hook-development

Claude Code plugin hooks validation combines LLM-driven reasoning with deterministic shell checks to block unsafe tool execution. You define PreToolUse and PostToolUse events that intercept destructive bash commands and validate file writes before execution.

Claude Code command hooks are event-driven scripts triggered at lifecycle stages like SessionStart, UserPromptSubmit, and Stop. You need them to automate workflows, load project context, and enforce security policies when agents interact with tools.

Prompt-based hooks can be used alongside command hooks for layered validation in Claude Code. This combines LLM reasoning for complex context checks with deterministic shell scripts for strict security enforcement and structured JSON exit codes.

Testing and linting Claude Code hook scripts requires running included validation scripts against hooks.json configurations. These scripts check structured JSON outputs, verify timeout controls, and lint implementations for path safety and security best practices.

Claude Code hook automation requires the jq and timeout dependencies. You use environment variables like ${CLAUDE_PLUGIN_ROOT} for path resolution and $CLAUDE_ENV_FILE for environment persistence across hook events.

PreToolUse hooks fail to block destructive commands when they lack proper structured JSON outputs or correct exit codes. Ensure your command hooks implement timeout controls and input validation to enforce safe tool usage decisions properly.