02-llm-editing-workflow/03-how-chatgpt-should-debug-an-app.md

How ChatGPT should debug an app

When debugging a Vroq app, find the owning layer before making changes.

Do not start by adding defensive conditions in random files.

Debugging order

Use this order when possible:

1. identify the visible symptom 2. identify the feature that owns the behavior 3. inspect state shape, actions, and reducer transitions 4. inspect API wrappers if remote data is involved 5. inspect sections or components that render the behavior 6. apply the smallest fix at the root cause

Typical ownership layers

Common sources of bugs:

reducer logic
action payload design
API response handling
feature section orchestration
component rendering or local state
layout or styling structure

The best fix usually happens in the layer that owns the behavior.

Prefer explicit debugging

Debugging is easier when the code is structured around:

clear feature folders
explicit actions
predictable reducers
isolated API helpers
small sections and components

This is why Vroq style strongly prefers these patterns.

Debug with structure, not guesswork

A good debugging edit does not guess. It checks:

whether the state is correct
whether the right action is dispatched
whether the reducer applies the right transition
whether the API wrapper returns the expected shape
whether the UI is reading and rendering the correct data

Avoid hacky bug fixes

Do not fix a reducer bug by adding UI-side workarounds.

Do not fix an API shape problem by duplicating normalization logic across several sections.

Do not fix a naming or ownership problem by adding more special cases.

Final rule

Debug by tracing ownership and data flow, then fix the problem at the layer that actually owns it.