Introduction

Dotboost is a drop-in Claude Code configuration for Dotkernel projects. It is not a Dotkernel application, a library, or something you composer require — there is no composer.json and no PHP source in this repository at all. Its entire payload is a .claude/ directory: settings, hooks, a status line, a review subagent, ten /dk-* slash commands and eighteen skills. You install it by copying that directory into the root of the Dotkernel project you're actually working on (API, Admin, Frontend, Light, Queue, or a project derived from one of them).

The problem it solves

Claude Code has no built-in knowledge of Dotkernel's conventions — handler naming, where an entity belongs (application module vs. the shared Core submodule), which PSR applies where, how authorization keys map to routes, or the difference between a HAL-based API response and a templated redirect. Without that knowledge, an agent working in a Dotkernel repo either asks too many questions or confidently produces code that doesn't match house style.

Dotboost closes that gap two ways:

• Skills teach the conventions themselves — naming, structure, validation, responses, security, testing, PSR standards — each loaded automatically when its description matches what you're doing.
• Hooks and permissions protect the repo while an agent works in it — blocking edits to vendor/, dependency manifests, migrations and secrets; blocking destructive or dependency-installing shell commands; and normalizing whitespace and line endings on every file Claude touches.

Who this is for

Anyone using Claude Code inside a Dotkernel repository — application developers, and anyone maintaining a project derived from one of the Dotkernel starter kits. It's equally useful whether you're building a new feature, reviewing a PR, or bootstrapping a fresh clone.

What's next

• Architecture Overview for how the pieces fit together.
• Getting Started to install it into a project.