generate-resume

generate-resume

Turn resume.json into a finished résumé file. Pure sh + awk splice, zero Node, zero network — the browser does the actual rendering when the user prints to PDF. This is the mechanism /cv-make drives; use it directly if you're asked to render outside that command's flow (e.g. from /cv-tailor in a later plan).

Inputs

• The active résumé: resume.json at the repo root, valid against ${CLAUDE_PLUGIN_ROOT}/schema/json-resume.schema.json (see the validation checklist in commands/cv-new.md — basics.name required, etc.). Stop and report the bad field rather than rendering invalid data.
• The active template slug: line 1 of cv/.active (default classic-ats if unset). Its template.html lives at "${CLAUDE_PLUGIN_ROOT}/templates/<slug>/template.html".
• The shared renderer: "${CLAUDE_PLUGIN_ROOT}/templates/_shared/render-core.js" — one renderResume(resume) -> htmlString function plus its helpers (escapeHtml, formatDate, dateRange, per-section renderers). Every template reuses this same file; a template only supplies its own CSS shell.

Procedure

1. Resolve and validate the inputs above.

2. Splice the three pieces into one file with the literal splice script:

   mkdir -p output
   sh "${CLAUDE_PLUGIN_ROOT}/lib/render.sh" \
     "${CLAUDE_PLUGIN_ROOT}/templates/<slug>/template.html" \
     "${CLAUDE_PLUGIN_ROOT}/templates/_shared/render-core.js" \
     resume.json > "output/resume-<slug-of-name>.html"
   

   See references/render-pipeline.md for exactly how the splice works, the two markers it fills, and why the result must stay self-contained.

3. Write the result to output/resume-<slug-of-name>.html (kebab-case the person's name, e.g. "Alex Rivera" → alex-rivera).

4. Report the saved path and hand the user the render contract: open the file in a browser, then Print → Save as PDF. The browser is the renderer; this skill only assembles the HTML it renders.

Notes

• Never write partial output. If the splice or a downstream step fails, don't leave a half-written file in output/.
• The output file must open correctly directly from disk (file://) with no server and no additional requests — that's what "self-contained" means here.
• This skill never produces a PDF itself. PDF generation via a headless browser/OS tool is optional and out of scope for Plan 1 (see Plan 5 in the project plan); the guaranteed, zero-dependency path is always the user's own browser doing Save as PDF.