Start by Telling Claude Who It Is and Who You Are
What it is
A reliable first-message shape for Claude: name the role you want it to take, give the context you're working in, state the constraints, then make the ask — in that order.
Why it works
Claude has no memory of you at the start of a chat. The first message is the entire world it reasons from. A little structure up front replaces a dozen clarifying round-trips later.
When to use it
The opening message of any non-trivial task. The higher the stakes or the more specific the output, the more the framing pays off.
When not to use it
Quick factual questions or casual chat — there the ceremony is slower than just asking.
Prompt
You are <role, e.g. a senior back-end engineer>.
Context: <what I'm building and why>.
Constraints: <language, style, length, what to avoid>.
Task: <the specific ask>.Example
Opening with 'You are a copy editor for a fintech blog; keep it plain and non-hyped; tighten this paragraph' gets a focused edit instead of a generic rewrite that adds buzzwords.
Advanced version
For work you repeat, save your best framing and reuse it. When you find yourself pasting the same setup often, graduate it into a Claude Project so the context is always on.
Common mistakes
- Opening with the bare task and then spending five messages adding the context you could have led with.
- Assigning a role but no constraints, so tone and length are a coin flip.
- Writing a paragraph of role-play backstory that adds flavour but no signal.