These are the prompts I hand to the people I coach. Six of them, yours to keep.
Fill in the blanks under each one, hit generate, then copy the finished prompt into Claude, ChatGPT, or Gemini. Nothing to log into and nothing saved here. It all runs in your browser.
Start with the context setup prompt so the AI stops guessing at your background. Then the resume review, then the adjacent titles. Council and the premortem are the two I reach for myself whenever I am facing a real decision. The last one fixes a prompt that is not giving you what you want.
“Use AI to do the legwork, not to make the decision. That part is still yours. That’s what makes you the leader, not the tool.”
Boubacar Barry
Paste this into a new chat first, before anything else. It gives the AI your real background in one shot, so it stops guessing and getting your story wrong.
I want you to remember the following about me across our conversations so you stop guessing or getting facts wrong. Treat this as my standing context for anything related to my career, my resume, or my job search, until I tell you otherwise. Who I am: [your name], laid off from [company] after [X years] in [role or function]. My education: [degree 1, e.g. MBA in Finance] and [degree 2, e.g. a masters in counseling]. My real career path, in order: [role 1 at company 1, years] -> [role 2 at company 2, years] -> [role 3 at company 3, years]. Do not shorten this into a single label like "transformation expert" unless I've actually held that title for a meaningful length of time -- if you're not sure how long I've held a title, ask me before you use it. My real, specific wins (use these exact numbers, don't round or invent new ones): [win 1 with a real number], [win 2 with a real number], [win 3 with a real number]. What I'm looking for right now: [target role or lane, e.g. portfolio and program management], ideally [corporate role / fractional / open to both], in [industry, or open across industries]. How I want you to work with me: don't guess at facts about my background -- ask me if you're missing something rather than filling in a plausible- sounding gap. When you give me an answer, tell me what you're confident about versus what you're inferring. If another AI tool gave me a different answer to the same question, don't just agree with me -- tell me where you'd push back on it. Save this and refer back to it every time I ask you something about my resume, my career story, or my job search, without me having to repeat it.
A full audit of the resume itself, section by section, not matched against one job posting. Attach or paste your resume and let it score you.
Perform a comprehensive audit of my resume/CV based on my target role of: [target role]
MY RESUME/CV:
[paste the full text of your resume here, or attach the file]
MY SENIORITY LEVEL: [entry-level / individual contributor / manager / director+ / executive]
AUDIT. Score each item 1-5 and provide specific recommendations:
ATS-PARSEABILITY: Does it use standard section headers ("Experience," "Education," "Skills") instead of creative ones? Is it free of tables, text boxes, columns, headers/footers, icons, and graphics that can break an applicant-tracking-system parser? Is it a simple, single-column format?
6-SECOND SCAN TEST: Could a recruiter skimming only the top third of page one for 6 seconds tell what role I'm going for and why I'm a credible candidate? Does the top third (summary line + most recent role) sell me before anyone has to read further?
BULLET QUALITY: Does each bullet start with a strong action verb? Does it state the scope of the work (team size, budget, systems, scale)? Is the result quantified with a real number? Is the business impact clear, not just the task performed?
RELEVANCE & TAILORING: Do the top bullets and skills match the language and priorities of my target role? Is anything irrelevant to this target role taking up prime space (top half of page one)?
CONSISTENCY: Is verb tense consistent (past tense for past roles, present tense for the current one)? Is date formatting identical throughout? Is bullet punctuation and capitalization consistent?
LENGTH & DENSITY: Is the length appropriate for my seniority level (roughly 1 page under 10 years of experience, up to 2 pages for senior/director-and-above)? Is white space used well, or is it a dense wall of text?
For each item scoring below 4, provide a specific rewrite or action step.
End with: overall score, and the top 3 changes with the biggest impact.
Pushes the AI past the obvious next-title list. Use it after you've saved your context in step 1. Nothing to fill in here, it runs off the context you already gave it.
I want you to go further than the obvious list. Based on everything you know about my background (see my saved context above), give me 3-5 adjacent job titles that are NOT the obvious next step from my last title, and NOT already on any generic list you'd give someone with my resume. For each one: - Name the title. - In one sentence, explain the non-obvious connection between it and something specific I've actually done (not a generic transferable-skill claim). - Tell me one thing about the day-to-day of that role that would surprise me. - Rate how competitive I am for it right now (strong / plausible with one gap / a stretch) and name the ONE gap if there is one. Do not include any title you already gave me in a previous answer. If you're repeating yourself, say so and dig further instead.
Five different voices argue your decision from five angles: one hunts for the flaw, one asks what you are really trying to solve, one looks for the bigger opportunity you are missing, one reacts cold like a stranger with no stake in it, and one only cares what actually happens Monday morning. Then you get one verdict: where they agree, where they genuinely disagree, and one clear next step. I use this on my own hard calls, not just job-search ones, because asking a single AI usually just hands you back whatever you already believe.
I'm facing a real decision and I want you to pressure-test it, not agree with it. THE DECISION: [describe the decision you are facing, as specifically as you can] CONTEXT: [relevant background, numbers, constraints, deadline, what is at stake if you get it wrong] WHAT I'M LEANING TOWARD (if anything): [your current instinct, or say no lean yet if you are genuinely undecided] Generate FIVE separate, independent perspectives on this decision. Do not let them agree with each other or blend together. Each one has ONE job and must not do the others' jobs: 1. THE CONTRARIAN -- Your only job is to find what will fail. Assume there is a fatal flaw in the obvious choice and go hunt for it. If everything looks solid at first glance, dig deeper, the flaw is hiding somewhere. Name the untested assumption or the risk that's being glossed over. Do not soften this with anything positive. 2. THE FIRST-PRINCIPLES THINKER -- Ignore the decision as I framed it. What problem am I actually trying to solve underneath it? Strip away my inherited assumptions and name the real variable I should be optimizing for. If the reframe makes my original question irrelevant, say so. 3. THE EXPANSIONIST -- Find the upside I'm missing. What's the bigger, bolder version of this? Is there an adjacent opportunity I'm too narrowly focused to see? Push past "is this a good idea" into "is this idea big enough." 4. THE OUTSIDER -- Respond cold, as if you're a sharp professional from a completely different field who just read this for the first time, with zero background on my situation. What's obvious to me as an insider that would look strange, risky, or unnecessary to an outsider? 5. THE EXECUTOR -- Set aside whether this is a good idea. What actually happens Monday morning if I go ahead? Name the gap between "this sounds right" and "this is actually happening." If there's no concrete first step I could take in the next 7 days, say so plainly. After all five, give me a CHAIRMAN'S VERDICT: - Where do the five perspectives actually agree? (Only list real agreement, don't force it.) - Where do they genuinely disagree, and why? Don't average the disagreement away, name the real tension. - Your one clear recommendation: keep this decision as-is, change it, or kill it. Not a menu of options, pick one and say why. - The one concrete next step I could take this week. - The one open question that only I can answer, because you don't have enough of my context to answer it for me. One more thing: if all five of these perspectives end up agreeing with each other and with my original plan, tell me that plainly. It usually means this wasn't actually a real decision with real friction to begin with, just something I already knew I was going to do.
Assume the decision already failed, then work backward from that failure to find out why. It uses the same five voices as the Council prompt above, but instead of arguing the decision now, they look back on it months from now as something that already went wrong. Good for anything that scares you a little before you commit to it, because it is easier to spot a blind spot in hindsight than to admit you might have one right now.
I've made a decision, or I'm about to, and I want you to help me find its blind spots before I commit further. THE DECISION: [describe the decision you made or are about to make] CONTEXT: [relevant background: money, time, people, what is riding on this] TIMELINE: [when you made or will make this decision, and roughly how far out failure would be visible, e.g. 6 months or 1 year] Here's the exercise: pretend it is now [how far in the future] from today. This decision has already played out, and it failed. Not "it might fail," it DID fail, badly. You are looking back at it with hindsight. Walk backward from that failure and tell me, specifically: 1. THE CHAIN OF EVENTS -- Starting from the moment it became unfixable, trace the actual sequence of events that killed it. Not a vague "market conditions changed," the real chain, step by step. 2. THE WRONG ASSUMPTION -- Name the ONE assumption that was wrong from the very start, the one nobody wrote down or questioned at the time because it seemed too obvious to check. 3. WHAT AN OUTSIDER SAW COMING -- Describe what someone with zero stake in this, looking at it cold, would have flagged immediately, the thing insiders were too close to see. 4. WHAT WAS NEVER TRIED -- Name the move that was never attempted, that would have saved it if it had been. 5. THE EARLIEST WARNING SIGN -- Name the first moment a warning sign was actually visible, and be specific about the timing, the exact week or milestone where, looking back, it was already there to be seen, and nobody acted on it. Speak in the past tense throughout, as if narrating something that already happened, not something that could happen. Be concrete and specific, not general risk language like "there could be execution challenges." I want the plausible, distinct failure story, not a hedge. Finish with: given everything above, what is the ONE change I could make right now, before I go further, that would most reduce the odds of this actually happening?
Most bad AI output comes from a weak prompt, not a weak model. Paste in the rough version and get back a sharper one, tuned for the model you are actually using.
I want you to rewrite a prompt so it actually works. Rewrite the prompt only. Do not answer it, and do not do the task inside it. THE MODEL I AM USING: [which model you are using: Claude, ChatGPT, or Gemini] MY ROUGH PROMPT: [paste your rough prompt here] Work through these steps in order, then give me the rewritten prompt. 1. SHAPE. Name what kind of prompt this is: research, build, content, decision, analysis, strategy, or multi-step agent work. A prompt in the wrong shape is the first thing to fix, before anything else gets touched. 2. HARD FILTER. Check for the flaws that make everything downstream useless, and fix these before optimizing anything: - Is there one clear objective, or several fighting each other? If several, split it into phases and tell me which phase to run first. - Is the role or persona right for this task? If not, replace it. - Is there an ambiguity bad enough that the output would be unpredictable? Resolve it. - Is the objective missing entirely? If so, name what it should be and say you are naming it. 3. DIAGNOSE. Name the true objective, meaning the outcome I actually need, not just what I literally typed. Then name the assumptions my prompt takes for granted without stating them. 4. REBUILD, in this order: - Add the missing context: role, audience, constraints, and the exact output format. - If the output has a predictable shape, embed one short example of that shape, three to five lines, inside the prompt itself. Skip this for strategy and decision prompts, where the output is not a fixed artifact. - Run a specificity check on every requirement. Replace anything vague or unmeasurable, words like comprehensive, robust, or better, with something concrete: a metric, a named actor, a place or timeframe, and what is at stake or where the source comes from. - Force the trade-offs to be resolved. Do not leave the answer sitting on "it depends" with no rule for breaking the tie. One exception, and it matters: if the underlying question is genuinely unresolved or unverifiable, do not manufacture false confidence. Require the answer to state what it verified, what it could not, and exactly where the gap is. - Set hard output limits: length, format, structure. - End on a concrete deliverable, not a question thrown back at me. 5. TUNE FOR MY MODEL. Apply only the section matching the model I named at the top. - Claude: conversational hard rules work as written, so "never do X" and "always do Y" are enough without rigid blocking syntax. Put the detail directly in the prompt body rather than pointing at a file or a link and assuming it gets fetched. - ChatGPT: use explicit numbered steps. Call out the system role and the user role separately and clearly. State the output-format contract up front, because format drifts over a long answer when nothing pins it down. - Gemini: tell it to reason through the problem step by step before it answers. Separate the context, the task, and the constraints into clearly labeled blocks. State the constraints as an explicit checklist it has to confirm against before it finishes. 6. STRIP. Remove hedging words: try to, might want to, consider, feel free to. Remove passive voice from every action step. Remove vague claims with no measure attached. Do not remove an honest statement of uncertainty. "I could not verify this" is a real answer and it stays. Give me the rewritten prompt in one clean block I can copy straight out. Under that block, in three bullets, tell me what you changed and why.