Snake, with context
Same prompt — now with context
Exercise 1.1 gave you WOW from an empty spec sheet. The agent filled in the blanks from memory. That's fine for a hook; it is not how you do real work.
Now we add what a real spec sheet looks like: a theme, an audience, a vibe, constraints. Still one question, but different output.
Step 1 — Start a new Claude Code session and re-run the exact same prompt
Prompt:
Create a HTML based snake game with the following features:
- Make at least three levels
- Add an onboarding flow
- Make sure it has a theme song on the main menu
Deploy this to Vercel and give me the live URL
(Use Vercel CLI, that is installed, I have a Vercel account)
Same model. Same prompt shape. The only thing that changed is the spec next to it.
Compare to what Exercise 1.1 produced. Notice how much more of you is in the result — and how little of it came from the prompt itself.
Stuck?Debrief — what should have happened
- Three levels, an onboarding screen, a theme song.
- A more interesting result not because the model got smarter but because the brief became richer.
The promotion to remember: prompt-fiddling is the amateur move. Editing contextContextEverything Claude can see right now — the conversation so far, the files it has read, the tool results, the system instructions. A big text buffer the model reads end-to-end before every reply. is the professional one. Every time you find yourself rewording the same prompt three different ways, ask: what would I add to a .md file that would make the next prompt unnecessary?
Stuck?FAQ for this exercise
Q: Where should the spec live? Do I write it as a file? A: Both, adding it to the prompt or a separate file is identical. As long as you point the agent to the file if you choose the file.
Q: Why was the first version so generic? A: Because you gave the agent nothing to anchor on. With no context, it draws on the average of what "snake game" means in its training data. With a spec, it draws on your specification.
OptionalGo deeper
Now make a game that teaches something. Same setup, same model — change only the brief.
Create an HTML game that teaches 7-year-olds the multiplication tables of 7
- Onboarding flow that explains the rules
- A score, three lives, a "well done" screen
- Cozy palette, encouraging tone
Deploy this to Vercel and give me the live URL
(Or pick your own topic: Spanish food vocabulary, capital cities, PythonPythonA programming language — the lingua franca of data science, machine learning, and scripting. A separate ecosystem from Node and npm, with its own runtime and package installer (pip). list comprehensions.) The game now has a purpose it didn't a minute ago — and the only thing that changed is the brief sitting next to the prompt.
ReflectInterlude A — What just happened
You just ran a controlled experiment on yourself. Same model, same prompt, two completely different outputs. The variable was a file.
a markdown file is the cheapest, most powerful unit of work you have. Every later exercise — skillsSkillA reusable bundle of know-how that Claude loads on demand. Lives as a folder of markdown on your machine; kicks in when the conversation matches what the skill is for., person.md, voice.md, program.md — is a variation on this theme. The agent does what you ask, but it does it in the style of the files around it.
The habit shift to install: every time you answer a question, write the answer in a file. Now every agent — yours and everyone else's — knows. Fewer questions, more productivity. Documentation is no longer a chore. It is fuel.