop execution.

Claude Code Hook Development

Key Features

Use Cases

Claude Code Hook Development

Key Features

Use Cases