Cover Letter Writer: Tailored, Human, and Under 300 Words

I've read a lot of cover letters. Some as a hiring manager, some as the friend people send drafts to at 11pm before an application deadline. And I can tell you the uncomfortable truth up front: most cover letters don't get read. They get scanned — for a handful of seconds — and then they either earn a real read or they don't.

That's not a reason to skip the cover letter. It's a reason to write one differently.

This skill turns your AI assistant into something closer to a senior recruiter paired with a career strategist. It doesn't start writing when you paste a job description. It starts analyzing — figuring out what the company is actually hiring for underneath the bullet points — and only then does it write a letter that connects your real experience to those priorities. The result is under 300 words, sounds like a human wrote it, and is deliberately structured for the way recruiters actually read: designed for the 6-second scan.

Note the wording. Designed for the scan — not guaranteed to beat it. Nobody can promise you an interview, and anyone who does is selling something. What I can promise is that this skill removes the failure modes that sink 90% of cover letters before a recruiter finishes their coffee.

Let's break down why it works, and then I'll walk you through using it end to end, including a full worked example.

Why most cover letters fail

Before we fix the cover letter, we need to be honest about why the typical one fails. In my experience, almost every weak letter commits at least one of three sins — and most commit all three.

Sin #1: It's generic

You know the letter. It opens with "I am excited to apply for the position of..." and could be sent to fifty companies with a find-and-replace on the company name. Recruiters have read this letter thousands of times. Generic letters don't offend anyone — they just evaporate. They transmit exactly one piece of information: this candidate didn't spend ten minutes thinking about us specifically. For roles with hundreds of applicants, that's disqualifying by default.

Sin #2: It's self-centered

Weak letters are about what the candidate wants: "This role would be a great opportunity for me to grow." "I'm looking for a challenge." "I've always wanted to work in fintech." Here's the thing — the company isn't hiring to fulfill your growth arc. They have a problem, a gap on a team, revenue they need to protect or capture. A strong letter is about what they need and the evidence that you can deliver it. The reader should finish the letter thinking about their problem being solved, not about your career aspirations.

Sin #3: It restates the resume

The third failure is the letter that dutifully walks through the resume in paragraph form: "I worked at X for three years, where I did Y. Before that, I was at Z..." The recruiter has your resume. It's attached. A cover letter that summarizes it adds zero information and costs the reader time. The cover letter's job is different: it's the argument — the connective tissue that explains why these specific experiences make you the answer to this specific opening.

A resume is evidence. A cover letter is the argument. If your letter just repeats the evidence, you've submitted two resumes and zero arguments.

The skill we're building here is engineered against all three sins. It can't be generic, because it starts from the specific job description's priorities. It can't be self-centered, because the writing rules force it to lead with the company's needs. And it can't restate the resume, because it's instructed to select two measurable achievements and connect them to hiring priorities — not to summarize your work history.

How recruiters actually read: the 6-second scan

There's a widely cited finding from eye-tracking research (Ladders ran the best-known study) that recruiters spend roughly six to seven seconds on an initial resume scan. Whether the exact number is six or ten seconds doesn't really matter. The point is that the first pass is a scan, not a read — and the same behavior applies to cover letters, arguably even more brutally, because the cover letter is optional reading.

Here's what a scan-driven reader is doing in those seconds:

• Pattern-matching against their mental checklist. Every recruiter opens a job requisition with three to five things they must see. They're scanning for signals of those things — keywords, titles, numbers, familiar company names.
• Reading the first line, maybe the first two. If the opening line is boilerplate, the scan often ends there. The opening line carries a wildly disproportionate share of the letter's total impact.
• Jumping to numbers. Digits stand out visually in a wall of text. "Reduced onboarding time by 40%" gets seen even in a skim. "Significantly improved processes" does not.
• Checking length. A letter that visibly runs past one screen signals effort for the reader, which reads as a cost. A tight letter signals judgment.

This is why the skill enforces the constraints it does. Under 300 words, because a scan-reader will actually engage with something short. A banned-openers rule, because the first line has to do real work. Two measurable achievements, because numbers survive skimming. One specific company detail, because it proves the letter couldn't have been sent anywhere else.

None of this guarantees the scan goes your way — the underlying fit has to be real. But when the fit is real, these design choices make sure the scan actually finds it. That's the whole game: designed for the 6-second scan.

The core move: analyze hiring priorities first

Here's the single biggest difference between this skill and "hey ChatGPT, write me a cover letter": it is forbidden from writing until it has analyzed.

When you paste a job description into a raw AI assistant and ask for a letter, the model does the statistically obvious thing — it produces a genre-average cover letter, lightly seasoned with keywords from the posting. That's exactly the generic letter from Sin #1, just faster.

This skill inserts a mandatory step in between: before drafting a single sentence, it reads the job description like a recruiter would and extracts the company's likely hiring priorities — the three to five things that will actually decide this hire.

What "hiring priorities" actually means

Hiring priorities are not the same as the requirements list. Job descriptions are noisy documents — half boilerplate, half wish list, often written by committee. The priorities are the signal underneath:

• Repetition is signal. If "cross-functional" appears four times, that team has been burned by silos. That's a priority.
• Position is signal. The first two or three responsibilities are usually the real job. The last five are the wish list.
• Specificity is signal. "Experience with HubSpot workflows" is a real, immediate need. "Familiarity with modern marketing tools" is filler.
• Pain leaks through. Phrases like "comfortable with ambiguity," "able to bring structure," or "thrives in a fast-paced environment" are euphemisms for actual problems: no processes, changing priorities, understaffing. Reading them as problems tells you what a great hire would fix.

Why this changes the output so much

Once the priorities are explicit, the letter almost writes itself — and it writes itself differently. Instead of "here is my background," the letter becomes "you need X; here is proof I've done X." Every sentence gets a job: it either maps an achievement to a priority or it gets cut. This is exactly the structure a scanning recruiter is pattern-matching for, and it's why priority-first letters feel sharp in a way keyword-stuffed letters never do.

This is the same principle behind good context engineering generally: the quality of AI output is determined less by the final instruction and more by the structured thinking you force before it. If that idea interests you beyond cover letters, my context-engineering skill goes deep on it.

The anatomy of a strong opening line

The opening line is worth its own section because, in a 6-second scan, it might be the only full sentence that gets read. The skill hard-bans the two most common openers — "I am excited to apply" and "I am writing to express my interest" — not because they're wrong, but because they're invisible. They spend your most valuable real estate saying nothing.

A strong opening line does one of three things: it leads with a relevant result, it leads with a specific point of connection to the company, or it names the company's problem and positions you against it.

Here are some before/after examples. To be clear, these are hypothetical examples I'm inventing to illustrate the pattern — the skill will generate openers from your actual materials, never from imagination:

Weak: "I am excited to apply for the Customer Success Manager position at Meridian." Stronger: "For the last two years I've managed a book of 45 B2B accounts through a product migration a lot like the one Meridian just announced — and kept churn under 3% doing it."

Weak: "I am writing to express my interest in the Data Analyst role." Stronger: "Your posting mentions turning messy operational data into decisions leadership actually uses — that's been the through-line of my last two roles."

Weak: "With five years of experience in marketing, I believe I would be a great fit for your team." Stronger: "I noticed Brightline is expanding into the DACH market; I spent 2024 building the paid acquisition program for exactly that expansion at my current company."

Notice what the stronger versions share: a specific fact, a direct link to the company, and zero throat-clearing. Notice also what they don't do — they don't invent enthusiasm, and they don't make claims that the resume can't back up. Every stronger opener above is built from a verifiable fact about the (fictional) candidate.

One more thing: the skill delivers a stronger alternate opening line with every letter. Openers are high-variance — sometimes the second option is clearly better, sometimes it just fits your voice more. Having two lets you choose instead of settle.

Measurable achievements beat adjectives

Adjectives are claims. Numbers are evidence. Recruiters are professionally allergic to claims, because every candidate makes the same ones. This is why the skill requires two measurable achievements pulled from your resume whenever your resume contains them — and why it bans the adjective clichés outright.

Here's the translation table the skill effectively applies:

Cliché (banned)	What to write instead
"I'm a hard worker"	A result that required sustained effort: "shipped the migration two weeks early despite a mid-project scope change"
"I'm a team player"	A concrete collaboration: "worked with design and legal to launch the feature across 3 markets"
"I'm a fast learner"	Proof of learning speed: "picked up SQL on the job and was writing production queries within a month"
"I'm detail-oriented"	A detail that mattered: "caught a billing config error that would have overcharged 200+ customers"
"Proven track record"	The actual track: "grew organic traffic 60% year over year, two years running"
"Passionate about X"	Evidence of the passion: a project, a certification, a side build — something that exists

The pattern is always the same: replace the self-assessment with the event that would make someone else assess you that way. If your resume says you reduced ticket resolution time by 35%, the letter should say that — number intact — and tie it to the hiring priority it serves.

What if my resume has no numbers?

This is common, and the skill handles it honestly: it uses the strongest concrete achievements available — scope, scale, outcomes, before/after states — without fabricating metrics. "Managed the region's largest client account through a contract renewal" is measurable in spirit even without a percentage. What the skill will never do is invent "increased efficiency by 27%" to fill the gap. Which brings us to the rule I care about most.

The ethics rule: never invent facts

Let me be blunt about this one, because it's where generic AI cover letters go from mediocre to dangerous.

Language models fill gaps. Ask an unconstrained model to write an impressive cover letter from a thin resume, and it will helpfully improve reality — a plausible-sounding metric here, a leadership claim there, a familiarity with tools you've never opened. The letter reads great. Then one of three things happens:

1. The interview. The interviewer asks about the "40% cost reduction" in your letter. You've never seen that number before, because the model made it up. You now get to choose between improvising a lie and admitting your application materials contain fiction. Both end the interview in spirit, even if it continues in form.
2. The background check. Larger companies verify claims. A fabricated title or invented company fact discovered late in the process doesn't just cost you this job — recruiters move between companies and remember names.
3. The offer that fits the fiction. Worst case, you get hired for the invented version of you, into a role calibrated to skills you don't have. That's a slow-motion failure with your name on it.

So the skill treats fabrication as a hard constraint, not a style preference: it never invents numbers, company facts, or experience. Every claim in the output must trace back to something you provided — your resume, your notes, the job description. If the model can't find a second measurable achievement, it works with what exists rather than manufacturing one. If you didn't provide company research, it doesn't hallucinate a recent funding round to name-drop.

This is also why the skill asks for your materials as inputs rather than "improving" from thin air. Its job is selection and framing of true things — which, honestly, is what great cover letter writing always was.

The 30-day sentence

There's one structural element in every letter this skill produces that deserves explanation: exactly one sentence about what you'd focus on in your first 30 days.

Why one sentence, and why at all?

Because it flips the letter's tense. Everything else in a cover letter is past tense — what you did, where, with what results. The 30-day sentence is the only line written in the future tense of this job, and that does three quiet but powerful things:

• It proves you understood the role. You can't name a sensible first-30-days focus without having actually parsed what the job needs. It's the hardest line in the letter to fake, which is exactly why it stands out.
• It lets the reader picture you in the seat. "In my first 30 days, I'd want to map the current handoff between sales and onboarding and find where deals stall" — the hiring manager reads that and involuntarily imagines you doing it. That mental image is worth more than three paragraphs of qualifications.
• It signals ownership without arrogance. One sentence is confident. Three sentences of "my 90-day plan" in a cover letter is presumptuous — you don't have enough information yet, and a good hiring manager knows it. The skill caps it at exactly one sentence for precisely this reason.

The skill derives this sentence from the hiring priorities it identified in step one, so it's always anchored to what the job description actually emphasizes — never a generic "I'd focus on learning the ropes and adding value."

How to use the skill, step by step

Time to get practical. Install it once:

npx skills add alexandrrich/skills --skill cover-letter-writer

Then each application follows the same five-minute loop.

Step 1: Gather your four inputs

The skill works from a simple fill-in template. Paste this and complete it:

Company: <company name>

Company research / notes (optional):
<anything you know: recent news, product details, why this company,
 a person you spoke to, mission points that resonate — or leave blank>

Job description:
<paste the full posting>

My resume:
<paste your resume as plain text>

Two notes on inputs. First, the company research field is optional but valuable — if you give it even one real detail ("they just launched in Brazil," "I use their API at my current job"), the skill will work exactly one specific company reference into the letter, which is one of the strongest anti-generic signals available. Second, paste the full job description, boilerplate and all. The noise is data — remember, the skill reads repetition and position as signals.

Step 2: Review the priority analysis

Before you see a letter, you'll see the skill's read of the job: the three to five hiring priorities it extracted, and which of your achievements it plans to map to them. Read this part. If the analysis is off — it overweighted a nice-to-have, or missed that the role is really about client retention — say so, and it will redo the mapping. Correcting the analysis is far more efficient than correcting the prose.

Step 3: Receive the three deliverables

Every run produces:

1. The final cover letter — under 300 words, all rules applied.
2. A stronger alternate opening line — a second option for the highest-stakes sentence.
3. A short version under 180 words — for application forms with a "why do you want to work here?" text box, email bodies to a hiring manager, or LinkedIn messages. Same argument, compressed.

Step 4: Edit for your voice

This matters: the output is a strong draft, not a finished artifact. Read it aloud. Swap in your natural phrasing where the letter sounds more polished than you'd ever be. Verify every fact against your resume one final time — you are the last line of defense on accuracy, and you're the one who'll be asked about it in the interview.

Step 5: Send it as text, not as an attachment (usually)

Unless the application specifically requests an attached document, a cover letter pasted into the email body or the application's text field gets read more often than a PDF that requires one more click. Reduce friction for the scanner.

A worked example: input to output

Let's make this concrete with a fully fictional example — invented candidate, invented company, invented job posting. This is illustrative only.

The setup. Maya Torres is a customer support team lead applying to "Northwind Labs" (fictional), a B2B SaaS company hiring a Customer Success Manager. Her notes mention Northwind recently launched a self-serve tier. Her resume includes: led a 6-person support team; cut average first-response time from 9 hours to 2; built a help-center that deflected ~30% of tickets; ran onboarding calls for the 50 largest accounts.

The skill's priority analysis (abridged):

Hiring priorities detected in the job description:
1. Reduce churn in mid-market accounts ("retention" appears 3x;
   listed first under responsibilities)
2. Scale success processes without scaling headcount
   ("do more with less", "self-serve motion" — matches their
   new self-serve tier launch)
3. Own onboarding for key accounts (specific, named tools)

Mapping:
- Priority 2 → help-center ticket deflection (~30%)
- Priority 1/3 → onboarding calls for top 50 accounts
- Supporting → response time 9h → 2h (operational credibility)

The final letter the skill produced (287 words in full; opening excerpted here):

Dear Northwind Labs team,

When your posting says you want to scale customer success without
scaling headcount, that's the problem I've spent the last two years
solving: the help-center program I built now deflects roughly 30%
of inbound tickets, which is what let a six-person team keep
first-response times at two hours while our customer base grew.

Your launch of a self-serve tier makes that experience directly
relevant. Self-serve customers still churn when they hit friction
silently — the difference is whether someone has built the systems
that catch it. [...]

In my first 30 days, I'd map where your mid-market accounts
currently stall between signup and first value, and pick the one
handoff most worth fixing. [...]

Maya Torres

The alternate opening line it offered:

"I ran onboarding for our 50 largest accounts through a pricing
change — the same retention pressure your CSM role is being hired
to own."

Notice what happened: the letter never mentions Maya being a "people person" or "passionate about customers." Every line is a true fact from her materials aimed at a detected priority, the self-serve tier detail proves the letter is Northwind-specific, and the 30-day sentence is one line, derived from priority #1. That's the method, executed.

Customizing per industry

The core method is industry-agnostic — priorities, evidence, brevity work everywhere — but the texture should shift. A few adjustments worth making:

• Tech and startups: Lean harder on metrics and shipping. The 30-day sentence can be more concrete and tactical. Slightly informal is fine; the banned-cliché rule matters most here because tech recruiters see the most AI-generated slop.
• Finance, legal, government: Keep the confident tone but raise the formality a notch. Precision language over energy. The "no invented facts" rule isn't just ethical here — it's compliance-adjacent.
• Creative roles: The letter itself is a writing sample. The alternate opening line deliverable earns its keep — pick the bolder one. Voice matters as much as evidence.
• Healthcare, education, nonprofits: Mission alignment is a real priority in these postings, not filler. This is where the optional company-research field does the most work — one genuine sentence about why this org lands harder than in any other sector.
• Career changers: Tell the skill explicitly in your notes ("I'm moving from teaching into L&D"). It will frame transferable achievements against the priorities instead of apologizing for the gap — the framing difference between "although I lack direct experience" and "here's the same problem solved in a different setting."

Common mistakes (and when not to send a letter at all)

Even with a good tool, a few habits sink people. The short list:

1. Skipping the review step. The skill's priority analysis is a hypothesis. You have context it doesn't — correct it before the draft, not after.
2. Sending the first output for every job. The skill makes tailoring cheap; it doesn't make it optional. Rerun it per job description. A tailored letter reused generically is just a slower generic letter.
3. Padding it back up. You'll be tempted to add "just one more paragraph" about that other experience. Resist. The 300-word ceiling is the feature. If it didn't map to a priority, it's cut for a reason.
4. Letting the AI voice through. If a sentence sounds like something you'd never say, change it. Recruiters in 2026 have finely tuned AI-detection instincts, and the tell is almost always vocabulary, not content.
5. Forgetting the fact-check. The skill won't invent facts — but it can only be as accurate as the resume you pasted. If your resume has a stale number, the letter will faithfully repeat it.

And sometimes, don't send one at all:

• The application says "optional" and the role gets 500+ applicants for a high-volume position. Warehouse, retail, high-volume support: resumes are often machine-screened and letters unread. Spend the time on more applications.
• A referral is carrying your application. A two-line note to the hiring manager via your referrer usually outperforms a formal letter.
• The form has no field for it. Don't wedge a cover letter into a "additional information" box designed for visa status. Use the 180-word short version only if there's a genuine "why us?" prompt.
• You'd be sending it purely out of guilt. A letter that exists to check a box reads like it. When you do send one, send one with an argument.

Resources & links

The skill file below is the complete, runnable SKILL.md — download it, drop it into your agent's skills directory, or install with npx skills add alexandrrich/skills --skill cover-letter-writer.

Related material on this site:

• Context Engineering — the deeper principle behind "analyze before you write": structuring what the model knows before asking it to produce.
• SEO Longform Writer — the same constraint-driven writing approach, applied to long-form content instead of 300-word letters.
• Grill Me — pressure-test yourself before the interview your new cover letter earns; it plays the skeptical interviewer.
• Productized AI Service — if you're a coach or freelancer, this cover-letter workflow is exactly the kind of repeatable deliverable you can productize.
• Prompt Engineering: The Complete Guide (2026) — why hard rules and banned phrases produce better output than "make it good," explained from first principles.
• The AI Content Creation track — the full learning path this skill belongs to.
• Make Money with AI in 2026 — for readers on the other side of the table: turning skills like this one into services.

External references worth your time:

• Ladders' eye-tracking research — the origin of the "6-second scan" finding on how recruiters review applications.
• Harvard Business Review: How to Write a Cover Letter — a classic, still-sound editorial take on letters that lead with the employer's needs.
• Ask a Manager's cover letter archive — Alison Green's long-running, hiring-manager-eye view of what actually reads well.