share-safe 1.1.2

Remove selected PDF metadata and actions, then verify the saved output. Not complete anonymization or content redaction.

Strip the hidden data from a PDF before you send it.

A PDF can carry far more than its pages: an author name, the software that made it, edit dates, XMP metadata, file attachments, and JavaScript. share-safe removes those, re-opens the result to confirm they are gone, and writes a report of exactly what it did and what it could not check.

It runs entirely on your machine. Nothing is uploaded.

npm install

Requires Node 20+. The only dependency is pdf-lib.

Preview first. A dry run changes nothing:

node clean-pdf.mjs examples/sample-dirty.pdf --dry-run --md preview.md

would remove: info=[Title, Author, Creator, Producer, CreationDate, ModDate] xmp=false attachments=1 javascript=0
not checked: annotation authors, form field values, EXIF inside page images

Then clean:

node clean-pdf.mjs examples/sample-dirty.pdf -o contract-share.pdf --report removed.json --md report.md

Pass several files with --out-dir. Each cleaned file keeps its name, and the Markdown report has one section per file:

node clean-pdf.mjs contracts/*.pdf --out-dir contracts-clean --md batch.md
node clean-pdf.mjs contracts/*.pdf --dry-run --md preview.md

The tool refuses to overwrite any existing output or input, including aliases and report paths. Duplicate batch filenames and collisions between output/report paths are rejected before processing. Choose a new output folder for each run. A batch exits with the worst code of any file; successful earlier files remain if a later file fails.

See examples/ for a sample input, its dry-run preview, and the report it produces.

• Document info: all keys, including custom fields; trailer document identifiers
• XMP metadata: metadata references in dictionaries throughout the PDF
• Embedded files: names-tree attachments, associated-file and embedded-file references
• Active actions: JavaScript, open/additional actions and action references throughout dictionaries (including link actions). /A is removed only when it resolves to an action dictionary; legitimate structure attributes and font glyph mappings are preserved.
• Orphan objects: objects no longer reachable after removal, so detached attachment/script payload objects are not serialized into the cleaned file

After saving, it re-opens the output and checks each of these. If anything is left, it throws and does not release the file.

These are listed in every report as not checked. Review them by hand:

• Annotation authors on comments and markup
• Form field values you may have filled in
• EXIF and GPS data inside images placed on the pages

Removing EXIF from embedded images would need a re-encode of each image. That is a planned feature, not a current one.

• Cleaning changes the file. Digital signatures on the original are invalidated by any save. Sign the cleaned copy instead.
• Action removal disables interactive links and action-driven behavior. Save invalidates signatures; verify the visible pages and expected behavior before sending.
• Inputs are limited to regular PDFs under 25 MiB. This is not a parser sandbox; hostile compressed PDFs still need OS memory/time limits.
• verified means the listed removals were rechecked, not malware clearance, complete anonymization or content redaction. Visible text, forms, annotation contents, rich media and image payloads can still disclose data.
• Encrypted PDFs need their password first. The cleaner does not try to open them.

npm test

MIT

Use a supported Node LTS (22 or 24). Run npm ci --ignore-scripts followed by npm run verify. The security workflow runs the regression suite on Linux and Windows; it has not yet been executed remotely for this local change. New report/output files are created exclusively; choose fresh filenames or output folders for repeat runs. See SECURITY.md for the threat model and disclosure guidance.