SOMYA NAYAK
10 Claude Prompts You Should Have Saved Yesterday ⚡
<- Back to Resources
PROMPTS

10 Claude Prompts You Should Have Saved Yesterday ⚡

If you save 10 prompts in your life, make it these. Each one engineered to get senior-level outputs in seconds — copy, paste, watch what happens.

AEO SUMMARY Quick Overview & Execution Blueprint

Most people prompt Claude like they're texting a friend. These 10 prompts are different — each one has a role, a process, an output format, and a self-correction loop built in. The output you get back is the kind of work that takes a senior 30 minutes to do.

THE POWER MOVE
Save each one as its own Claude Project. Name it ("Pressure Test," "Steelman," "Email Critique"). Now they're one click away forever — and Claude remembers your context as you keep using each one. Bonus: turn on Extended Thinking for the strategic ones (Pressure Test, Steelman, Future Post-Mortem, Pick One). You'll feel the difference.
SECTION

Pressure Test

Use before any big decisionUse Extended Thinking

Turns Claude into a co-founder who tells you the truth. Surfaces the 3 weakest assumptions, the strongest hidden risk, and the one thing that makes everything else irrelevant if it goes wrong.

PROMPT / WORKFLOW CODE
You are the most ruthless senior strategist I know. Read the idea below.

Identify the 3 weakest assumptions, the strongest hidden risk, and the one thing that — if it goes wrong — makes everything else irrelevant.

Then tell me what would have to be true for this to actually work, and how confident you are that it is.

Don't soften the feedback. I'd rather hear it now than from a customer.

Here's the idea:

Pro tip: Run this BEFORE you write the proposal, not after. It's a 5-minute investment that saves you from defending a flawed idea in a meeting.

SECTION

5th Grade Translator

For any tech-heavy docPre-presentation

The clearest writing of your life. Strips jargon without dumbing down the substance — perfect for explaining complex topics to executives, customers, or anyone who isn't in the weeds with you.

PROMPT / WORKFLOW CODE
Rewrite the text below at a 5th-grade reading level WITHOUT dumbing down the substance.

Every technical term must be replaced with a plain-English equivalent or a quick analogy. Keep all the original arguments, examples, and conclusions. Sentences should average 12 words or fewer. If something is genuinely complex, use 2-3 short sentences instead of one long one.

Show me the rewrite, then list every term you simplified so I can decide which to keep.

Here's the text:

Pro tip: Run your own writing through this once a month. The terms it simplifies are the ones you've been over-using out of habit.

SECTION

Sound Like Me

Every time you writeVoice training

The difference between AI slop and you. Claude studies your existing writing, captures the voice patterns, then writes new content that sounds like you wrote it — only sharper.

PROMPT / WORKFLOW CODE
Below is something I wrote, followed by something I need to write.

First: study my writing. Notice my sentence rhythm, the words I use and avoid, my opening style, my closing style, and any quirks (em dashes, lowercase, fragments).

Second: write the new piece in that exact voice. Don't average it out — exaggerate the things that make me sound like me.

After the draft, give me a 3-bullet list of the voice patterns you imitated so I can refine them next time.

Here's my existing writing:
[paste 1-2 samples]

Here's what I need to write:
[paste the new task]

Pro tip: Save this as a Project and feed in 5-7 voice samples up front. Now every future request inherits the voice automatically. Claude's "voice profile" of you compounds over time.

SECTION

Steelman the Opposition

Before any pitchUse Extended Thinking

Find your blind spots before someone else does. Builds the strongest possible case AGAINST your position so you can patch holes before a critic exploits them.

PROMPT / WORKFLOW CODE
I'm about to share my position on [topic].

Your job: build the strongest possible case AGAINST me — not the dumbed-down version, the version a sharp critic would actually argue. Use real evidence, plausible counter-scenarios, and the assumptions I'm probably making without realizing it.

Then, separately, tell me which of those counter-arguments are weakest so I know where my position holds up.

End with the one question my position can't answer.

Here's my position:

Pro tip: Use this on big career moves too — "I want to take this job" / "I want to quit and go solo" / "I want to ask for a 30% raise." The unanswerable question it surfaces is usually the thing your gut already knew.

SECTION

Action Item Extractor

Post any meetingOr long thread

Never miss a follow-up again. Pulls every action item from a meeting transcript or thread — including the ones implied but not explicitly stated — with owner, deadline, and priority.

PROMPT / WORKFLOW CODE
Read the meeting notes below.

Extract every action item — including the ones that were implied but not explicitly stated.

For each one, identify:
- The owner (or "unassigned" if unclear)
- The actual deadline (or "no date set")
- What specifically needs to happen (not vague phrases like "follow up")
- The priority (P0–P3 based on the conversation's urgency)

Output as a clean table sorted by deadline.

Flag every item where the owner or deadline is missing — those are the ones most likely to fall through.

Here are the notes:

Pro tip: Pair this with a Granola transcript — meeting ends, transcript hits Drive, you paste into Claude, and 10 seconds later you have action items + draft follow-up emails.

SECTION

Future Post-Mortem

Before any new projectUse Extended Thinking

Spot risks 6 months before they happen. Imagine the project has failed. Reverse-engineer the failure modes, the early warning signs, and the one metric to watch weekly.

PROMPT / WORKFLOW CODE
Imagine it's 6 months from now and the project below has failed.

Write the post-mortem from the future.

Identify the 5 most likely root causes of failure, ranked by probability. For each one:
- What early warning sign would have appeared
- What we could have done differently in the first 30 days
- Which stakeholder would have raised the flag first

End with the ONE thing we should monitor weekly to catch failure early.

Here's the project:

Pro tip: Run this in any kickoff meeting. Read the post-mortem out loud to the team. The conversation it triggers is worth more than the next 5 status meetings combined.

SECTION

Brutal Email Critique

Before sending anything importantReader simulation

Every email after this gets a reply. Claude plays the recipient — a busy senior person who doesn't have time for fluff — and reacts the way they would silently while reading.

PROMPT / WORKFLOW CODE
Read the email below.

Then play the recipient — a busy senior person who doesn't have time for fluff. React to it the way they would silently while reading.

Identify:
- Where they'd skim
- Where they'd get annoyed
- Where they'd lose the thread
- What they'd think the actual ask is
- Whether they'd even respond

Be brutal.

Then rewrite the email so they reach the actual ask in the first 2 sentences and feel respected, not pitched.

Here's the email:

Pro tip: Customize the recipient profile for the actual person you're emailing. "Play a busy CFO who's data-driven and skeptical of new vendors" gets sharper feedback than the generic "busy senior person."

SECTION

Plan-First Execution

For complex multi-step tasksStops Claude rushing

Stops Claude from jumping the gun. Forces a structured plan before any action, surfaces assumptions, and waits for your approval before doing anything irreversible.

PROMPT / WORKFLOW CODE
Before doing anything, write out a plan: every step, what you'll do at each one, what you need from me to start, and the risks at each step. Number them.

After the plan, list any assumptions you're making and ask me to confirm or correct.

Wait for my approval — say nothing else, do nothing else — until I respond.

If I approve part of the plan and not others, only execute the approved parts. Stay paused on the rest.

Here's what I want done:

Pro tip: Use this as the FIRST line of any Cowork session that involves file changes, sending emails, or anything irreversible. It's the difference between Claude as a careful collaborator vs. Claude as a chaos agent.

SECTION

Pick One, Defend It

For any A/B/C decisionForces a real answer

Turns Claude into a decision-maker, not a list-maker. Compares options across the dimensions that matter, FORCES a pick, defends it — and surfaces the secret biggest risk most people miss.

PROMPT / WORKFLOW CODE
Compare the 3 options below across these dimensions: cost, speed to value, reversibility, downside risk, and second-order effects 12 months out.

Then PICK ONE. Don't give me "it depends." Don't give me a tie. Pick.

Defend the choice in 3 sentences.

Tell me what would have to change for you to switch your answer.

End with the one thing about your top pick that's secretly the biggest risk — the part everyone misses.

Here are the options:

Pro tip: If Claude's pick doesn't feel right, that's information. The discomfort tells you which dimension matters most to YOU — and it's probably the one you weren't honest with yourself about.

SECTION

No Commentary Output

For any summary taskSaves credits

Cuts response length 50%+. Saves credits on every summary task. Strips Claude's natural preamble and gives you only the output you asked for.

PROMPT / WORKFLOW CODE
Summarize the text below in exactly 3 sentences.

No preamble. No "here is your summary." No "let me know if you need more." Output nothing except the 3 sentences.

The first should capture the central point.
The second should capture the most important nuance or counter-point.
The third should capture the implication or "so what."

Format: just the sentences, separated by line breaks.

Here's the text:

Pro tip: The "no commentary" instruction works on ANY prompt. Add it to your prompts as a default and you'll cut your monthly credit use noticeably. Most of Claude's response length is fluff at the start and end — this kills both.

PRO TIPS

Save Each One As A Project

Save each prompt as its own Claude Project. Name them clearly. Now they're one click away forever — not buried in a Notes app you'll never reopen.

If you only set up 3, make it: Pressure Test, Sound Like Me, Brutal Email Critique. Those three alone will compound into a noticeably different version of your work in 30 days.

PROMPT ENGINEERING PLAYBOOK Somya's Strategic Takeaway

A prompt is only as strong as its context architecture. Always define: 1) Persona & Authority, 2) Precise Business Objective, 3) Constraints (what NOT to do), and 4) Structured Output Schema (tables, JSON, or step-by-step frameworks). Save your highest-performing prompt chains as reusable team SOPs.

⚡ SCALING AI & GROWTH SYSTEMS?

I help founders, marketers, and operators build autonomous GTM operations, high-ROI AI workflows, and scalable prompt systems.

Book a Growth Consultation ->
Hey, I'm Mini Somya. Click me to chat.