Welcome! Type "help" for available commands.
$
Summary
Python 3.10+ stdlib scripts and agent skill that strip multi-vendor AI provenance marks—Unicode text carriers, token-sampling watermarks, and C2PA/metadata—from text, images, and documents (PNG, JPEG, SVG, PDF, DOCX, HTML, Markdown).
Targets Claude, Gemini/SynthID, OpenAI, and open-LLM marking schemes via three layers: deterministic text cleaning, statistical rewrite hooks, and file metadata removal. An optional external SynthID pixel scorer reports confidence scores but does not remove pixel watermarks.