create-ats-resume

name	create-ats-resume
description	Generates per-job-tailored ATS-friendly resumes and cover letters. Produces single-column HTML, PDF (via Chrome CDP), and DOCX (via python-docx). Reads a user-maintained content database for experiences/skills; extracts keywords from a target job description; outputs are named by {date}-{company}-{role}. Use when the user pastes a job description (or URL) and wants a tailored application package. Do NOT use for the dark-sidebar visual resumes (those are the existing chrisli-resume*.html files).
license	MIT

Create ATS Resume

A skill for producing Applicant Tracking System (ATS)-friendly application packages, tailored per job.

When to use

Use when the user:

• Pastes a job description (or a URL to one) and asks for a tailored resume, cover letter, or both.
• Says things like "tailor this for [role]", "ATS resume for [company]", "make a one-pager for [job]".

Do not use for:

• The dark-sidebar visual resumes (chrisli-resume*.html / .pdf) — those are for human readers, not parsers.

Outputs (one job application = one folder, three artifacts)

For each application, write to the user's resume folder and produce:

<resume-folder>/ats/<YYYY-MM-DD>-<company>-<role>/
├── resume.html       # single-column ATS-safe HTML
├── resume.pdf        # rendered via CDP
├── resume.docx       # rendered via python-docx
├── cover-letter.html
├── cover-letter.pdf
└── cover-letter.docx

<resume-folder> defaults to ~/Downloads/resume. The user can override.

Inputs the user provides

Input	What it is	Required
Job description	Pasted text or URL	✅
Target role title	Job title exactly as posted	✅
Target company	For filename + cover-letter greeting	✅
Content database	User's master experiences/skills file	✅ (uses default)
Resume folder	Where to write outputs and read the content DB from	optional (default: current working directory)

If a URL is provided, fetch it via the web-browser skill first.

Overriding the resume folder

The default is the current working directory (whatever pwd returns at the time you invoke the skill). To use a different folder, pass it on the command line:

/skill:create-ats-resume --resume-folder /path/to/folder <JD>

Rules:

• Use an absolute path.
• The folder must already exist (the skill will create the ats/<date>-<company>-<role>/ subfolder underneath, but not the parent).
• The content DB is read from <resume-folder>/chrisli-content.md (or whatever the user has named it).
• Outputs are written to <resume-folder>/ats/<YYYY-MM-DD>-<company-slug>-<role-slug>/.
• If --resume-folder is omitted, use the current working directory.

If the path contains a tilde (~), expand it to the home directory. Resolve symlinks. Normalize to a clean absolute path before creating the ats subfolder.

Finding the content DB from the CWD

If the user didn't pass --resume-folder and the content DB isn't at <cwd>/chrisli-content.md, walk up parent directories (e.g., <cwd>/.., <cwd>/../..) looking for a chrisli-content.md at each level. Use the first match. This means the user can invoke the skill from any subdirectory of their resume folder (e.g., from inside ats/<some-job>/) and the content DB is still found.

If no chrisli-content.md is found within ~5 parent levels, fall back to ~/Downloads/resume/chrisli-content.md (the legacy default). If that's also missing, tell the user once and ask them to either point you at the right file or run the bootstrap flow from jobsdb-extracted.md.

Master content database

Default location: <resume-folder>/chrisli-content.md. The file holds:

• Contact info
• A base summary variant plus a short note on how to retune
• A categorized skills matrix (Advanced / Intermediate / Tools)
• Work experience — every role with all bullets in a bullets pool (so you can pick what to surface)
• Education, Certifications, Languages
• Optional tailoring notes per role (e.g., "when the JD emphasizes logistics, lead with picker/driver apps")

If the content DB doesn't exist yet, tell the user once and offer to bootstrap it from jobsdb-extracted.md (which is also in the resume folder from an earlier session).

Workflow

1. Resolve resume folder.
   • If --resume-folder was passed, use that (expand ~, absolutize, resolve symlinks). Verify the folder exists; if not, tell the user and ask them to create it or pick a different path.
   • Otherwise, start from the current working directory (CWD) and look for chrisli-content.md in the CWD, then in each parent directory up to 5 levels. Use the first match. If nothing is found, fall back to ~/Downloads/resume/chrisli-content.md and continue the search; if still nothing, tell the user and ask them to point you at the right file.
2. Load content DB from the resolved folder. If missing entirely, bootstrap it from jobsdb-extracted.md (or ask the user).
3. Get the JD. If a URL was given, fetch + extract the description text.
4. Extract keywords from the JD:
   • Required / preferred skills
   • Job title (use verbatim)
   • Tools, technologies, methodologies mentioned
   • Repeated terms (likely important)
   • Acronyms — capture both forms
5. Plan the resume:
   • Choose a summary angle (one tight paragraph mirroring the JD's framing, naming the user honestly)
   • Build a Skills section that mirrors JD terminology (but only the user's real skills)
   • For each role, select the 2–4 most relevant bullets from the pool; reorder them so the strongest match leads
   • Decide which roles to surface vs condense to a one-line earlier-role entry
   • Keep titles, dates, employers exactly as in the content DB
6. Draft the resume HTML following the structure in references/html-structure.md.
7. Draft the cover letter HTML as a one-page, 3–4 paragraph note (see "Cover letter" below).
8. Render PDF + DOCX using the scripts:
   • PDF: node <skill>/scripts/render-pdf.mjs <resume-folder>/ats/<dir>/resume.html
   • DOCX: python3 <skill>/scripts/render-docx.py <resume-folder>/ats/<dir>/resume.html
   • Run them in parallel for the resume and cover-letter.
9. Summarize what you changed vs the base content (which bullets you surfaced, which keywords you mirrored) so the user can sanity-check.

Cover letter (always paired)

• Header: user's contact block + date + company name + "Dear Hiring Team,"
• Paragraph 1: hook — name the role, name the company, one sentence on why this role is interesting to the user
• Paragraph 2: the strongest 1–2 experiences that directly match the JD's top requirements — paraphrase the resume, don't repeat it verbatim
• Paragraph 3: the user's broader differentiator (CTO/founder experience, business grounding, full-stack depth) in one sentence
• Closing: availability, contact, "Best regards," + name
• Length: ~250–350 words

ATS-safe render rules (load references/ats-checklist.md for full details)

The HTML must:

• Single column, no sidebar, no two-column layouts
• No tables, no images, no SVG icons, no text boxes
• No content in headers or footers (ATS often can't read them)
• Section headings, in this order:
  1. Name (h1) + contact line
  2. Summary (p)
  3. Skills (h2) → bulleted list of skill tags
  4. Professional Experience (h2) → chronological, most recent first
  5. Education (h2)
  6. Certifications (h2, optional)
  7. Languages (h2, optional)
• Fonts: Calibri (with Arial fallback). Body 11pt, name 16–18pt, headings 12pt bold.
• Margins: ~0.75" / ~19mm.
• Filename safety: no spaces, ASCII only, slugs from {company}-{role} (kebab-case, no special chars).

Render scripts

Script	Does
scripts/render-pdf.mjs	CDP-based, opens the HTML in a background tab and prints to PDF (one PDF per HTML). Reuses the existing web-browser skill's CDP client pattern.
scripts/render-docx.py	Parses the HTML with html.parser and builds a DOCX with python-docx. Headings map to docx heading levels; <ul>/<li> map to list bullets; <strong> becomes bold runs.

Both scripts accept one or more HTML file paths as args.

References (load on demand)

• references/ats-checklist.md — full ATS rules, sources, and common pitfalls. Read before drafting any HTML.
• references/html-structure.md — the exact HTML skeleton to produce, with allowed tags and CSS defaults.

What this skill does NOT do

• It doesn't auto-apply anywhere. Outputs are local files.
• It doesn't optimize the existing visual resumes. Those are intentionally different and live alongside.
• It doesn't write fictional content. If a JD requires a skill the user doesn't list in the content DB, surface the transferable skill instead and tell the user.

Quick start (from the user's first request)

User: "ATS resume for Acme's senior frontend role. JD: <paste JD or URL>"
You:
  1. Read SKILL.md (this file).
  2. Read references/ats-checklist.md.
  3. Read references/html-structure.md.
  4. Resolve <resume-folder>:
     - If `--resume-folder <path>` was passed, expand `~` and use that.
     - Otherwise start from the current working directory and walk up looking for `chrisli-content.md`. Fall back to `~/Downloads/resume`.
  5. Read <resume-folder>/chrisli-content.md.
  6. Extract keywords from JD.
  7. Plan + draft resume HTML.
  8. Draft cover letter HTML.
  9. mkdir + write 2 HTML files.
 10. Run render-pdf.mjs and render-docx.py (in parallel).
 11. Summarize tailoring choices back to the user.