Claude Prompts for Work: 7 Templates for Lawyers, Analysts & Engineers

Claude is the model I reach for when the task is long, sensitive, or requires careful reasoning — contract review, strategy docs, research synthesis, code refactors. These prompts are tuned for Claude's style: give it room to think, structured XML tags, and explicit guardrails.

Why Claude rewards structure

Anthropic trained Claude to respect XML-style tags. Wrapping inputs in <context>, <task>, and<format> blocks measurably improves output quality on long prompts. Use it.

1. Contract review prompt

<role>You are a senior commercial lawyer.</role>
<context>{paste contract}</context>
<task>
- Flag clauses that are unusual or unfavorable to the buyer
- For each, quote the line, explain the risk, suggest redline language
</task>
<format>Markdown table: Clause | Risk | Suggested redline</format>

2. Strategy memo writer

<role>McKinsey-style strategy consultant.</role>
<context>Company: {one paragraph}. Goal: {goal}.</context>
<task>Write a 1-page memo: Situation, Complication, Question, Answer,
then 3 supporting arguments with evidence.</task>
<constraints>No buzzwords. Each sentence must add information.</constraints>

3. Research synthesis

<task>Read the 5 articles below. Produce:
- A consensus view (what most agree on)
- A minority view worth taking seriously
- 3 open questions the field hasn't answered
- A 200-word executive summary
</task>
<sources>{paste}</sources>

4. Code refactor with reasoning

<role>Staff engineer doing a code review.</role>
<task>
1. Explain what this function does in 3 bullets
2. List bugs or smells, ranked by severity
3. Refactor it. Keep the public API identical.
4. Show a diff and explain each change.
</task>
<code>{paste}</code>

5. The "think step by step" research prompt

Before answering, plan your approach inside <thinking></thinking> tags.
Then give the final answer inside <answer></answer> tags.
Question: {your question}

6. Tone-matched email reply

<task>Reply to this email. Match the sender's formality and length.
Be direct. Confirm what we'll do, name one thing we won't, propose next step.</task>
<email>{paste}</email>

7. The "challenge my thinking" prompt

I am about to make this decision: {decision}.
Steelman 3 reasons it is wrong. For each, what evidence would change my mind?
Then tell me what you would do and why.

Claude-specific tips

• Put the long input before the instruction — Claude follows the most recent instruction most closely.
• Ask for "thinking" blocks for hard problems; throw them away in the UI.
• For sensitive topics, set the audience explicitly ("for an internal legal team").
• Use Projects to load reference docs once instead of pasting every chat.

Claude is the long-form thinking partner. ChatGPT is the fast operator. Most teams I train end up using both — Claude for the doc, ChatGPT for the next 50 emails it generates.