{
"name": "resume-beautifier",
"description": "Create polished, privacy-safe A4 resume designs from structured resume content or existing resume files. Use when Codex needs to extract resume facts, remove personal example data, choose a visual style, generate an editable HTML resume, add client-side PNG\/PDF export, audit for private information before open sourcing, or prepare a reusable resume-design workflow."
}
Resume Beautifier
Use this skill to turn resume content into a polished A4 HTML resume while keeping personal data out of reusable templates and open-source artifacts.
Core Workflow
Extract facts without rewriting truth.
Preserve names, dates, awards, roles, schools, companies, and claims exactly unless the user asks to edit them.
If the input is .docx, use the documents skill or python-docx to extract paragraphs/tables first.
If the input is HTML, treat the visible resume content as source data and ignore export UI/script controls.
Normalize into JSON.
Use references/resume_schema.md for the expected structure.
Keep personal content in the JSON input/output, not in the skill itself.
For open-source examples, replace real personal details with fictional placeholders.
Choose a style.
Use references/theme_tokens.md for available brand-inspired style tokens.
Do not claim a theme is an official brand implementation unless the user supplied the original brand guidelines or DESIGN.md.
Prefer restrained themes for formal resumes; use expressive themes only when the user's target audience fits.
Generate HTML.
Run scripts/build_resume_html.py with a resume JSON file.
The generated HTML is a self-contained A4 page with print CSS and client-side export buttons.
If a headshot is included, embed it as a data URI so export works from file://.
Treat export runtime assets the same way: if the page depends on html2canvas, download that browser-side script once and embed it into the delivered HTML instead of relying on CDN loading at open time.
Export and verify.
Treat self-contained export as a hard requirement for local file:// delivery. Do not assume remote CDN access will be available when the user opens the HTML later.
Prefer browser-native rendering of the HTML. Avoid separately redrawing the resume in another PDF library unless browser export is blocked.
Use the HTML toolbar to export PNG/PDF from the rendered page, or use browser automation when available.
Visually inspect the result: no clipping, overlap, unexpected page breaks, stray design credits, or distorted headshots.
Before handing off the result, explicitly remind the user whether html2canvas has already been embedded into the HTML. If it has not, warn clearly that file:// export may fail, buttons may appear unresponsive, or PDF export may produce tainted-canvas errors.
Audit privacy before packaging.
Run scripts/audit_pii.py on skill/repo files before publishing.
Remove real names, phone numbers, emails, addresses, portraits, and bespoke career facts from examples/templates.
Non-Negotiable Export Rule
For resumés delivered as local HTML files, do not leave html2canvas as a runtime CDN dependency.
Download the browser-side html2canvas script once, embed it into the output HTML, and keep the export path self-contained.
If a session ends before this is done, say so plainly in the final handoff instead of implying export is ready.
Scripts
scripts/build_resume_html.py: build a self-contained A4 resume HTML from JSON.
scripts/audit_pii.py: scan files for likely emails, phone numbers, ID-like numbers, and user-provided denylist terms.
References
references/resume_schema.md: JSON schema and content mapping guidance.
references/theme_tokens.md: available theme names and token rules.
references/export_workflow.md: export and verification workflow for HTML, PNG, and PDF.
Privacy Rules
Never commit a user's real resume, photo, or contact details into the skill package.
Keep examples fictional.
If deriving the skill from a live resume, explicitly state that the reusable skill has been stripped of personal content.
If publishing to GitHub, run the PII audit against the whole repo and show the result.
