Reference
Checklists, prompt patterns, a command cheat sheet, and a glossary — return here during the practical work.
Project-start checklist
Before building
Set the foundations
Before building0 of 6 done
During building
Keep quality visible
During building0 of 6 done
Before sharing
Final review
Before sharing0 of 6 done
Prompt patterns
Inspect a project
prompt
Inspect the project before editing. Identify its stack, structure, design-system
sources, relevant patterns, test commands, and risks. Tell me what you need to
clarify. Then propose a small plan and wait.
Implement a feature
prompt
Implement only [feature]. Reuse [components/patterns]. Handle [states]. Do not
change [boundaries]. Validate with [commands and manual checks]. Summarise files,
decisions, results, and remaining uncertainty.
Review a result
prompt
Do not edit yet. Review the rendered experience against these criteria: [list].
Separate critical issues, improvements, and optional polish. Cite the screen,
state, or file supporting each observation.
Debug a problem
prompt
Expected: [result]. Actual: [result]. Reproduction: [steps]. Evidence: [error or
screenshot]. Recent change: [change]. Diagnose before editing, then make the
smallest safe fix and repeat the reproduction steps.
Command cheat sheet
bash
# Where am I?
$ pwd
$ ls
# Start the project
$ npm install
$ npm run dev
# Stop the server
Ctrl + C
# Check a production build
$ npm run build
# Git checkpoint
$ git status
$ git diff
$ git add .
$ git commit -m "Describe the outcome"
$ git push
Glossary
Every key term from the guide, in one place. Use it during the practical classes when a word comes up and you just need a one-line reminder.
- Acceptance criteria
- Observable conditions that define whether work is complete.
- Agent
- AI system that can plan and use tools to take actions.
- API
- A contract through which software systems exchange requests and responses.
- Backend
- Server-side logic, data, and protected operations.
- Branch
- An independent line of work in Git.
- Commit
- A named checkpoint in a Git repository.
- Component
- A reusable piece of an interface with defined structure and behaviour.
- Context engineering
- Designing the information, instructions, examples, tools, and feedback available to AI.
- Context window
- The information a model can use for one response.
- Database
- Organised persistent storage for data.
- Dependency/package
- Reusable external code included in a project.
- Deployment
- Building and hosting an application so others can access it.
- Design token
- A named design decision such as a colour, spacing value, or motion duration.
- Environment variable
- Configuration supplied outside the main code, often used for secrets or environment-specific values.
- Few-shot prompting
- Giving examples for a model to imitate.
- Framework
- An organised set of conventions and tools for building applications.
- Frontend
- The part of an application rendered for and operated by the user.
- Git
- Version-control software that records project history.
- GitHub
- An online platform for Git repositories and collaboration.
- Hallucination
- Plausible AI output that is unsupported or false.
- JSON
- A common text format for structured data.
- LLM
- A model trained to process and generate language.
- Localhost
- An application running on your own computer.
- MCP
- A standard for connecting AI applications to tools and data sources.
- Model
- A trained mathematical system that recognises or generates patterns.
- Motion.dev
- A JavaScript animation library suited to interface motion and interactions.
- Node.js
- A runtime that executes JavaScript outside the browser.
- npm
- The package manager commonly bundled with Node.js.
- Prompt
- The immediate request or instruction given to AI.
- Pull request
- A proposal to review and merge one branch into another.
- Repository
- A project tracked by Git.
- Responsive design
- Designing behaviour and layout that adapt to available space and input methods.
- Semantic HTML
- HTML elements chosen according to meaning and function.
- Token
- A chunk of information processed by an AI model.
- TypeScript
- JavaScript extended with type information.
- Vibe coding
- Building software conversationally with AI while directing and reviewing the work.
- Vertical slice
- A small end-to-end path through a product that works completely.