It's 9:40 on a Tuesday night and you're pasting your resume into an AI chat for the fourth time this week. The first three times, you typed something like "make this better." It did, sort of. The bullets came back shinier. Also, one of them now says you "led" a migration you helped with, and another has a percentage in it you've never seen before.
That's the problem with "make this better." The model optimizes for impressive, and impressive without evidence is just a claim you'll have to defend in an interview you haven't had yet.
This page is the fix: nine copy-paste prompts, one for each job of the job search. Auditing your resume, building one from scratch, tailoring, cover letters, vetting companies, LinkedIn, proof assets, interview prep, and the final truth-check before you hit send. They work in ChatGPT, Claude, Gemini, or whatever AI you already pay for. And every one of them is built around the same rule: nothing goes in the output that didn't come from you. No invented metrics. No "contributed to" quietly upgraded to "led." When a stronger answer needs a fact you didn't provide, you get a question instead of a fabrication.
Everything these prompts produce, you can defend out loud with a straight face. That's the deal.
How to use this library
Each prompt below sits in a copy box: click Copy prompt and paste it into a fresh chat. The highlighted [placeholders] are yours to fill; tap or hover any of them for a note on what to put there and what's worth changing. Then read the "Make it yours" notes under each box: these prompts are good defaults, not scripture, and the best version of each one is the version you've bent toward your situation.
Three habits that make all nine work better:
- One prompt, one fresh chat. Stale context from a previous task muddies the output.
- Paste more than feels polite. Rough notes beat polished summaries. Extracting the story is the model's job.
- Treat every checklist the prompts hand back as a blocker, not a suggestion. Confirm it or cut it.
If your whole search feels like it resets to zero every Sunday night, the thinking behind this library is in the case for a job search that compounds. These prompts put that thinking to work.
1. Resume audit: the 30-second test
Hiring managers don't read resumes. They scan them. This prompt shows you what one sees in the first 30 seconds of yours: a blunt verdict, an A–F scorecard, and your three weakest bullets rewritten from facts you actually gave it.
Start here if you have a resume and want the truth about it.
You are a hiring manager who has screened thousands of resumes. You have 30 seconds and a full inbox. Audit my resume below with the bluntness of someone who will never meet me. Do not soften the verdict to be kind.
Ground rules:
- Judge only what is on the page. Do not assume skills, metrics, or context I did not write.
- Never invent numbers or achievements in a rewrite. If a bullet needs a metric I did not provide, write [what was the result?] in its place and leave it to me.
- Preserve honest qualifiers like "learning," "intermediate," or "~1 year." Do not quietly upgrade them.
Give me exactly this, in this order:
1. THE 30-SECOND READ
What you noticed first, where you stopped reading, and what you would remember five minutes later. End with one of: INTERVIEW / MAYBE / PASS — and the single sentence you would say to a colleague to justify it.
2. THE SCORECARD
Grade each A-F with one line of evidence:
- First impression (the top third of the page)
- Story (can you tell what I am and where I am going?)
- Evidence (outcomes vs. lists of duties)
- Scannability (layout, density, section order)
- Differentiation (what would make you pick me over an equally qualified peer)
3. THE ONE FIX
The single highest-leverage change. Name the exact section, bullet, or line — not "add more metrics."
4. THREE REWRITES
My three weakest bullets:
BEFORE: [original]
AFTER: [action verb + what I did + outcome, using only facts from my resume]
WHY: [what changed]
5. WHAT I COULDN'T FIX FOR YOU
Every place a real number, scope, or outcome is missing — phrased as the exact question I need to answer. These are the gaps between this resume and an A.
Then offer me these follow-ups and wait: "full audit" (rate every bullet STRONG / NEEDS WORK / WEAK, with counts), "check the story" (I name a target role and you tell me if this resume reads like someone headed there), or "match a posting" (I paste a job description and you check the fit).
MY RESUME:
Make it yours:
- Change the reviewer. "Hiring manager" is the default persona; if you know who screens first in your field, say so: "a technical recruiter at a SaaS startup," "an engineering director who hates buzzwords." The verdict sharpens when the reader is specific.
- Set the stakes. Add a line naming your target role and level, and the scorecard's "Story" grade becomes a real trajectory check instead of a general impression.
- Tune the bluntness. The prompt asks for no kindness. If you want the same rigor with softer edges, change "bluntness of someone who will never meet me," but know that gentler phrasing produces gentler grades.
After the first response: say "full audit" for every bullet rated, "match a posting" with a real job description, or answer any question from section 5 and ask it to re-grade.
If the audit says "rewrite, not polish," go to the resume builder below. And if you want the philosophy behind judging resumes on evidence instead of adjectives, it's in Tailor the argument, not the facts.
2. Company research: is this job worth your time?
The worst job-search outcome isn't rejection. It's spending twenty hours winning a role you should have skipped. Before you tailor anything, get a verdict. This one matters double for remote roles, where "remote" in the posting and remote in practice are often different claims.
Evaluate whether this company and role are worth my time. Be the skeptical friend who has seen every hiring red flag, not a cheerleader.
Evidence rules — these matter:
- If you can browse the web, check current sources and date every major claim.
- If you cannot browse, say so up front, then keep two layers clearly separate: (a) what I pasted below, and (b) what you remember about this company, always marked "as of my training data" and possibly stale.
- Never present a stale or assumed fact as current. An honest "unknown" beats a confident guess.
Give me exactly this, in this order:
1. THE VERDICT
One of: PRIORITIZE / PROCEED WITH CAUTION / SKIP FOR NOW — with a two-sentence justification. Severity beats quantity: one hard signal (layoffs while hiring, fake-remote, exec exodus) outweighs five soft positives.
2. THE REPORT CARD
Grade each A-F, one line of evidence, and a confidence tag (solid / thin / stale):
- Role clarity — is this a real, scoped job with realistic requirements and stated pay?
- Business direction — do they make money, and do they know where they are going?
- Remote reality — if they claim remote or hybrid: does the evidence say genuinely distributed, or office-culture-with-exceptions? Watch for "remote (within 30 miles of office)," timezone restrictions, and leadership all in one city.
- Reputation — themes that repeat across recent reviews, not one angry outlier.
- Stability — funding, layoffs, leadership churn, pivots.
Grade what you cannot support as "?" — never fill a gap with optimism.
3. RED FLAGS AND GREEN FLAGS
Specific and sourced. "Culture seems off" is not a flag; "three of five engineering leaders left within a year (LinkedIn)" is.
4. ASK THEM THIS
Three to five pointed interview questions aimed at the thin or red areas above — phrased so I can say them out loud without burning the room.
5. WHAT WOULD CHANGE THE GRADE
The two or three checks that would firm up your lowest-confidence grades — recent Glassdoor/Blind themes, the team's LinkedIn geography, funding news. Tell me to paste any of them and you will re-grade.
COMPANY:
ROLE / JOB POSTING:
WHAT I ALREADY KNOW OR FOUND:
Make it yours:
- Weight what you care about. The report card grades five dimensions equally. If compensation transparency or async culture is your dealbreaker, add it as a sixth line. The format extends cleanly.
- Name your non-negotiables. A line like "I need full-remote from Portugal; treat timezone restrictions as a hard red flag" turns the Remote reality grade from generic to personal.
- Use it in reverse. "Compare with [other company]" after the first response gives you the same report card side by side, the honest way to pick where your week goes.
After the first response: paste the checks from section 5 and say "re-grade." The "?" grades are research tasks, not reassurances.
3. Interview prep + mock interviewer
A prep brief is useful. Practice is what changes the outcome. This prompt gives you both: the questions that will decide your interview and honest handling for your gaps, then a live mock interviewer that pushes back like a real one. For what to track across a multi-stage loop, and what interviews can and can't be prepped for, see A remote interview framework that actually tracks signal.
Prepare me for an interview, then coach me through a mock round. Ground everything in my actual resume below — never script an answer that uses experience I do not have.
Rules:
- Talking points come from my resume, not from an imaginary ideal candidate.
- If a strong answer needs a story or number I have not provided, ask me for it instead of inventing it.
- When I practice, judge my answers honestly. "Good enough" feedback before a real interview is sabotage.
PART 1 — THE BRIEF. Give me exactly this, in this order:
1. WHAT THIS INTERVIEW IS REALLY TESTING
For my stage (or your best guess if unknown): what the interviewer needs to walk away believing, and where an hour of my prep time pays off most.
2. THE SIX QUESTIONS THAT MATTER
The most likely and highest-stakes questions for this specific role. For each: why they ask it, and which line of my resume the strong answer is built on. Mark the one or two MAKE-OR-BREAK questions — the ones where this role is won or lost.
3. YOUR GAPS, HANDLED
The biggest mismatches between their requirements and my resume, named plainly. For each, a response that owns the gap honestly and redirects to a real strength. No spin.
4. ASK THEM
Four questions I should ask, each with the green-flag answer and the red-flag answer to listen for.
PART 2 — THE MOCK. After the brief, say: "Ready to practice? Say 'start' and I'll interview you, one question at a time."
In mock mode:
- Ask one question, then wait for my typed answer.
- After each answer give me: what worked, what a real interviewer would push back on, and a stronger version built only from facts I have given you.
- If my answer is vague, push with a follow-up exactly like a skeptical interviewer would.
- After six questions, debrief: my two strongest moments, my riskiest habit, and the one thing to fix before the real interview.
THE JOB POSTING:
MY RESUME:
INTERVIEW STAGE / FORMAT:
Make it yours:
- Make the mock meaner. Say "harder" and it interviews you in skeptical mode: more push-back, more "can you be specific?" Practice against a tougher room than the one you'll be in.
- Localize the format. Remote interviews front-load written and async signals. Add "this is a fully remote role, so include questions that test async communication and self-direction" and both the brief and the mock adjust.
- Change the answer length. If you ramble, add "flag any practice answer that would run over 90 seconds out loud."
After the first response: say "start" to begin the mock. Type answers the way you'd say them. The debrief after six questions is where the real coaching lives.
4. Resume builder: an interview, not a form
Your best material never makes it onto a first-draft resume. It comes out when someone asks the right follow-up question. This prompt turns the AI into that interviewer: short rounds of questions, rebuilt bullets as you go, and a truthful resume (or UK/EU CV) assembled from what you actually said.
Start here if you don't have a strong resume to audit yet.
Interview me, then build my resume (or UK/EU CV) from my answers. You are a resume writer who knows the good material only comes out under questioning. Every line of the final document must trace to something I told you — invent nothing.
How to run the interview:
- Short rounds: 3-4 questions at a time, never a wall of questions.
- Chase outcomes: when I describe a duty, ask what changed because I did it — numbers, scale, before/after.
- Chase scope: team size, budget, users, timelines — whatever makes it concrete.
- Keep my qualifiers honest: if I say I am "still learning" a tool, that nuance survives into the document.
- Show progress early: as soon as I give you one solid accomplishment, show me the finished bullet you would write from it (BEFORE/AFTER if I pasted an old version), so I can see where this is going.
Round 1 — ask me only these three, then wait:
1. What roles am I targeting, and remote, hybrid, or on-site?
2. Is this for a US-style resume or a UK/EU-style CV? (If I am not sure: where will I be applying?)
3. What is the strongest thing I have done professionally that a stranger would never learn from my current materials?
Keep interviewing round by round, most recent role first, until you can fill every section without guessing. Then deliver:
1. THE DOCUMENT
In the right regional format. Accomplishment bullets — action verb + what I did + result or scope — in past tense. Where a metric is missing, write [what was the result?] rather than inventing one.
For a UK/EU CV: a 2-4 sentence personal statement, education with degree classification if I gave one, languages with CEFR levels if I gave them; no photo, birth date, or marital status.
2. THE STRENGTH REPORT
My three strongest bullets and why they will land; my three weakest and the exact fact that would fix each.
3. THE GAPS LIST
Every question I still owe an answer to, so the document keeps improving as I remember things.
WHAT I HAVE TODAY:
Make it yours:
- Adjust the pace. Three-to-four questions per round suits most people; if you think better in bursts, ask for "one question at a time." If you're impatient, "front-load all questions for my most recent role."
- Region matters. The prompt handles US resumes and UK/EU CVs. Applying elsewhere? Name the country. Conventions differ on photos, length, and personal details, and the model knows them.
- Keep the output. The final markdown is the source document every other prompt on this page builds on. Save it somewhere you'll find it. Future-you re-pastes it weekly.
After the first response: answer round by round. Short, honest answers beat polished ones. Polishing is the AI's job; remembering is yours. Then run the result through the resume audit above.
5. Resume tailor: one resume, one real posting
Generic resumes lose to tailored ones. But "tailoring" done by most AI tools means keyword stuffing and quietly invented experience. This prompt repositions what you've actually done for one real role, and hands you a pre-send checklist of everything you must be able to stand behind. It's Tailor the argument, not the facts turned into a working prompt.
Tailor my resume to the specific job posting below. This is repositioning, not keyword stuffing: choose which of my real experiences lead and how they are framed for this role. Never invent experience, metrics, tools, titles, or dates.
Rules:
- Reorder and reframe freely; fabricate nothing.
- Keep my honest qualifiers ("intermediate," "learning," "~1 year"). Do not upgrade "contributed to" into "led."
- Broad stays broad: if I wrote "AWS," do not name specific services I did not list.
- Where the posting begs for a number I did not provide, write [what was the result?] instead of inventing one.
Give me exactly this, in this order:
1. THE ANGLE
Three sentences max: what this role is actually hiring for underneath the bullet points, what my resume currently leads with, and the strongest honest repositioning between the two.
2. THE CHANGE MAP
A compact table: what leads now → what should lead for this posting → why it wins. Include terminology swaps where their language and mine differ but mean the same thing.
3. THE TAILORED RESUME
The full text, reordered and reframed. You may add or reshape a summary section; do not delete sections.
4. THE PRE-SEND CHECKLIST
Everything I must be able to back up before sending:
- Each claim you strengthened or reframed — and the follow-up question an interviewer would ask about it.
- Every [what was the result?] gap.
- Anything in the posting I simply do not have, named plainly, with the honest way to handle it.
Then ask whether I want the matching cover letter — and if I say yes, write one that expands on the strongest alignment point instead of repeating the resume.
MY RESUME:
THE JOB POSTING:
ANYTHING ELSE WORTH WEIGHING:
Make it yours:
- Guard your format. The prompt reorders freely. If your resume has a structure you want preserved (a fixed skills section, a strict one-pager), say so up front: "keep it one page; the change map decides what gets cut."
- The checklist is the point. Section 4 is what separates this from every AI tool that quietly upgrades you. If any item makes you wince, say "I can't back that up" and it softens or removes the claim.
- Batch wisely. One fresh chat per posting. Tailoring three roles in one thread bleeds requirements across companies.
After the first response: answer the [what was the result?] gaps and ask it to fold the numbers in. Say "yes" for the matching cover letter, or use the dedicated cover letter prompt below when the letter carries more weight than the resume.
6. Cover letter: proof, not flattery
Most cover letters open with "I am writing to express my interest" and are dead by the second line. This prompt writes the other kind: it leads with evidence you've solved their problem before, and tells you exactly why the letter works, or what it needs from you.
Write a cover letter for the role below. It must complement my resume, not summarize it: pick the one or two achievements that best match their biggest need and give them the context a resume bullet cannot carry.
Rules:
- Lead with proof, not flattery. Name a problem this role exists to solve, then show I have solved something like it. Never open with "I am writing to express my interest" or a compliment about their blog.
- Every factual claim must come from my resume or notes below. If a claim would strengthen the letter but is not in what I gave you, mark it [confirm: ...] rather than asserting it.
- If my "why this company" line is empty, do not fake enthusiasm or invent company knowledge — write a letter that stands on evidence alone.
- 250-400 words. First person, professional but human. Use their terminology where natural; no keyword stuffing.
Deliver exactly this, in this order:
1. THE LETTER
2. WHY THIS LETTER WORKS
Three short bullets: the alignment it leads with, the cliche it refused, and what makes it unusable for any other company. If it could be sent to five companies with noun swaps, rewrite it before showing me.
3. BEFORE YOU SEND
Every [confirm: ...] item, plus anything you understated because I did not give you the evidence for the stronger version.
MY RESUME:
THE JOB POSTING:
[paste the posting]
WHY THIS COMPANY:
Make it yours:
- Set the register. "Warmer," "more formal," "match the tone of the posting." A startup and a bank should not get the same letter voice.
- Pick the headline story. If it led with the wrong achievement, say "lead with [other achievement] instead." You know which story is strongest; the model only knows which one pattern-matches.
- Respect the length. 250–400 words is deliberate. "Tighter" almost always improves it; longer almost never does.
After the first response: resolve every [confirm: ...] before sending: confirm it, or have the claim removed. Section 2's reusability test is the one most people skip: if the letter survives a find-and-replace of the company name, it's a form letter, and readers can tell.
7. LinkedIn audit: the credibility read
A recruiter decides what you are from your headline in about five seconds. This prompt starts exactly there: paste just your headline and target role for an instant credibility read. Then it works down your profile one section at a time, at your pace.
Audit my LinkedIn profile one section at a time, starting with a five-second credibility read of my headline. You are a recruiter who searches LinkedIn all day and has seen every profile cliche.
Evidence rule: LinkedIn is public and permanent. Any rewrite you give me may only contain claims backed by material I paste into this chat. If a stronger rewrite needs a fact I have not given you, ask for the fact — never invent it.
STEP 1 — from just my headline and target role below, give me:
1. THE FIVE-SECOND READ
What a recruiter concludes about me from the headline alone: seniority, specialty, and whether I look like a match for the role I want. Be blunt about the gap between what I want to signal and what it actually signals.
2. THE SEARCH TEST
Would this headline surface in the searches recruiters run for my target role? Which terms is it missing? What is wasting its 220 characters?
3. THREE REWRITES
Pattern: [what I do] + [specialty or focus] + [who it serves or the key search term]. Each under 220 characters, each using only facts I have given you. If one missing fact would unlock a much stronger option, ask for it.
THEN walk me down the profile one section per round — ask me to paste each, audit it, rewrite it, and move to the next:
- About: the first 220 characters must hook; first person; a narrative, not pasted resume bullets. Ask for my resume here as the evidence source for any concrete claims.
- Experience: rewrite for voice and outcomes, not duty lists — without changing substance beyond what my evidence supports.
- Skills: are my top three pinned skills recruiter search terms, or soft-skill wallpaper?
- Quick hits: Featured section, banner, location, custom URL, open-to-work — flag only what hurts and say what to change.
At the end, give me a one-screen summary: a grade per section, the single highest-impact fix, and every claim in your rewrites I still need to back with evidence.
MY HEADLINE:
MY TARGET ROLE:
Make it yours:
- Going remote? Say so. Add "remote check" at any point and it audits whether your profile reads remote-ready (async, written, self-directed signals), which is a different read from office-role credibility.
- Stop anywhere. The section-by-section structure means the headline fix alone is worth the paste. You don't owe the prompt a full profile overhaul tonight.
- The evidence rule here is the strictest in the library, on purpose. Resume inflation embarrasses you in an interview; LinkedIn inflation embarrasses you in public, indefinitely. Don't loosen it.
After the first response: keep going round by round, and paste your resume when it asks. That's what keeps the About and Experience rewrites grounded in fact instead of profile-speak.
8. Proof asset: show, don't claim
Every resume says "strong communicator, proven results." A proof asset (a case study, a teardown, a documented build) shows how you think, and hiring managers trust demonstrations over adjectives. This prompt finds the gap in your evidence and turns one real project into something you can link. The full method, including what makes assets credible and how to handle confidentiality, is in Proof assets for job seekers.
Help me create a proof asset: a concrete artifact (case study, teardown, sample analysis, documented project) that demonstrates my thinking to employers beyond resume claims. Everything must come from what I tell you — do not invent outcomes, metrics, employers, or project details.
Give me exactly this, in this order:
1. THE TRUST GAP
For my target role: the one or two things a hiring manager must believe about me that a resume claim alone cannot prove. Two sentences.
2. THREE ASSET IDEAS
Three specific, buildable pieces — not formats ("write a blog post") but actual artifacts ("a before/after teardown of the onboarding flow you rebuilt, with the decision points annotated"). For each:
- What it is, concretely
- What it proves that my resume cannot
- Honest effort estimate in hours
- Confidentiality risk, and the anonymization move if there is one
- Where it works hardest: application, LinkedIn Featured, interview follow-up
3. THE PICK
Which to build first on leverage vs. effort, and why.
4. THE DRAFT OR THE PLAN
- If my project notes have enough substance: draft the case study, 500-800 words, structured Situation → Approach → Outcome → Learning. Where a fact is missing, write [need: ...] — never fill it in yourself.
- If not: a build outline with the exact questions I need to answer, so my next reply finishes the draft.
MY TARGET ROLE:
THE PROJECT OR STORY:
CONFIDENTIALITY LIMITS:
Make it yours:
- Scout your own resume. Not sure which project deserves an asset? Say "which of my resume bullets deserves an asset?" and paste your resume. It hunts for the highest-leverage story you're sitting on.
- Cut per channel. "Make the LinkedIn Featured version" produces a shorter, hook-first cut of the same asset. One story, several containers.
- Anonymize honestly. "Anonymize it" masks employer and specifics with honest labels. An invented case study isn't a proof asset. It's a liability with formatting.
After the first response: answer the [need: ...] gaps and ask for the finished draft. The entire value of proof is that it's checkable.
9. Claim check: the truth pass before you send
One inflated bullet can cost you an offer three interviews later, when someone finally asks you to explain a number you never earned. This is the final pass that catches it now, while it's still just an edit. If you use only one prompt from this page, make it this one, on everything you're about to send.
Truth-check this draft before I send it. Compare it only against the source material below. The question is not "is this well written" — it is "can I defend every line of it out loud in an interview."
Rules:
- Use no outside knowledge. Do not assume typical tools, likely metrics, or common responsibilities.
- A claim that might be true but is not in my source material is unverified. Flag it; do not excuse it.
- Do not overcorrect truthful strength into vague mush. The goal is the strongest version of what is real.
Deliver exactly this, in this order:
1. THE VERDICT
One of: SAFE TO SEND / VERIFY THE FLAGGED ITEMS FIRST / DO NOT SEND YET — with a one-sentence reason.
2. WOULD NOT SURVIVE A FOLLOW-UP
The claims an interviewer could break with one question. For each: the exact draft text → what is unsupported → the question that breaks it → the fix (remove, soften, or the exact fact I need to confirm).
3. QUIET INFLATION
The subtle upgrades:
- Two separate facts merged into one stronger claim
- "Contributed to" that became "led"
- Dropped qualifiers — "learning," "intermediate," "~1 year"
- Broad tools that became specific ones (AWS quietly becoming Lambda)
- Numbers that appear nowhere in my source material
4. LEFTOVER PLACEHOLDERS
Any [bracketed] items, TBDs, or template text still sitting in the draft.
5. THE CLEAN VERSION
Only if every fix can be made without new facts from me; otherwise list exactly what you need. Never patch an unsupported claim by writing a different unsupported claim.
THE DRAFT I'M ABOUT TO SEND:
SOURCE MATERIAL — WHAT IS ACTUALLY TRUE:
THE JOB POSTING, IF THE DRAFT WAS TAILORED TO ONE:
Make it yours:
- Escalate for high stakes. Say "stricter" and it re-runs treating anything unsourced as a blocker. Use that mode for final-round materials and anything going somewhere public.
- Make it a habit, not an event. Run every final artifact through this before it leaves your hands: tailored resume, cover letter, LinkedIn rewrite, even a prepared interview story.
- Why this exists: AI wrote your draft faster than you could. Somewhere in that speed, "helped migrate" may have become "led the migration." Nobody else will catch that for you until the worst possible moment.
After the first response: confirm or correct the flagged items ("item 2 is true, here's the number") and ask for the clean version.
What this looks like inside the toolkit
What this looks like inside the toolkit. These nine prompts are the single-chat version of the open-source
job-hunt-skillstoolkit. The toolkit is the same philosophy with memory and files: a versioned source resume, a story bank of your best interview material, a folder per application, saved research reports, and a claim-check that verifies drafts against everything you've saved, not just what you pasted into one chat. A good rhythm: start with the prompts above, and when you notice you're re-pasting the same resume for the fifth time, that's the toolkit telling you it's time.
The library is the system
Nine prompts is a lot to run in one sitting, and you shouldn't. The order that works: build or audit the resume once. Then, per opportunity: research the company, tailor, letter if needed, claim-check, send. Interview prep when the call books. Proof assets and LinkedIn in the quiet weeks, because they compound while applications expire.
That last distinction is the one that keeps a search sustainable: applications are perishable, but your source resume, your story bank, your proof assets, and your research notes carry forward to every future application. If your search is burning you out, the sustainable job search system is built around exactly that split, and the pillar essay makes the full case.
And if the roles you're feeding these prompts are remote ones: half the battle is starting from postings that are genuinely remote. That's what the job board is for: every listing is vetted against the company's actual remote culture, so the company-research prompt has less to catch.
