resume-studio

A Claude skill that turns everything you've ever done at work into resumes that win interviews.

Most resume tools polish words. This one fixes the real problem: you don't remember half of what you achieved, and a generic resume gets ignored. resume-studio builds a permanent memory of your career, digs out the achievements you forgot, and tailors a fresh resume for every job you apply to.

What it does

1. Remembers your career. Feed it anything — old resumes, your LinkedIn, performance reviews, or just messy notes. It builds a career-corpus.md file: your full professional history, with every number tied to where it came from. This file is yours, it's plain text, and it grows every time you use the skill.

2. Interviews you like a great recruiter. It drafts first, then asks a few sharp questions about the weak spots: "What was the number before you improved it?" "What did your manager praise you in your last review?" A few answers at a time, never a long form.

3. Tailors a resume for each job. Paste a job description. It researches the company, maps your history against what the role actually needs, and produces a clean, ATS-safe resume (.md + .docx) — plus a report that explains every change and honestly lists what the job wants that you don't have. No hiding, no padding.

4. Never makes things up. Every number on your resume traces back to something you said or shared. If there's no evidence, it asks you — it doesn't invent.

5. Learns your taste, permanently. Say "never use em dashes" or "that number was actually 3.8%" once — it's saved and applied to every future resume. Corrections never need repeating.

Set it up — 2 clicks

  1. Download resume-studio.skill from this repo (click the file, then the download icon).
  2. Drop it into any Claude chat and click Save skill.

Even lazier: paste this into Claude and let it set itself up:

Set up the resume-studio skill from https://github.com/Sidgit11/resume-studio — follow the repo's AGENT_SETUP.md.

Start — 1 message

Build my professional profile. Here's my old resume and some notes about what I actually did: …

That's it. Claude takes over from there.

How a typical journey goes

  1. Dump everything. Old resumes, LinkedIn PDF, performance reviews, appraisal letters, even a rambling voice note about your proudest project. The messier the better — more raw material means a better resume.
  2. Get your career file. Claude drafts career-corpus.md immediately, with weak spots visibly marked, and asks you a handful of questions to fill them. Answer what you can, skip what you want.
  3. Apply to a job. Paste the JD (plus the hiring manager's LinkedIn post if there is one) and say "tailor my resume for this." You get the resume, the Word file, and the change report.
  4. Correct anything, once. Style, facts, framing — every correction is remembered for all future resumes.
  5. Next application takes minutes, because your career file already exists and keeps getting richer.

Tips for the best results

  • Share performance reviews and self-assessments — they're the single best input, and the one nobody thinks of.
  • Keep your resume-studio/ folder somewhere permanent (a connected local folder, a synced drive). That folder is your career memory — as long as Claude can see it, everything carries over between sessions.
  • The resume rules inside are research-backed (ATS parsing mechanics, recruiter surveys, role-specific norms — see references/resume-playbook.md), including the myth-busting: ATS systems almost never auto-reject resumes; humans do.

What's inside

Path What it is
SKILL.md The skill's logic: modes, protocols, guardrails
references/resume-playbook.md The research: ATS mechanics, formatting rules, role-specific playbooks, debunked myths
references/corpus-schema.md File formats for your career memory
references/question-bank.md How the interview questions work
scripts/build_docx.py Renders your resume to an ATS-safe Word file
resume-studio.skill The one-click install file
AGENT_SETUP.md Instructions for a Claude agent to install this automatically
evals/ Test cases used to benchmark the skill

License

MIT. Contributions welcome — especially role-specific additions to the playbook. If Claude keeps repeating a mistake your style rules should prevent, that's a bug: open an issue.