Watermarks Remover is an agent skill and Python service for removing AI provenance marks from text and files that users own. It cleans invisible Unicode characters, exotic spaces, bidirectional controls, and other hidden text artifacts through deterministic processing. Statistical text watermarks can be addressed through agent rewriting and an optional rewrite hook. The service also strips provenance and metadata such as C2PA, EXIF, XMP, and document properties from many media and document formats. Its agent skill acts as a thin HTTP client, allowing the processing service to run separately without requiring Python inside the agent host. The project supports local operation, optional Docker infrastructure, Cursor integration, and a standard-library Python server.
Features
- Invisible Unicode and text artifact cleanup
- Statistical text watermark rewrite hooks
- C2PA provenance metadata removal
- EXIF, XMP, and document property stripping
- Broad document, image, audio, and video support
- Local HTTP service and agent skill integration