02-llm-editing-workflow/00-overview.md

LLM editing workflow overview

This section defines how ChatGPT and other LLMs should work when reading, editing, debugging, and refactoring Vroq apps.

The goal is not just to produce working code. The goal is to produce code that stays consistent, understandable, and maintainable over time.

Main goals

When editing a Vroq app, prefer these outcomes:

understand existing structure before changing it
keep edits small and focused
preserve good existing patterns
move messy code toward the standard house style
avoid accidental rewrites of unrelated logic
fix problems at the owning layer

Core workflow

The preferred editing workflow is:

1. inspect the app structure 2. identify the owning feature and files 3. look for an existing pattern to reuse 4. patch existing files where possible 5. create new files only when there is a clear responsibility boundary 6. keep the change easy to review and debug

Main rule

Before changing code, understand what already exists.

A fast edit that ignores the existing structure often causes architecture drift and duplicate patterns.

What this section covers

This section documents:

how ChatGPT should read an app
how ChatGPT should edit files
how ChatGPT should debug issues
how ChatGPT should debug a running app using Vroq MCP
how ChatGPT should refactor code safely
how to use Vroq MCP correctly
when to patch vs rewrite
how to handle file deletion safely

LLM editing workflow overview

LLM editing workflow overview

Main goals

Core workflow

Main rule

What this section covers

Read next