Hookify - Hook Management

Manage hookify rules that prevent unwanted Claude behaviors in the Wythm project.

Usage

• /hookify help - Show documentation on how hookify works
• /hookify list - List all configured rules
• /hookify configure - Enable/disable rules interactively
• /hookify <behavior> - Create rule for specific behavior
• /hookify - Analyze conversation to find behaviors to prevent

Parse $ARGUMENTS to determine which sub-command to run.

Reference

For complete rule writing guide, syntax, examples, and testing instructions, see references/rule-guide.md.

Sub-command: help

Show comprehensive hookify documentation.

Read and present the content from references/rule-guide.md.

Explain:

1. How hookify works (hook events, rule loading)
2. Configuration file format (frontmatter + message)
3. Pattern syntax (Python regex)
4. Available commands
5. Example rules

Key points to cover:

• Rule files location: .claude/hooks/hookify/rules/*.local.md
• Events: bash, file, stop, prompt, all
• Actions: warn (default), block
• Changes take effect immediately

Sub-command: list

List all configured hookify rules.

Steps:

1. Glob for .claude/hooks/hookify/rules/*.local.md
2. Read each file, extract frontmatter (name, enabled, event, pattern)
3. Present as table:

4. Show totals (X enabled, Y disabled)
5. Add footer with edit/disable instructions

If no rules found, suggest /hookify <behavior> to create first rule.

Sub-command: configure

Enable/disable rules interactively.

Steps:

1. Glob for .claude/hooks/hookify/rules/*.local.md
2. Read each file, get name and enabled state
3. Use AskUserQuestion with multiSelect to let user pick rules to toggle
4. For each selected: toggle enabled: true <-> enabled: false using Edit tool
5. Confirm changes

Sub-command: create (default when $ARGUMENTS is not a sub-command)

Create hook rules from behavior description or conversation analysis.

If $ARGUMENTS contains behavior description:

• Parse the specific instruction
• Still scan recent conversation for examples

If $ARGUMENTS is empty:

• Launch conversation-analyzer agent to find problematic behaviors
• Agent scans user prompts for frustration signals

Creation steps:

1. Gather behaviors (from args or agent)
2. Present findings with AskUserQuestion:
   • Which behaviors to hookify? (multiSelect)
   • Warn or block? (for each selected)
3. Generate rule files in .claude/hooks/hookify/rules/{name}.local.md
4. Confirm creation, remind no restart needed

Rule file format:

---
name: {kebab-case-name}
enabled: true
event: {bash|file|stop|prompt|all}
conditions:
  - field: {command|new_text|file_path|user_prompt}
    operator: {regex_match|contains|equals|starts_with|ends_with|not_contains}
    pattern: '{regex pattern}'
action: {warn|block}
---

{Message to show Claude when rule triggers}

Quick Reference

Events:

• bash: Bash tool commands
• file: Edit/Write tools
• stop: When agent wants to stop
• prompt: User prompt submission
• all: All events

Regex tips:

• rm\s+-rf - rm -rf
• console\.log\( - console.log()
• \.env$ - .env files
• chmod\s+777 - chmod 777

File location: .claude/hooks/hookify/rules/*.local.md

No restart needed - rules apply immediately!