Writing Hookify Rules

Custom AI guardrails are defined using a declarative YAML frontmatter format within Markdown files, specifying event triggers like file operations and pattern matching to block sensitive data. You create rules in a .local.md file within the .claude directory to enforce these workflow behaviors.

Custom AI behavior rules can be triggered on bash commands, file operations, agent stop events, and user prompts. This event handling capability allows you to apply conditional logic and pattern matching across various AI interaction stages.

To block API keys from .env files, set up a custom rule by defining a file event that matches the pattern API_KEY and blocks any edits containing it. This proactive rule is loaded dynamically to prevent accidental sensitive information logging during AI interactions.

Yes, AI prompt enforcement supports pattern matching and conditional logic for user prompt events. You can define declarative rules in YAML frontmatter to manage and constrain AI behavior dynamically based on the matched input patterns.

No special dependencies are required; declarative AI rules are created and managed using YAML frontmatter within standard Markdown files. This approach facilitates proactive AI safety and workflow enforcement through dynamic rule loading without external libraries.

The best way to manage proactive AI safety rules for bash commands is by defining declarative rules with bash event triggers and pattern matching in Markdown files. This allows dynamic loading and conditional logic to handle specific command patterns appropriately.