evolver
Claude hook for agent self-evolution — turns ad hoc prompt tweaks into auditable, reusable evolution assets
Hooks let you run your own scripts on Claude Code lifecycle events. Add permission gates, audit logs, notifications, or completely custom behavior without touching the agentic loop itself.
Claude hook for agent self-evolution — turns ad hoc prompt tweaks into auditable, reusable evolution assets
Desktop notifications Claude hook — alerts when Claude needs input or completes tasks, with click-to-jump
Claude hook for code quality — performs project-specific TypeScript, ESLint, and Prettier checks with auto-fixes
Python SDK for Claude Code hooks — simplifies building, parses JSON, and validates all 9 hook types
Prompt injection scanner Claude hook — detects injection attacks, secrets, and data exfiltration in tool I/O
Hooks are user-defined scripts that run on Claude Code lifecycle events — PreToolUse, PostToolUse, Stop, Notification, and others. They let you add permission gates, audit logs, or custom behavior without modifying the agent loop.
Hooks receive event metadata via stdin and can approve, deny, or modify tool calls. Common uses: blocking writes to sensitive files, logging tool usage, sending notifications when Claude is waiting for input.
Add an entry under hooks in your settings.json with the matching event and the command to run. The command is invoked synchronously and can return a JSON decision.