resumint

Mint perfect resumes from code. One page, every time.

Classic · Executive · Compact · Modern

---

Drop any file. Get 4 resume PDFs + editable source code.

git clone https://github.com/WHATDOESTHEFOXSAY2U/resumint.git
cd resumint
make mint FILE=your_resume.docx

Accepts .docx, .txt, .pdf, .json, or .yaml. If both a .docx and .pdf exist side by side, resumint cross-references them for the best extraction.

Want a company theme? Add THEME=stripe (15+ presets -- run make themes to see them all).

---

The core idea

resumint generates resume code, then validates its own output. The code-to-PDF feedback loop is the product:

  your file (docx, txt, pdf, yaml)
          |
          v
      [ parse ] --> structured YAML (the "code")
          |
          v
      [ build ] --> 4 PDFs + Typst source via Typst
          |
          v
      [  QA   ] --> 11 checks: overlap, margins, alignment, density, ATS
          |
       pass? --yes--> done (PDFs + editable source in output/)
          |
          no
          v
      [ fix  ] --> trim bullets, shorten text, drop sections
          |
          +-------> rebuild (up to 5 iterations until clean)

AI agents generate the code. The QA engine validates it. The fix loop corrects it. Humans get a perfect PDF.

---

Your resume is code

resumint converts your resume into structured data (YAML) and typeset source (Typst). AI agents and humans edit the code. PDFs are the rendered output.

output/
  pdf/                   # what you submit to jobs
    resume_classic.pdf
    resume_executive.pdf
    resume_compact.pdf
    resume_modern.pdf
  code/                  # what you (or your AI agent) edit
    resume_classic.typ
    resume_executive.typ
    resume_compact.typ
    resume_modern.typ
  resume_data.yaml       # the structured source of truth

Edit the YAML or .typ files, run make build, get updated PDFs. Your resume lives as code -- versioned, diffable, and agent-friendly.

---

Why resumint

• Any file, one command -- drop a .docx, .txt, .pdf, or structured data. resumint parses it, builds 4 templates, and QA-checks the output. .docx gives the cleanest extraction; .pdf works but may lose precision on numbers.
• Always fits one page -- the feedback loop trims bullets, shortens text, removes sections, and drops old jobs until every template fits. No manual tweaking.
• Built for AI agents -- point Cursor, Claude Code, Copilot, or any agent at AGENTS.md and let it handle everything. The QA engine gives agents deterministic pass/fail signals.
• 111 tests -- 9 fixture profiles, 12 test categories, covering builds, page counts, ATS extraction, QA checks, theme switching, content trimming, cross-referencing, and docx input.

No API keys. No accounts. Runs locally.

---

Commands

make mint FILE=resume.docx             # the main command -- any file in, PDFs out
make mint FILE=resume.txt THEME=stripe # same, with a company color preset
make build                             # rebuild after editing YAML or .typ files
make qa                                # run 11 layout checks on your PDFs
make fix                               # auto-fix issues + rebuild
make test                              # run the full 111-test suite
make themes                            # list all company presets
make setup                             # first-time setup with example data

---

Contributing

resumint ships with 4 templates and 15 themes. We want 40 of each.

Add a template -- create a .typ file in 1_Templates/, export #let render(data), add the name to 3_Engine/build.py.

Add a company theme -- one line in 3_Engine/build.py:

"tesla": "#CC0000",

Ideas we'd love PRs for:

Two-column layout / Minimal template / Academic CV / Dark mode / Company themes (Tesla, Nvidia, Figma, Vercel, Shopify, Uber, Airbnb, Datadog, Palantir, Jane Street)

See CONTRIBUTING.md for the full guide.

---

Prerequisites

Typst (brew install typst) and Python 3.8+.

---

MIT License · Contribute · Agent docs