jokerman89 /make-pdf
PDF generation via the same managed Chromium that powers /browse. Accepts URLs, local HTML, or markdown (converted to HTML first via a deterministic toolchain). Use for deliverables: rendered design docs, customer-facing one-pagers (after the active pack's compliance gates), printable runbooks.
When to use
- Operator wants a PDF copy of a
/office-hoursdesign doc for distribution - Customer-facing deliverable that's been voice-checked and needs to leave the chat as a static artifact
- Print-friendly version of a runbook or onboarding doc
- Snapshot of a rendered page for archival (combine with
/scrape --difffor monitoring)
When NOT to use
- Live preview during authoring — use the IDE's markdown preview
- Multi-page report assembly with TOC + cross-refs — out of scope. Use a dedicated typesetting tool.
- Customer-bearing data not yet voice-checked — STOP. Run the active pack's compliance gates first (if any are configured).
Inputs
- Required
--input <path-or-url>—.md,.html, orhttp(s)://... - Optional
--out <path>— output PDF path (default:<input-stem>.pdfin cwd) - Optional
--format <a4|letter|custom>— paper size (default: a4) - Optional
--orientation <portrait|landscape>— default portrait - Optional
--header-footer <yaml>— declarative header/footer (page number, title, date) - Optional
--print-css <path>— override print stylesheet - Optional
--no-background— skip CSS backgrounds (smaller, B&W-friendly)
Workflow
- Resolve input. URL → fetch via Playwright. Local file → check extension.
- Markdown path. If
.md: convert to HTML using a deterministic pipeline (pandoc if available, else markdown-it). Apply default print CSS unless overridden. - Compliance gate. If input is a URL matching Layer 2 prod-host list: BLOCK unless overridden.
- Render. Launch headless Chromium, navigate to file:// or http(s)://, wait for
networkidle, render PDF with the requested format. - Verify. PDF written + readable + non-zero bytes. If less than 5KB: warn (likely blank).
- Report. Path, file size, page count.
Report format
Make-PDF: design-doc-v2.md → design-doc-v2.pdf
Input: 14KB markdown, 1 image embedded
Render: 2.3s (chromium headless, A4 portrait)
Output: 142KB PDF, 8 pages
Path: ./design-doc-v2.pdf
Compliance integration
- Layer 2 customer-data gate on the input (markdown content scanned for customer-data patterns; URL hostname checked against prod list).
- Output PDFs land where operator specified — they are NOT auto-uploaded anywhere. Distribution is the operator's responsibility.
- If customer-facing voice content detected in input markdown: surface reminder to run the active pack's compliance gates if not already done.
Voice tier note
voice: internal. The skill itself is internal-voice; the PDF contents may use a customer-facing voice tier or internal — the skill does not transform voice.
Failure modes
- Markdown converter not installed (no pandoc, no markdown-it): report missing toolchain + bail. Do not silently degrade to a worse converter.
- Page render timeout (60s default): capture whatever's there, warn about partial render.
- Output PDF < 5KB: warn explicitly — almost certainly an empty or broken render.
- Print CSS reference broken: fall back to default print stylesheet, log warning.
- Headless Chromium crash mid-render: retry once. Second crash: report + bail.
Examples
Markdown to PDF:
> /make-pdf --input design-doc.md
✓ 142KB PDF, 8 pages → ./design-doc.pdf
URL to PDF with header/footer:
> /make-pdf --input https://docs.example.com/runbook --header-footer hf.yaml
hf.yaml:
header: "Runbook — confidential — {{date}}"
footer: "Page {{page}} of {{total}}"
✓ 312KB PDF, 24 pages → ./runbook.pdf
Letter landscape:
> /make-pdf --input report.html --format letter --orientation landscape
✓ 87KB PDF, 4 pages → ./report.pdf
See also
/browse— page rendering without PDF output/scrape— extracting structured data instead of producing PDF- The active pack's compliance gates — voice gate before customer-facing PDF leaves
/design-html(batch 7) — generate the HTML that feeds this skill