Give ChatGPT Your Stack and Conventions Up Front

What it is

A prompting habit: before asking for code, state your language, framework and versions, plus the conventions the code must follow — so what comes back fits your project instead of a generic template.

Name the stack, versions, and rules so the output drops in cleanly.

Why it works

Without a stack, ChatGPT picks defaults: the latest syntax, popular libraries, its own naming. That code compiles but clashes with your project, and you spend the saved time reconciling style. Front-loading the stack and rules makes the first output paste-ready far more often.

When to use it

Any code that has to live in an existing codebase with established patterns — a component in your design system, an endpoint matching your API style, a query in your ORM.

When not to use it

Throwaway scripts or standalone snippets where fit doesn't matter, and quick 'how do I do X' syntax questions where conventions are irrelevant.

Prompt

Stack: <language + version>, <framework + version>, <key libraries>.
Conventions: <naming, error handling, patterns to follow or avoid>.
Task: <what the code should do>.

Match the conventions exactly; if something's ambiguous, ask before assuming.

Example

'Stack: TypeScript 5, React 18, TanStack Query; conventions: no default exports, errors via Result type' — the generated hook fits the codebase with no reformatting.

Advanced version

Paste one representative file from your codebase as a style exemplar and say 'match this file's structure and naming' — the model mirrors your patterns better than any written rule list.

Common mistakes

Omitting versions and getting syntax your toolchain rejects.
Not naming your libraries, so it reinvents what you already use.
Describing conventions vaguely ('clean code') instead of concretely.