Security hooks flag hardcoded secrets, eval patterns, SQL injection vectors
Under the hood
How it works
π
Agent writes code
Claude Code generates TypeScript, runs linters, commits changes.
πͺ
Hooks intercept at boundaries
20 automated shell scripts run on Read/Edit/Write operationsβoutside the LLM, zero token cost, impossible to bypass.
β
Enforcement or rejection
Type safety checks, file size limits, security pattern detection, and decomposition rules either auto-fix violations or block commits.
What it does
Core capabilities
β‘
Fast
Optimized for speed and efficiency.
π―
Reliable
Built to work consistently.
π
Integrates
Works with your existing tools.
I got tired of Claude shipping 600-line components with `as any` everywhere. These hooks make it impossibleβthe agent literally cannot commit code that violates them.
arc-web/claude-plugins
Codebase
Built with care
π»
TypeScript
Built in TypeScript for type safety and reliability.
ποΈ
Architecture
Configuration Tool
π¦
Key deps
Claude Code, composure, sentinel
π
Open Source (MIT / PolyForm)
Fork it, extend it, ship it. It's yours.
Built by
Mike Ensor Arc Web
AI automation, marketing technology, building tools that work.
Mike builds open-source AI tools for agencies, marketers, and operators who are done waiting for software to catch up. All Arc Web projects are open source by default β fork them, extend them, ship them.