claude-code-hooks-daemon
Over a hundred handlers enforcing deterministic guardrails on AI coding agents: the policy layer
A daemon-based hooks system for Claude Code: well over a hundred handler implementations spread across pre-tool-use, post-tool-use, session-start, and stop events, covering everything from blocking destructive git commands to catching security antipatterns to enforcing project-specific plan workflows.
The test tree dwarfs the source tree: tens of thousands of lines of source, with several times as many lines of tests. Well over a hundred releases since the first alpha, with continuous active development.
This site is a dogfooding consumer: this exact repository runs the daemon as part of its own development workflow, including several of the handler guardrails described in this project catalogue.
Pipeline
- Pre-tool-use guardrails
- Blocks destructive git commands, security antipatterns, error-suppression patterns, and sensitive-content leaks before a tool call executes.
- Post-tool-use & session hooks
- Tracks backgrounded processes, audits hook registration, checks plan-document QA, and surfaces drift reports at session start.
- Project-level extensibility
- Consuming repos can add their own handlers without forking the daemon itself.
- Well over a hundred releases
- Continuous delivery from a 2025-01 alpha to the current version. Actively maintained, not a one-off script.