export-pdf

Export PDF

Convert an HTML CV to PDF using headless Chromium; every workspace PDF should use this skill for consistent rendering.

When to use

• After any edit to a CV HTML file
• When regenerating the master PDF after the master HTML changes
• When initial scaffolding needs a PDF export
• Invoked manually by the user after editing HTML. Also called by prepare-to-send to verify PDF freshness. NOT called by new-application (user tailors HTML first, then runs export-pdf).

Inputs

• HTML file (argument or inferred): path to the source HTML. If omitted:
  • If exactly one *_CV.html exists in the current working directory, use it.
  • If multiple, ask the user which one.

Preconditions

1. Chromium available

Check in order:

which chromium || which chromium-browser || which google-chrome || which 'Google Chrome' || ls /Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome 2>/dev/null

Prefer this order on macOS:

1. /Applications/Google Chrome.app/Contents/MacOS/Google Chrome
2. /Applications/Chromium.app/Contents/MacOS/Chromium
3. chromium on PATH
4. chromium-browser on PATH

If none found, fail loudly:

ERROR: No Chromium-based browser found. Install Chrome (brew install --cask google-chrome) or Chromium (brew install --cask chromium).

Do NOT silently fall back to another PDF tool — cross-application PDF consistency is critical.

2. HTML file exists and is readable

Fail fast if not.

Workflow

1. Resolve paths

• HTML: absolute path.
• PDF: same directory, same stem, .pdf extension.
• If a PDF already exists, note that it will be overwritten.

2. Run Chromium headless

Use scripts/html-to-pdf.sh which wraps the browser invocation. Call it with absolute paths:

bash ${PLUGIN_ROOT}/skills/export-pdf/scripts/html-to-pdf.sh <html-absolute-path> <pdf-absolute-path>

Where ${PLUGIN_ROOT} resolves to the plugin's root directory.

3. Verify output

• PDF file exists at the target path
• File size > 1KB (anything smaller is a failed render)
• If exiftool is installed, print a quick summary of PDF metadata so the user sees what leaked in

3b. Sanity-check PDF content

Use Read to inspect the produced PDF text. These strings almost certainly mean Chromium rendered a browser error page rather than the CV:

• This page isn't working
• ERR_CONNECTION_REFUSED
• chrome-error://
• This site can't be reached
• HTTP ERROR

If any such string is found, fail immediately:

ERROR: PDF content suggests render failure; re-run export-pdf or inspect HTML.

Do NOT report success or proceed to scrubbing if this check fails.

4. Scrub metadata

Auto-invoke scrub-pdf-metadata on the produced PDF as the final step:

$job-hunt-toolkit:scrub-pdf-metadata <pdf-absolute-path>

Every exported PDF is scrubbed, even when attached directly without prepare-to-send.

5. Report

✓ Exported: <html-filename> → <pdf-filename>
  Size: <bytes>
  Metadata scrubbed.

Hard rules

• Use Chromium every time. Never fall back to weasyprint or wkhtmltopdf; prompt the user to install Chrome or Chromium if missing.
• Use absolute paths. Chromium's --print-to-pdf writes to CWD otherwise, which is unpredictable across tool calls.
• Always scrub metadata after export. Auto-invoke scrub-pdf-metadata; prepare-to-send also verifies scrubbing.
• Warn if the HTML has never-rendered markers like TODO, <!-- draft -->, [placeholder] — these will appear in the PDF unless scrubbed at HTML level.

Error handling

Scenario	Action
Chromium not found	Fail loudly with install instructions
HTML file missing or unreadable	Fail loudly
Chromium exits non-zero	Print stderr, fail loudly
Output PDF < 1KB	Treat as failure; delete partial PDF
Target PDF is read-only / directory not writable	Fail loudly

After export

Remind user:

1. Visually review the PDF (open it, check layout)
2. Metadata has already been scrubbed by this skill
3. Run $job-hunt-toolkit:prepare-to-send before attaching to any application for a final freshness and content check