writing-hookify-rules

Hookify rules guard prompts by defining frontmatter fields like name, enabled, event, and pattern, combined with a Markdown body for user-facing warning messages. You create local hook rule files that intercept and block sensitive prompt interactions.

Guardrails for Claude interactions are automated safety checks that enforce consistent agent behavior. Hook patterns define trigger conditions using fields like event and pattern, while the Markdown body delivers corrective messages when a prompt matches restricted content.

Multi-condition rules in Hookify gate triggers based on file_path, new_text, old_text, and other fields. You define these conditions within the rule frontmatter to precisely control when guard patterns activate during development and production workflows.

The best way to structure a hook rule file for prompt security is using frontmatter with name, enabled, event, and pattern fields, followed by a Markdown body. This format supports automated policy enforcement and clear user-facing messages.

You need Markdown to author the body of Hookify rules, as it provides the user-facing message displayed when a policy violation occurs. The frontmatter handles machine-readable configuration, while the Markdown body handles human-readable guidance.

Hookify rules may not trigger if the enabled field is false or if the pattern field does not accurately match the incoming prompt text. Multi-condition rules also require exact matches on gating fields like file_path or new_text to activate.