Lesson Content

The prompt is your instruction set.

Claude does not have intentions, goals, or background knowledge about you unless you provide it. Everything Claude knows about what you need comes from what you put in the prompt. This means the quality of your input directly determines the quality of your output – not because Claude is unforgiving, but because it genuinely does not know what you meant unless you tell it.

A vague prompt does not produce a wrong answer – it produces a generic answer. Claude fills in gaps with assumptions. Sometimes those assumptions align with what you wanted. Often they do not. Power users do not rely on Claude to guess; they tell Claude exactly what they need.

What Claude reads in your prompt:

When you submit a prompt, Claude processes the entire text as context. It is picking up on:

• The task – what you are asking it to do
• The role or frame – who you are, what role you want Claude to play, what domain this is in
• The constraints – length, format, tone, what to include or exclude
• The audience – who the output is for
• The goal – what success looks like for this response
• Examples – if you give Claude an example of the kind of output you want, it will match that pattern

The more of these signals you provide, the more precisely Claude can target its response.

The context window.

Claude can only see what is currently in its context window – the active conversation. Think of the context window as Claude's working memory for that session. It has a limit. In a long conversation, Claude holds the entire thread in context, which means very long conversations can eventually push earlier content out of range depending on the model version you are using.

Practically, this means:

• Paste in the documents, notes, or background you want Claude to work with
• Do not assume Claude remembers something you mentioned 40 messages ago in a very long session
• For complex multi-step projects, keep your key reference information near the top of the session or re-paste it when needed

Claude calibrates to what it sees.

If you write in formal language, Claude will tend to respond formally. If you use bullet points, Claude often mirrors that structure. If you ask a brief question, Claude tends to give a proportionally brief answer. You can override this by explicitly stating the format you want – but understanding that Claude mirrors your style helps you control the output.

Iteration is normal and expected.

Claude's first response is rarely the final product. Professional users treat the first output as a draft, not a deliverable. They refine it with follow-up prompts: "Make this more concise," "Add a section on risk," "Rewrite the opening paragraph to be more direct," "This is good but the tone is too formal for our audience." Iteration is how Claude becomes genuinely useful, not a sign that something went wrong.