pdf

name	pdf
description	Read, create, merge, split, rotate, and extract text from PDF files using SLICC's built-in pdftk and convert commands. No Python required. Use when the user asks to work with PDFs or .pdf files, needs to combine, merge, split, rotate, or extract text/pages from documents, wants to check page count or metadata, or needs to convert a PDF page to an image.
triggers	["extract text from pdf","extract text from this pdf","read this pdf","merge pdfs","merge these pdfs","combine pdfs","combine these pdfs","split pdf","split this pdf","extract pages from pdf","page extraction","rotate pdf","rotate pages","create a pdf","make a pdf","generate a pdf","pdf page count","how many pages in this pdf","pdf metadata","burst pdf into pages","document merging",".pdf file"]

PDF Skill

Work with PDF files using SLICC's built-in pdftk (backed by @cantoo/pdf-lib + unpdf) and convert (magick-wasm) commands.

pdftk Syntax

pdftk uses positional syntax: input file(s) come before the operation.

pdftk <input.pdf> <operation> [options]

For multi-file operations, assign handle labels first:

pdftk A=first.pdf B=second.pdf cat A B output merged.pdf

Operations

Extract text

var r = await exec('pdftk /mnt/file.pdf dump_data_utf8');
console.log(r.stdout);

Extract metadata

var r = await exec('pdftk /mnt/file.pdf dump_data');
console.log(r.stdout);
// Output: NumberOfPages, InfoKey/InfoValue pairs, etc.

Merge PDFs

// Assign handle labels A, B, C — then cat them in order
await exec('pdftk A=/mnt/a.pdf B=/mnt/b.pdf C=/mnt/c.pdf cat A B C output /shared/merged.pdf');

// Verify: confirm page count matches sum of inputs
var check = await exec('pdftk /shared/merged.pdf dump_data');
console.log(check.stdout); // Look for NumberOfPages

open('/shared/merged.pdf', '--download');

Split — extract specific pages

// Extract pages 2–5
await exec('pdftk /mnt/file.pdf cat 2-5 output /shared/pages2to5.pdf');

// Extract a single page
await exec('pdftk /mnt/file.pdf cat 3 output /shared/page3.pdf');

// Extract from page 4 to the end
await exec('pdftk /mnt/file.pdf cat 4-end output /shared/from4.pdf');

// Verify the output page count
var check = await exec('pdftk /shared/pages2to5.pdf dump_data');
console.log(check.stdout); // Should show NumberOfPages: 4

Split — every page into its own file

// Creates /shared/page_01.pdf, /shared/page_02.pdf, etc.
await exec('pdftk /mnt/file.pdf burst output /shared/page_%02d.pdf');

// Verify by listing produced files
var ls = await exec('ls /shared/page_*.pdf');
console.log(ls.stdout);

Rotate pages

// Rotate all pages 90° clockwise
await exec('pdftk /mnt/file.pdf rotate 1-end right output /shared/rotated.pdf');

// Rotate a single page (page 3 only)
await exec('pdftk /mnt/file.pdf rotate 3 right output /shared/rotated.pdf');

// Rotate directions: right (90° CW), left (90° CCW), down (180°)

Create a PDF from scratch

Say "create a pdf [title]" — SLICC will run create-pdf.jsh directly and download the result.

To generate programmatically, run the script via node:

await exec('node /workspace/skills/pdf/scripts/create-pdf.jsh "My Report Title"');

The script produces a 2-page US Letter PDF (612×792pt) with:

• Centered gray header + hairline rule on every page
• 24pt bold Helvetica titles, 13pt body, 72pt margins
• "Page N of M" footer centered at the bottom of every page

Convert PDF page to image

// Convert page 1 to PNG (0-indexed — page 1 = [0])
await exec('convert /mnt/file.pdf[0] /shared/page1.png');
open('/shared/page1.png', '--view');

Error Handling

• File not found: If pdftk reports a file error, double-check the path — uploaded files are typically at /mnt/<filename>, not /tmp/ or relative paths.
• Page range out of bounds: If a requested page range exceeds the document's page count, run pdftk /mnt/file.pdf dump_data first to confirm NumberOfPages, then adjust the range accordingly.
• Empty stdout from dump_data_utf8: If the command returns nothing at all, treat it as a command/file problem first (wrong path, malformed PDF, or pdftk error) — even a scanned PDF normally emits metadata such as NumberOfPages. Check the path and any stderr before assuming OCR is needed.
• No extractable text: If dump_data_utf8 runs cleanly but yields no usable text (or only metadata with no page text), the PDF is likely a scanned image with no embedded text layer. OCR is not available in the current SLICC environment.
• convert not available: The convert (ImageMagick) command requires a tray runtime; it is not available in the browser-only float.

Notes

• All output files should go to /shared/ or /mnt/ — not /tmp/.
• pdftk cat page ranges are 1-based. end means last page: 3-end.
• pdftk burst zero-pads page numbers — use %02d or %03d in the output pattern.
• OCR and PDF form filling are not available in the current SLICC environment.
• For PPTX → PDF conversion with font embedding and layout fidelity, use the pptx2pdf skill.