deepfusionlabs/deep-init

README preview

<!-- DEEPINIT:HUMAN-AUTHORED — not a DeepInit-managed region -->
<p align="center">
  <img src="docs/deepinit-readme-banner.png" alt="DeepInit — grounded, verified code-truth your agent can load" width="100%">
</p>

<p align="center">
  <img alt="License: MIT" src="https://img.shields.io/badge/license-MIT-C4E934?style=flat-square&labelColor=070E1B">
  <img alt="100% local, read-only" src="https://img.shields.io/badge/100%25_local-read--only-2a3a52?style=flat-square&labelColor=070E1B">
  <img alt="No data egress" src="https://img.shields.io/badge/data_egress-none-2a3a52?style=flat-square&labelColor=070E1B">
  <img alt="Claude Code plugin" src="https://img.shields.io/badge/Claude_Code-plugin-2a3a52?style=flat-square&labelColor=070E1B">
  <img alt="Every claim grounded and verified" src="https://img.shields.io/badge/every_claim-grounded_%2B_verified-C4E934?style=flat-square&labelColor=070E1B">
</p>

<h1 align="center">DeepInit</h1>

<p align="center"><b>Your coding agent reads every file — and still breaks rules that were never written down anywhere.</b></p>

It rewrites a function with total confidence, then violates a business rule no comment mentions. It trusts your schema file when the live database already dropped that column. It changes one service and breaks another through a table it never knew they shared. It "cleans up" a workaround that was load-bearing.

DeepInit reads your codebase and **writes down the truth** — the real rules, the live database, the *why*, and the problems it finds — with **every claim grounded to a `file:line` and checked against your code before it's written.**

View full repository on GitHub →