Hookify
148,649 starsOfficial Anthropic plugin for Claude Code: create custom hooks in natural language — /hookify analyzes your description (or uses the conversation-analyzer agent to mine unwanted behavior patterns from the conversation) and generates Markdown + YAML-frontmatter rule files (e.g. .claude/hookify.warn-rm.local.md) that regex-match tool usage and show warnings on hit; effective on the very next tool call with no restart, plus list/configure/help subcommands for rule management.
- First published：2025-11-17
- data as of：
- 2026-09-30
Host
Reads your session conversation history (conversation-analyzer mines it for behavior patterns) and writes rule files under .claude/ — files that then change how Claude Code intercepts tool calls, i.e. it helps configure your permission gates.
The rules it generates are loadables for Claude Code's hook machinery (written under .claude/, bound to Claude Code's tool-use events). The hook event model is host-proprietary: on OpenClaw or Hermes there are neither the same event points nor a loader that reads these rule files.
Risk
Rules auto-generated from conversation analysis become part of your permission gates: an over-broad regex blocks legitimate work (or breeds alert fatigue), an over-narrow one is theater — and generation is model-driven, so every rule needs line-by-line human review. Handing gate configuration to a model is a variant of the "a prompt is not control" warning in the permission-gate glossary entry.
Install prompt
在 Claude Code 的插件管理介面中，從官方 marketplace（claude-plugins-official）安裝 hookify；然後直接用自然語言描述你想攔截的行為（例如「當我使用 rm -rf 時警告我」），它會生成對應規則檔並立即生效。Paste this to your agent. It deliberately contains no tool subcommands — check the tool's official docs for commands.