Hook Create

Name: hook-create
Rating: 92
Author: hyxklee

Create hooks for deterministic automation that LLMs might forget.

When to Create a Hook

•Action must ALWAYS happen (format, lint, test)
•Need to block dangerous operations
•Want to inject context automatically
•Workflow integration with external tools

Key Insight: LLMs can forget; hooks never forget.

Hook Events

Event	When	Can Block?
`PreToolUse`	Before tool executes	Yes
`PostToolUse`	After tool succeeds	No
`UserPromptSubmit`	Before Claude processes	Yes
`SessionStart`	Session begins	No
`Stop`	Claude finishes responding	Yes

Configuration Location

json

// .claude/settings.json (project)
// ~/.claude/settings.json (personal)
{
  "hooks": {
    "{EventName}": [
      {
        "matcher": "{regex}",
        "hooks": [
          {
            "type": "command",
            "command": "{script}"
          }
        ]
      }
    ]
  }
}

Exit Codes

Code	Meaning	Behavior
0	Success	Action proceeds
2	Block	Action blocked, stderr to Claude
Other	Warning	Action proceeds, stderr in verbose

Example: Block Dangerous Commands

.claude/settings.json:

json

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          {
            "type": "command",
            "command": "\"$CLAUDE_PROJECT_DIR\"/.claude/hooks/block-dangerous.sh"
          }
        ]
      }
    ]
  }
}

.claude/hooks/block-dangerous.sh:

bash

#!/bin/bash
COMMAND=$(jq -r '.tool_input.command' < /dev/stdin)

if echo "$COMMAND" | grep -qE 'rm -rf|sudo|chmod 777'; then
  echo "Blocked: Dangerous command" >&2
  exit 2
fi

exit 0

Example: Auto-Format After Edit

json

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Write|Edit",
        "hooks": [
          {
            "type": "command",
            "command": "npx prettier --write \"$CLAUDE_TOOL_INPUT_FILE_PATH\""
          }
        ]
      }
    ]
  }
}

Hook Types

Type	Use Case
`command`	Shell script execution
`prompt`	LLM evaluation (yes/no)
`agent`	Multi-turn verification

Checklist

Before creating:

• Must this ALWAYS happen?
• Is it deterministic (not judgment-based)?
• Keep it fast (especially SessionStart)
• Script handles errors gracefully?

Reference: @claude/hooks/README.md for detailed patterns.