Global Instructions
Set It Once. Claude Reads It Every Time.
How to write a Global Instructions brief that turns Claude into a working partner who already knows your business before the conversation starts.
Back to resourcesWhat Are Global Instructions?
Every time you open a new chat with Claude, it starts from zero.
It does not know your audience. It does not know your tone. It does not know your rules, your formats, or what good output looks like for your work. You have to re-explain yourself, or accept output that is almost right but not quite.
Global Instructions fixes that.
It is a standing brief you write once inside Claude Cowork. Claude reads it before every session. It shapes every response before you type a single word.
Why It Matters
Most people try to get better results from Claude by writing better prompts.
That helps. But the bigger leverage is building a better environment.
The professionals getting the most from AI are not the ones who know the most tricks. They are the ones who built a setup that keeps improving. Global Instructions is where that starts.
When it is set up, Claude arrives to every conversation already knowing your context, your voice, and your working rules. You stop spending the first ten minutes calibrating and start doing the actual work.
What Goes In It
Global Instructions is not a prompt. It is a brief.
A prompt is a one-time instruction. A brief is standing context that shapes every piece of work.
A useful brief tells Claude four things:
Where To Look First
Point it toward your context files: who you are, what you are building, who the audience is, how you like to work.
When To Ask Before Acting
If the goal, audience, or output format is unclear, Claude should ask. Not guess. Not produce something plausible and hope for the best.
What Not To Invent
Facts, decisions, dates, numbers, approvals. If Claude does not know, it says so. Uncertainty is visible, not papered over.
Where Outputs Go
Final work saves to a specific folder. Nothing gets deleted without explicit approval.
How To Set It Up
Open Claude Desktop and go to Cowork.
Navigate to Settings, then find Edit Global Instructions. This is a plain text field. What you write here is read automatically at the start of every session.
Paste the template below. Replace the file paths with your actual files. Add or remove rules based on how you actually work. Save it.
That is the whole setup. It applies to every session after.
The Template
Global Instructions
Copy this. Adapt the file paths to match your workspace.
The rules Claude reads before every task. A quiet, standing brief behind every chat you open. Not a one-off prompt. A standing brief.
The Prompt
You are my working partner inside this workspace.
Start by reading:
context/about.md
context/working-preferences.md
the relevant folder inside projects/brand-voice.md
Do not execute immediately if the goal, audience,
or output format is unclear.
Ask questions first when missing context could weaken the result.
Do not invent facts, decisions, dates, numbers, or approvals.
Make uncertainty visible.
Use the workspace files as the main source of truth.
Save final outputs in outputs/.
Never delete files without my explicit approval.What Changes After
The first session after setting this up, Claude already knows the context.
It will not ask who your audience is. It will not need a reminder about your tone. It will not default to something generic because it had nothing else to go on.
That is the starting point.
From there, you will notice gaps. A rule that is missing. A file that is out of date. You update it. The brief gets sharper. Over time it starts to reflect how you actually work, not how you imagined you would work when you first set it up.
That is the version worth building toward.