Scoping an Agent Build

What it is for

An agent given a broad brief builds a broad half-finished app. Scoping to one thing that works, and keeping a checkpoint you can return to, is the difference between a session that ends with software and one that ends with a rollback.

The skill Copy

The agent builds what you describe, including the parts you did not mean. So
describe a slice.

First message — one user journey, end to end, and nothing else:
"A visitor opens the page, enters an email, and it is stored. Show a success
message. No accounts, no admin, no styling beyond readable."

Then, one change per message. After each one that works, take the checkpoint.
The agent's failure mode is not writing bad code, it is confidently rewriting
something that already worked while fixing something else — and without a
checkpoint that is unrecoverable.

Say explicitly what NOT to do:
- "Do not add authentication."
- "Do not change the database schema."
- "Do not install new packages without saying which and why."

Each of those is something an agent will otherwise do helpfully and expensively.

Before deploying, ask it to list every environment variable and every external
service the project now depends on. That list is almost never what you expected,
and it is what breaks the deploy.

How to install it Copy

1. Open the tool and start a new conversation, project or template.
2. Paste the text below wherever the tool keeps standing instructions
   — usually called system prompt, instructions, or persona.
3. Save it so it applies to every run rather than a single message.

Did it work for you?

Where it came from