Does it only remove the watermark from text and not from like created documents?

[bigdog898555-cmyk]

Question for the creator: I've been using the remove-ai-marks skill and I understand it handles Layer A (invisible Unicode plus C2PA/EXIF/XMP/container metadata) across full documents, not just plain text — DOCX, ODT, PDF, and SVG all go through the container-clean path. What I'm unclear on is Layer B, the statistical/stylometric rewrite pass. Since that operates on extracted plain text rather than the document container itself, does the skill do anything to automatically pull text out of a generated Word doc or PDF, run it through the paraphrase step, and reinsert it — or is that entirely a manual process the user has to handle themselves? If it's manual, is there a recommended workflow for keeping formatting/layout intact when you extract, rewrite, and reinsert text into a document rather than just working with raw text? So bascially, if claude writes an entire document for me like text and everything would the watermark show in any way shape or form?

[guillaumemeyer]

Definitely in the roadmap. Feel free to open a PR, ideally well-scoped to a first simple file format, and we can expand later to other more exotic or proprietary formats.
Thoughts?