Glossary of Codex prompts for front-end work

Codex

OpenAI's coding agent. It works in a repository: it reads and writes files, runs commands, reads an AGENTS.md file for instructions and supports MCP servers. For current features, check Codex's documentation.

Task brief

The prompt for one piece of work, written so it stands on its own: goal, constraints, design values, files in scope and a definition of done. See how to brief Codex for design.

AGENTS.md

A Markdown file in a repository that holds instructions for coding agents: the stack, the commands, the rules. See design rules in AGENTS.md.

Acceptance check

A condition with a yes or no answer that the agent can test itself, such as a passing build or no horizontal scroll at 390px.

Definition of done

The list of acceptance checks at the end of a brief. It tells the agent when to stop and tells you what was verified.

Scope

The files and folders a task may change, and those it must not. Scope keeps the diff small enough to review.

Design tokens

The named values a design is built from: colours, type sizes, spacing, radii, durations. Kept in one file, they give every task the same system.

Fixture data

A small file of realistic sample data in the repository, used to build and test a screen before a real data source is connected. See dashboard prompts.

Follow-up task

A small task that corrects one issue in a build: where, observed, expected, check. See follow-up prompts that fix a build.

Stop rule

A line in a brief or in AGENTS.md that tells the agent to stop and ask when it would otherwise invent a value or step outside the scope.

MCP

Model Context Protocol, a way for an AI agent to use outside tools. With uberdesigner's MCP server ↗, an agent can fetch a template by name. See the Codex setup guide.

Restyle

Changing how an existing app looks without changing what it does. See restyle prompts.