How do I install the Opennomark skill?

Run npx skillmds@latest add nanmicoder/opennomark in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Opennomark safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: makes network calls. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Opennomark?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Opennomark free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Opennomark?

nanmicoder (@nanmicoder) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Opennomark by nanmicoder — Agent Skill

Name: Opennomark
Rating: 4.2 (5 reviews)
Author: nanmicoder

Opennomark

Remove visible watermarks from user-provided or authorized images, including Gemini sparkle marks, AI-generator corner signatures, and compact high-confidence text watermarks. Use when a user asks to remove, clean, erase, or batch-process visible image watermarks, including requests containing "remove watermark", "去水印", "去除水印", or "水印去除".

nanmicoder Updated Sep 17, 2026

File contents

OpenNoMark

Use the OpenNoMark CLI as the deterministic execution layer. Preserve every source image and write cleaned files to a separate output directory.

Run

Confirm that the inputs are user-provided or that the user is authorized to edit them.
Resolve every requested image path. Accept PNG, JPEG, and WebP files, directories, and glob patterns.
Prefer an installed opennomark executable. Otherwise run it from the official repository with uvx:

opennomark <inputs...> --output <output-directory> --json

uvx --from git+https://github.com/NanmiCoder/OpenNoMark.git opennomark <inputs...> --output <output-directory> --json

Set --device cuda only when CUDA is available. Allow automatic device selection otherwise. Do not force mps; LaMa safely falls back to CPU on Apple Silicon.
Parse the JSON response. Treat exit code 0 with status: "ok" as success, 1 as invalid or missing input, and 2 as a model-loading or processing failure.
Report the absolute paths in results[].output, grouped by cleaned and no_watermark. Treat partial as a failure that requires review: it can mean unresolved residual evidence or that dense/tiled candidates exceeded the automatic-removal safety budget. Do not claim an image was cleaned when its status says otherwise.
Inspect at least one output when visual inspection is available. For batch requests, inspect representative outputs and any low-confidence or unexpected result.

The first run can take longer because model weights are downloaded and cached. Keep the command running while downloads or inference are active.

Useful options

# Batch a directory without overwriting the originals
opennomark /path/to/images --output /path/to/cleaned --json

# Save detector boxes and masks when diagnosing a miss
opennomark /path/to/image.png --output /path/to/debug --debug --json

# Inspect the installed CLI version
opennomark --version

Use --debug only for diagnosis because it creates extra artifacts beside the cleaned output.

---name: opennomarkdescription: Remove visible watermarks from user-provided or authorized images, including Gemini sparkle marks, AI-generator corner signatures, and compact high-confidence text watermarks. Use when a user asks to remove, clean, erase, or batch-process visible image watermarks, including requests containing "remove watermark", "去水印", "去除水印", or "水印去除".---# OpenNoMarkUse the OpenNoMark CLI as the deterministic execution layer. Preserve every source image and write cleaned files to a separate output directory.## Run1. Confirm that the inputs are user-provided or that the user is authorized to edit them.2. Resolve every requested image path. Accept PNG, JPEG, and WebP files, directories, and glob patterns.3. Prefer an installed `opennomark` executable. Otherwise run it from the official repository with `uvx`:```bashopennomark <inputs...> --output <output-directory> --json``````bashuvx --from git+https://github.com/NanmiCoder/OpenNoMark.git opennomark <inputs...> --output <output-directory> --json```4. Set `--device cuda` only when CUDA is available. Allow automatic device selection otherwise. Do not force `mps`; LaMa safely falls back to CPU on Apple Silicon.5. Parse the JSON response. Treat exit code `0` with `status: "ok"` as success, `1` as invalid or missing input, and `2` as a model-loading or processing failure.6. Report the absolute paths in `results[].output`, grouped by `cleaned` and `no_watermark`. Treat `partial` as a failure that requires review: it can mean unresolved residual evidence or that dense/tiled candidates exceeded the automatic-removal safety budget. Do not claim an image was cleaned when its status says otherwise.7. Inspect at least one output when visual inspection is available. For batch requests, inspect representative outputs and any low-confidence or unexpected result.The first run can take longer because model weights are downloaded and cached. Keep the command running while downloads or inference are active.## Useful options```bash# Batch a directory without overwriting the originalsopennomark /path/to/images --output /path/to/cleaned --json# Save detector boxes and masks when diagnosing a missopennomark /path/to/image.png --output /path/to/debug --debug --json# Inspect the installed CLI versionopennomark --version```Use `--debug` only for diagnosis because it creates extra artifacts beside the cleaned output.

Frequently asked questions