resume-tailor

name	resume-tailor
description	Tailors a résumé to a specific job description — parses the JD for priorities, reorders and rewrites bullets to match, quantifies achievements, and passes ATS filters. Use when a user is applying to a specific role and wants their résumé to match the JD, or when their bullets are weak ("responsible for...") and need tightening.

Résumé Tailor

A résumé that matches the job description beats a generic one. Not by keyword-stuffing; by showing the exact evidence the hiring team is looking for.

When to use

• User is applying for a specific role and wants their résumé tuned to that JD.
• Résumé reads generic ("responsible for") and needs punch.
• User has 5+ years of experience and a résumé overflowing onto page 3 — needs cutting.
• ATS-compatibility is a concern (large companies; public sector; many online portals).

Before you start

Gather:

1. The job description — full text, not a summary. Priorities are buried in specific wording.
2. The current résumé — ideally in editable form (Markdown, Google Doc, Word). PDF only means you'll retype.
3. The user's recent work + achievements that aren't on the résumé yet — often there's a strong story hidden in Slack DMs and project docs.
4. Which companies they'd be most excited about — the one where they'll push hardest and get the most value from tailoring.

Tailoring workflow

1. Parse the JD. Separate must-haves, nice-to-haves, culture signals. See jd-parsing.md.
2. Audit the current résumé against the must-haves. What's already there? What's missing? What's there but buried?
3. Reorder. Most relevant experience first — within the same role, reorder bullets. Across roles, keep reverse-chronological but consider a summary / highlights section at the top.
4. Rewrite weak bullets. Apply STAR (Situation / Task / Action / Result). See bullet-rewrites.md.
5. Quantify everything possible. "Improved performance" < "Cut p95 latency 40% (800ms → 480ms)" < "Cut p95 latency 40%, unblocking SLA target for 3M monthly users."
6. Trim. One page if <10 years experience, two pages max otherwise. Recent and relevant stays; old and tangential goes.
7. ATS pass. Remove fancy formatting, tables, text-in-images, two-column layouts. Single column, clean headers, normal fonts. See ats-formatting.md.
8. Generate a short cover letter / email that leads with the specific reason this role is a match — not a restatement of the résumé.

Non-negotiable rules

• Never fabricate. Achievements, titles, metrics, dates — all real. Fabricated résumés get discovered in reference checks, probation reviews, or worse.
• Lead every bullet with a verb. "Led a team of 5..." not "Was responsible for leading..."
• Quantify every bullet you can. Numbers, percentages, counts, dollars. "Shipped feature" is weaker than "Shipped feature used by 2,400 users in week 1".
• Match the JD's vocabulary — where truthful. If the JD says "partner with stakeholders", use "partnered with stakeholders", not "worked closely with folks".
• Cut filler sections. "Objective" / "References available on request" / generic skills lists — delete. Wasted space.
• One page if possible. Hiring teams spend 30 seconds on first pass. Two pages means they might not reach page 2.
• Never use first person. "Shipped X" not "I shipped X". Not "Yaroslav shipped X" either — implied first person, no pronoun.
• Never use a photo (in the US/UK/Canada). In some European countries and APAC it's expected — match local norm.
• Dates on the right, roles on the left — makes scanning easy for recruiters.

Output

• The tailored résumé — full document, ready to copy into the user's source.
• A change log — what you rewrote, what you cut, what you reordered, and why. So the user can push back on your edits.
• A short cover email/letter — optional but almost always worth including.

References

• Parsing a JD — how to extract must-haves, nice-to-haves, and culture signals; common JD patterns.
• Bullet rewrites — STAR pattern, strong verb lists, quantification tactics, common bullet failures.
• ATS formatting — the layout choices that survive parsing, the ones that don't; Word vs PDF; file-naming.