Agents at Work / The tool stack / Claude Code
01 · Build · Primer
Claude Code, the builder.
Claude Code is a coding agent from Anthropic. You don't write code. You tell it what you want, the way you'd tell a new hire, and it writes the files, runs them, and fixes what breaks. Every piece of your agent gets made this way, one prompt at a time.
What you get
- Every prompt from the Scout build, with Brightline swapped for your business.
- A file you can open and read for every piece it makes.
- An agent that asks you when a fact is missing, instead of making one up.
How it fits in your agent.
What it is
Anthropic's coding agent. It lives in the terminal and in the Claude desktop app.
What it does for your agent
Writes every file from your prompts, runs them, and reads them back to you.
When you touch it
Every session. It is the thing you type into.
The one idea.
You don't write the job description. Your business facts do. Claude Code turns what you tell it into files, and you check every one. If a line isn't you, you say so and it fixes it.
What you'll do with it.
- A week before, paste the setup prompt. It gets your computer and your accounts ready.
- Session 1. Paste the briefs prompt and the job prompt. Read what it wrote.
- Session 2. Paste the tools and skills prompts. Try it on one real company.
- Session 3. Paste the memory, deploy, and schedule prompts. Watch the first full run.
Words to know.
- Prompt
- What you type to the AI. In the accelerator, one prompt makes one piece of your agent.
- Folder
- Your agent. Every piece is one file in this folder. Nothing more.
- Key
- A password for software. It lets your agent use a service like Apify. Never paste one into a chat.