![The file that tells Claude [or ChatGPT] how you work](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,format=auto,width=720,onerror=redirect/uploads/asset/file/ff2c87a4-2ea7-4fd9-9936-5fef16e74561/context-file-that-tells-claude-how-you-work-beehiiv.png?t=1785967066)
🧭 THIS WEEK AT BuildProven
Howdy, going to publish on schedule this week, a quick one about context!
as usual, would love it if you could share this newsletter, give feedback via the poll at the end, or best, just hit 'reply' and tell me what you need/want with AI. I will definitely reply and try to help best i can.🧰 Worth Your Click
Here are a few things I found recently:
Common workflows in Claude Code — A walkthrough of the everyday moves in Claude Code, including how it reads project context before it answers.
Manage Claude's memory with CLAUDE.md — Give Claude persistent instructions with CLAUDE.md files, and let Claude accumulate learnings automatically with auto memory.
Github stacked pull requests
🗺️ FEATURED INSIGHT
Today, a way of giving your AI tool of choice some common information/background, ie ‘context’ that is continuously known/used.
Build it first: your context file
Open a plain text editor. Make a file called context.md. Write these four sections about your actual job, not a generic version of it. It takes about 10 minutes, and you can build it from the template below.
# How I work
## My role
What you do, who you do it for, and whether you write code.
## What I bring you most often
Raw meeting notes. An email thread. A messy list.
## How I need output
Lead with the ask. Short paragraphs. No lists over five items.
## Rules I always end up repeating
1. Never invent a date, a name, or a number. Write "not stated" instead.
2. Separate what someone committed to from what someone only mentioned.
3. Flag anything that reads like a decision but has no owner.Now load it so it applies without being pasted. In Claude or ChatGPT, make a Project and add context.md to the project knowledge, it loads on every message in that project.
In Claude Code, save the same file as CLAUDE.md in the folder you work in and it gets read the same way. ChatGPT Projects and custom GPT instructions can use the file AGENTS.md similarly.
Above is just an example, you can give the file further detailed background/context and instructions you want.
If using the chat tools, Claude or ChatGPT, creating a project and adding above ‘context’ as project instructions causes all chats in that project to operate with that same basic knowledge. You could include things like:
project scope & basic information
any key background rules/instructions/informations
type of communication you like [long, shorter, formality type etc]
If using Claude Code [CLAUDE.md] or Codex [AGENTS.md] to build SW, you can include project background as before, but you can also add more detailed details about the SW project
architecture
build and test procedure, etc.
(one thing you can do if you already have a bunch of code build, just type ‘/init’, and Claude Code or Codex will analyse the code and create the .md file for you)
That’s all this week, just a quick one. So main thing is, next time you’re creating a project, start with a common context file. That will give the AI consistent instructions and background/basis. As a result, you’ll get better results whether you’re working on a specific project in ‘chat’ or building a certain SW project.
Weekly build logs from a 25-year program manager who codes with AI.
— Brett
👉 Hit “Reply” and share your experience — I read every one!