What is watermarks-remover?
watermarks-remover is an open-source project combining agent skills with a standard-library Python service to remove AI provenance marks from owned content. The repository describes the purpose as privacy and hygiene, covering both text and files rather than a single watermark format or vendor.
watermarks-remover separates cleaning into distinct mechanisms. Deterministic scripts address invisible Unicode and related characters, while statistical text watermark handling relies on agent rewriting and an optional hook. File cleaning targets provenance and metadata surfaces, including C2PA, EXIF, XMP, and document properties.
What watermarks-remover can do
watermarks-remover covers text characters, statistical watermark classes, and metadata across several file families. The documented ecosystems include Claude, Gemini and SynthID-Text, OpenAI provenance surfaces, and open-LLM watermark approaches. These are described as class-level targets, with different cleaning mechanisms used for different kinds of marks.
- watermarks-remover cleans invisible Unicode, exotic spaces, bidirectional characters, and tag characters.
- watermarks-remover uses agent rewriting for statistical text watermarks.
- watermarks-remover targets C2PA, EXIF, XMP, and document properties.
- watermarks-remover supports formats including PNG, JPEG, PDF, DOCX, HTML, MP4, WAV, and FLAC.
- watermarks-remover exposes health, inspection, and cleaning endpoints.
- watermarks-remover provides file hooks with check and clean modes.
How watermarks-remover works
watermarks-remover uses a thin-client skill that calls the cleaning service over HTTP, so the agent host does not need Python. Users install the skill, start the service, and configure WATERMARKS_SERVICE_URL when using a different service address. The client also supports WATERMARKS_SERVER_API_KEY when authentication is required.
watermarks-remover provides installation targets for personal and project-level Claude Code and Codex setups, plus Cursor and uploadable Cowork bundles. The Claude Code plugin installs through a marketplace workflow. A separate clean-user-facing-text skill includes its own scripts and does not need the service, but handles text only.
Who watermarks-remover is for
watermarks-remover is intended for people cleaning content they own, including developers working with coding assistants and users preparing authorized text. The optional text-only skill specifically names manuscripts, documentation, and web copy. File-based workflows also suit users who need provenance checks during agent editing or before changes enter Git.
- watermarks-remover can clean authorized manuscripts, documentation, and web copy.
- watermarks-remover can inspect files immediately after supported agent editing operations.
- watermarks-remover can support provenance checks before files enter Git.
- watermarks-remover can clean metadata from supported owned documents and media.
watermarks-remover licensing and access
watermarks-remover is available as a public GitHub repository under the MIT license. The source describes installation through repository scripts, a Claude Code plugin marketplace, and skill bundles for upload. The full service-backed skill requires users to run a reachable service; the self-contained text-only skill offers a different setup path.
watermarks-remover strengths and limitations
watermarks-remover combines broad file-format coverage with explicit separation between deterministic cleaning and instruction-based rewriting. The file hook defaults to checking rather than changing content. In clean mode, the hook swaps a temporary file only when a real difference exists, preserving modification times for files that were already clean.
- watermarks-remover requires a reachable HTTP service for the main skill.
- watermarks-remover cloud workflows need a service reachable from the remote session.
- watermarks-remover hooks cannot modify assistant chat messages before display.
- watermarks-remover chat-only cleaning remains model-instruction-based and best-effort.
watermarks-remover FAQ
What is watermarks-remover?
watermarks-remover is an open-source agent-skill and Python-service project for removing AI provenance marks from content users own. The documented targets include invisible text characters, statistical text watermarks, and metadata such as C2PA, EXIF, XMP, and document properties.
How is watermarks-remover licensed and accessed?
watermarks-remover is available from a public GitHub repository under the MIT license. Users can install skills through repository scripts, install the Claude Code plugin through its marketplace, or create a skill bundle for upload to supported account-based environments.
Which AI assistants does watermarks-remover support?
watermarks-remover documents installation workflows for Claude Code, Codex, Cursor, and Cowork, along with Grok setup instructions. The repository also describes skill uploads for claude.ai and service-reachability requirements for cloud sessions and routines.
Does watermarks-remover need a local service?
watermarks-remover requires a reachable HTTP service for the main remove-ai-marks skill, but the service does not have to run on the agent host. The separate clean-user-facing-text skill is self-contained and does not require the service.
Can watermarks-remover automatically clean assistant chat messages?
watermarks-remover cannot use hooks to rewrite assistant chat messages before users read them. The deterministic hook workflow covers files written by the agent, while text that exists only in the chat transcript depends on the instruction-based skill workflow and remains best-effort.
What files can watermarks-remover clean?
watermarks-remover lists support for image, document, web, video, and audio formats, including PNG, JPEG, PDF, DOCX, XLSX, PPTX, HTML, Markdown, MP4, WAV, MP3, and FLAC. File cleaning targets provenance and metadata such as C2PA, EXIF, XMP, and document properties.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




